fix: reject class-typed embedders at serialize time, drop unused validator

* feat: add Skills Repository — registry, cache, CLI, and SDK integration

Adds a Skills Repository feature allowing users to publish, install,
and use skills from the CrewAI registry with @org/skill-name refs.

## What's New

### SDK (lib/crewai/)
- SkillFrontmatter: added optional 'version' field (backward compatible)
- SkillCacheManager: manages ~/.crewai/skills/{org}/{name}/ with
  .crewai_meta.json tracking, path-traversal-safe tar extraction
- SkillRegistry: parse @org/skill-name refs, local-first resolution
  (./skills/ > cache > download), interactive prompt on first use,
  CI-mode guard (CREWAI_NONINTERACTIVE/CI env vars)
- Agent.skills and Crew.skills widened to accept str refs (@org/name)
- set_skills() resolves registry refs with org-prefixed dedup keys
- New events: SkillDownloadStartedEvent, SkillDownloadCompletedEvent

### CLI (lib/cli/)
- crewai skill create <name> — context-aware (project vs standalone)
- crewai skill install @org/name — downloads to ./skills/ or cache
- crewai skill publish — ZIP + upload to org registry
- crewai skill list — show installed skills

### PlusAPI (lib/crewai-core/)
- Added SKILLS_RESOURCE, get_skill(), publish_skill(), list_skills()

### Scaffolding
- crew and flow templates now include skills/ directory

### Tests
- 91 SDK skill tests + 15 CLI skill tests, all passing

* fix: address all CI failures and CodeRabbit review comments

Lint:
- Remove unused imports (click, pytest, json)
- Replace try-except-pass with logging (S110)
- Fix unprotected zipfile.extractall (S202)

Security:
- Path traversal: startswith → is_relative_to for tar extraction
- Add path traversal protection to ZIP extraction via _safe_extract_zip
- Both cache.py and CLI main.py hardened

Type checker:
- Fix import path: crewai.events.event_bus (not crewai_event_bus)
- Remove unused type: ignore comments
- Fix type mismatches in set_skills() variable types

Code quality:
- Fix f-string interpolation in SkillNotCachedError
- Use ValidationError instead of Exception in test

* style: ruff format + autofix remaining lint errors

* refactor: reuse SDK parser and SkillCacheManager in CLI

- _parse_frontmatter() now delegates to crewai.skills.parser.parse_frontmatter
  when available, with a minimal fallback for CLI-only installs
- install() global cache path now reuses SkillCacheManager.store() instead
  of duplicating metadata writing logic

* refactor: add _print_current_organization to SkillCommand (matches ToolCommand pattern)

* fix: write .crewai_meta.json in fallback install path

CodeRabbit caught that the ImportError fallback in install() didn't write
cache metadata, making skills invisible to 'crewai skill list'.

* fix: tighten @org/name ref validation to prevent path traversal

Reject refs with multiple slashes (@org/a/b), dot segments (@../skill),
or leading dots in org/name. Applied to both CLI install() and SDK
parse_registry_ref() so the contract is enforced consistently.

* fix: update test assertions to match tightened error messages

* fix: align OSS client with AMP API contract

- download_skill(): fetch download_url (presigned URL) instead of
  expecting inline base64. Falls back to 'file' field for compat.
- Read 'latest_version' field, fall back to 'version'
- Same fixes applied to CLI install() command

* fix: publish as tar.gz (matches AMP content_type validation) + add zip fallback to SDK cache

CLI publish:
- _build_skill_zip → _build_skill_tarball (tar.gz format)
- Content type: application/x-gzip (matches SkillVersion validation)

SDK cache:
- store() now tries tar.gz first, falls back to zip extraction
- Added _safe_extract_zip for path-traversal-safe zip handling
- Both formats work for download/install regardless of server format

---------

Co-authored-by: João Moura <joaomdmoura@gmail.com>

.github/workflows/build-uv-cache.yml

@@ -26,7 +26,7 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: ${{ matrix.python-version }}

.github/workflows/generate-tool-specs.yml

@@ -22,10 +22,10 @@ jobs:
     steps:
       - name: Generate GitHub App token
         id: app-token
-        uses: tibdex/github-app-token@v2
+        uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 # v3.2.0
         with:
-          app_id: ${{ secrets.CREWAI_TOOL_SPECS_APP_ID }}
-          private_key: ${{ secrets.CREWAI_TOOL_SPECS_PRIVATE_KEY }}
+          app-id: ${{ secrets.CREWAI_TOOL_SPECS_APP_ID }}
+          private-key: ${{ secrets.CREWAI_TOOL_SPECS_PRIVATE_KEY }}
 
       - name: Checkout code
         uses: actions/checkout@v4
@@ -34,7 +34,7 @@ jobs:
           token: ${{ steps.app-token.outputs.token }}
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: "3.12"

.github/workflows/linter.yml

@@ -13,7 +13,7 @@ jobs:
       code: ${{ steps.filter.outputs.code }}
     steps:
       - uses: actions/checkout@v4
-      - uses: dorny/paths-filter@v3
+      - uses: dorny/paths-filter@d1c1ffe0248fe513906c8e24db8ea791d46f8590 # v3
         id: filter
         with:
           filters: |
@@ -41,7 +41,7 @@ jobs:
             uv-main-py3.11-
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: "3.11"

.github/workflows/nightly.yml

@@ -5,6 +5,10 @@ on:
     - cron: '0 6 * * *' # daily at 6am UTC
   workflow_dispatch:
 
+concurrency:
+  group: nightly-publish
+  cancel-in-progress: false
+
 jobs:
   check:
     name: Check for new commits
@@ -18,10 +22,11 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: Check for commits in last 24h
+      - name: Check for recent commits
         id: check
         run: |
-          RECENT=$(git log --since="24 hours ago" --oneline | head -1)
+          # 25h window absorbs cron-vs-commit timing skew at the boundary.
+          RECENT=$(git log --since="25 hours ago" --oneline | head -1)
           if [ -n "$RECENT" ]; then
             echo "has_changes=true" >> "$GITHUB_OUTPUT"
           else
@@ -38,34 +43,42 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.12"
-
       - name: Install uv
-        uses: astral-sh/setup-uv@v4
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
+        with:
+          version: "0.11.3"
+          python-version: "3.12"
+          enable-cache: false
 
       - name: Stamp nightly versions
         run: |
           DATE=$(date +%Y%m%d)
+
+          # All workspace packages share the same base version and are released together.
+          BASE=$(python -c "
+          import re
+          print(re.search(r'__version__\s*=\s*\"(.*?)\"', open('lib/crewai/src/crewai/__init__.py').read()).group(1))
+          ")
+          NIGHTLY="${BASE}.dev${DATE}"
+          echo "Nightly version: ${NIGHTLY}"
+
           for init_file in \
             lib/crewai/src/crewai/__init__.py \
+            lib/crewai-core/src/crewai_core/__init__.py \
             lib/crewai-tools/src/crewai_tools/__init__.py \
-            lib/crewai-files/src/crewai_files/__init__.py; do
-            CURRENT=$(python -c "
-          import re
-          text = open('$init_file').read()
-          print(re.search(r'__version__\s*=\s*\"(.*?)\"\s*$', text, re.MULTILINE).group(1))
-          ")
-            NIGHTLY="${CURRENT}.dev${DATE}"
+            lib/crewai-files/src/crewai_files/__init__.py \
+            lib/cli/src/crewai_cli/__init__.py; do
             sed -i "s/__version__ = .*/__version__ = \"${NIGHTLY}\"/" "$init_file"
-            echo "$init_file: $CURRENT -> $NIGHTLY"
+            echo "Stamped $init_file -> $NIGHTLY"
           done
 
-          # Update cross-package dependency pins to nightly versions
-          sed -i "s/\"crewai-tools==[^\"]*\"/\"crewai-tools==${NIGHTLY}\"/" lib/crewai/pyproject.toml
+          # Update all cross-package dependency pins to the nightly version.
           sed -i "s/\"crewai==[^\"]*\"/\"crewai==${NIGHTLY}\"/" lib/crewai-tools/pyproject.toml
+          sed -i "s/\"crewai-core==[^\"]*\"/\"crewai-core==${NIGHTLY}\"/" lib/crewai/pyproject.toml
+          sed -i "s/\"crewai-cli==[^\"]*\"/\"crewai-cli==${NIGHTLY}\"/" lib/crewai/pyproject.toml
+          sed -i "s/\"crewai-tools==[^\"]*\"/\"crewai-tools==${NIGHTLY}\"/" lib/crewai/pyproject.toml
+          sed -i "s/\"crewai-files==[^\"]*\"/\"crewai-files==${NIGHTLY}\"/" lib/crewai/pyproject.toml
+          sed -i "s/\"crewai-core==[^\"]*\"/\"crewai-core==${NIGHTLY}\"/" lib/cli/pyproject.toml
           echo "Updated cross-package dependency pins to ${NIGHTLY}"
 
       - name: Build packages
@@ -85,15 +98,12 @@ jobs:
     runs-on: ubuntu-latest
     environment:
       name: pypi
-      url: https://pypi.org/p/crewai
     permissions:
       id-token: write
       contents: read
     steps:
-      - uses: actions/checkout@v4
-
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: "3.12"
@@ -116,7 +126,8 @@ jobs:
               continue
             fi
             echo "Publishing $package"
-            if ! uv publish "$package"; then
+            # --check-url skips files already on PyPI so manual re-runs on the same day are idempotent.
+            if ! uv publish --check-url https://pypi.org/simple/ "$package"; then
               echo "Failed to publish $package"
               failed=1
             fi

.github/workflows/pr-size.yml

@@ -10,7 +10,7 @@ jobs:
     permissions:
       pull-requests: write
     steps:
-      - uses: codelytv/pr-size-labeler@v1
+      - uses: codelytv/pr-size-labeler@095a41fca88b8764fd9e008ad269bcdb82bb38b9 # v1
         with:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           xs_label: "size/XS"

.github/workflows/pr-title.yml

@@ -12,7 +12,7 @@ jobs:
   pr-title:
     runs-on: ubuntu-latest
     steps:
-      - uses: amannn/action-semantic-pull-request@v5
+      - uses: amannn/action-semantic-pull-request@e32d7e603df1aa1ba07e981f2a23455dee596825 # v5
         continue-on-error: true
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/publish.yml

@@ -34,7 +34,7 @@ jobs:
           python-version: "3.12"
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v4
+        uses: astral-sh/setup-uv@38f3f104447c67c051c4a08e39b64a148898af3a # v4
 
       - name: Build packages
         run: |
@@ -63,7 +63,7 @@ jobs:
           ref: ${{ inputs.release_tag || github.ref }}
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: "3.12"
@@ -159,7 +159,7 @@ jobs:
 
       - name: Notify Slack
         if: success()
-        uses: slackapi/slack-github-action@v2.1.0
+        uses: slackapi/slack-github-action@b0fa283ad8fea605de13dc3f449259339835fc52 # v2.1.0
         with:
           webhook: ${{ secrets.SLACK_WEBHOOK_URL }}
           webhook-type: incoming-webhook

.github/workflows/tests.yml

@@ -13,7 +13,7 @@ jobs:
       code: ${{ steps.filter.outputs.code }}
     steps:
       - uses: actions/checkout@v4
-      - uses: dorny/paths-filter@v3
+      - uses: dorny/paths-filter@d1c1ffe0248fe513906c8e24db8ea791d46f8590 # v3
         id: filter
         with:
           filters: |
@@ -51,7 +51,7 @@ jobs:
             uv-main-py${{ matrix.python-version }}-
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: ${{ matrix.python-version }}

.github/workflows/type-checker.yml

@@ -13,7 +13,7 @@ jobs:
       code: ${{ steps.filter.outputs.code }}
     steps:
       - uses: actions/checkout@v4
-      - uses: dorny/paths-filter@v3
+      - uses: dorny/paths-filter@d1c1ffe0248fe513906c8e24db8ea791d46f8590 # v3
         id: filter
         with:
           filters: |
@@ -48,7 +48,7 @@ jobs:
             uv-main-py${{ matrix.python-version }}-
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: ${{ matrix.python-version }}

.github/workflows/update-test-durations.yml

@@ -38,7 +38,7 @@ jobs:
             uv-main-py${{ matrix.python-version }}-
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: ${{ matrix.python-version }}

.github/workflows/vulnerability-scan.yml

@@ -31,7 +31,7 @@ jobs:
             uv-main-py3.11-
 
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@d0cc045d04ccac9d8b7881df0226f9e82c39688e # v6
         with:
           version: "0.11.3"
           python-version: "3.11"
@@ -46,9 +46,39 @@ jobs:
       - name: Run pip-audit
         run: |
           uv run pip-audit --desc --aliases --skip-editable --format json --output pip-audit-report.json \
-            --ignore-vuln CVE-2026-3219
+            --ignore-vuln PYSEC-2024-277 \
+            --ignore-vuln PYSEC-2026-89 \
+            --ignore-vuln PYSEC-2026-97 \
+            --ignore-vuln PYSEC-2025-148 \
+            --ignore-vuln PYSEC-2025-183 \
+            --ignore-vuln PYSEC-2025-189 \
+            --ignore-vuln PYSEC-2025-190 \
+            --ignore-vuln PYSEC-2025-191 \
+            --ignore-vuln PYSEC-2025-192 \
+            --ignore-vuln PYSEC-2025-193 \
+            --ignore-vuln PYSEC-2025-194 \
+            --ignore-vuln PYSEC-2025-195 \
+            --ignore-vuln PYSEC-2025-196 \
+            --ignore-vuln PYSEC-2025-197 \
+            --ignore-vuln PYSEC-2025-210 \
+            --ignore-vuln PYSEC-2026-139 \
+            --ignore-vuln PYSEC-2025-211 \
+            --ignore-vuln PYSEC-2025-212 \
+            --ignore-vuln PYSEC-2025-213 \
+            --ignore-vuln PYSEC-2025-214 \
+            --ignore-vuln PYSEC-2025-215 \
+            --ignore-vuln PYSEC-2025-216 \
+            --ignore-vuln PYSEC-2025-217 \
+            --ignore-vuln PYSEC-2025-218
         # Ignored CVEs:
-        #   CVE-2026-3219   - pip 26.0.1 (GHSA-58qw-9mgm-455v): no fix available, archive handling issue
+        #   PYSEC-2024-277      - joblib 1.5.3: disputed; NumpyArrayWrapper only used with trusted caches
+        #   PYSEC-2026-89       - markdown 3.10.2: DoS via malformed HTML; fix 3.8.1 — already past, advisory range is stale
+        #   PYSEC-2026-97       - nltk 3.9.4: arbitrary file read in filestring(); no fix available
+        #   PYSEC-2025-148      - onnx 1.21.0: path traversal in save_external_data; no fix available
+        #   PYSEC-2025-183      - pyjwt 2.12.1: disputed weak-encryption claim; key length is application-chosen
+        #   PYSEC-2025-189..197 - torch 2.11.0: memory-corruption/DoS in functions only reachable via untrusted models; no fix available
+        #   PYSEC-2025-210, PYSEC-2026-139 - torch 2.11.0: profiler/deserialization issues; no fix available
+        #   PYSEC-2025-211..218 - transformers 5.5.4: deserialization/code injection via malicious model checkpoints; no fix available
         continue-on-error: true
 
       - name: Display results

docs/ar/changelog.mdx

@@ -4,6 +4,137 @@ description: "تحديثات المنتج والتحسينات وإصلاحات
 icon: "clock"
 mode: "wide"
 ---
+<Update label="19 مايو 2026">
+  ## v1.14.5
+
+  [عرض الإصدار على GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5)
+
+  ## ما الذي تغير
+
+  ### الميزات
+  - إلغاء استخدام `CrewAgentExecutor`، وتعيين وكلاء الطاقم الافتراضيين إلى `AgentExecutor`
+  - تحسين أدوات صندوق الرمل Daytona
+  - إضافة معلمة بدء `restore_from_state_id`
+  - إضافة تسليط الضوء على `ExaSearchTool`، وإعادة تسميته من `EXASearchTool`
+
+  ### إصلاحات الأخطاء
+  - إصلاح تسرب الذاكرة في `git.py` باستخدام `cached_property`
+  - عرض استدعاءات الأدوات المتدفقة عندما تكون `available_functions` غائبة
+  - ضمان تحميل أحداث `skills` للتتبع
+  - تصحيح مسار نقطة النهاية للحالة من `/{kickoff_id}/status` إلى `/status/{kickoff_id}`
+  - استعادة كتلة الشيفرة المفقودة في دليل التدفق الأول للغة البرتغالية (pt-BR)
+  - منع `result_as_answer` من إرجاع رسائل الخطأ أو الكتل المرتبطة كإجابة نهائية
+  - الحفاظ على مخرجات المهام عبر تفريغ الدفعات غير المتزامنة
+  - دائمًا استعادة `task.output_pydantic` في كتلة finally
+  - التعامل مع إدخال `BaseModel` في `convert_to_model`
+
+  ### الوثائق
+  - تحديث سجل التغييرات والإصدار لـ v1.14.5
+  - إضافة دليل ترقية OSS و انتقال الطاقم إلى التدفق
+  - توثيق متغيرات البيئة الإضافية لأدوات المطور
+  - إضافة وثائق لـ `TavilyGetResearch`
+
+  ### إعادة الهيكلة
+  - استخراج واجهة سطر الأوامر إلى حزمة مستقلة `crewai-cli`
+
+  ## المساهمون
+
+  @NIK-TIGER-BILL, @akaKuruma, @cgoeppinger, @github-actions[bot], @greysonlalonde, @heitorado, @irfaan101, @iris-clawd, @lorenzejay, @manisrinivasan2k1, @minasami-pr, @mislavivanda, @theCyberTech, @theishangoswami, @wishhyt
+
+</Update>
+
+<Update label="18 مايو 2026">
+  ## v1.14.5a7
+
+  [عرض الإصدار على GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a7)
+
+  ## ما الذي تغير
+
+  ### الوثائق
+  - تحديث سجل التغييرات والإصدار لـ v1.14.5a6
+
+  ### تغييرات كسرية
+  - إلغاء حقل function_calling_llm
+
+  ## المساهمون
+
+  @greysonlalonde, @heitorado
+
+</Update>
+
+<Update label="15 مايو 2026">
+  ## v1.14.5a6
+
+  [عرض الإصدار على GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a6)
+
+  ## ما الذي تغير
+
+  ### إصلاحات الأخطاء
+  - إصلاح استدعاءات الأدوات المتدفقة عندما تكون available_functions غائبة
+  - رفع اعتماد langsmith إلى الإصدار >=0.8.0 لمعالجة GHSA-3644-q5cj-c5c7
+  - حل مشاكل الأماكن الشاغرة لكتل التعليمات البرمجية غير المترجمة في وثائق البرتغالية البرازيلية
+
+  ### الوثائق
+  - إضافة وثائق لـ TavilyGetResearch
+  - تحديث سجل التغييرات والإصدار لـ v1.14.5a5
+
+  ## المساهمون
+
+  @greysonlalonde, @heitorado, @iris-clawd, @lorenzejay, @manisrinivasan2k1
+
+</Update>
+
+<Update label="13 مايو 2026">
+  ## v1.14.5a5
+
+  [عرض الإصدار على GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a5)
+
+  ## ما الذي تغير
+
+  ### الميزات
+  - إلغاء استخدام CrewAgentExecutor، وتعيين وكلاء Crew الافتراضيين إلى AgentExecutor
+  - تحسين أدوات صندوق الرمل Daytona
+
+  ### إصلاحات الأخطاء
+  - إصلاح كتلة الكود المفقودة في دليل التدفق الأول باللغة البرتغالية (pt-BR)
+  - تسجيل أخطاء المراجعة المسبقة والتقطير HITL، إضافة learn_strict
+  - تصحيح urllib3 للثغرات الأمنية
+  - تصحيح gitpython و langchain-core؛ تجاهل CVE paramiko غير المصححة
+  - تحديث جميع حزم مساحة العمل المنشورة على uv lock/sync
+
+  ### الوثائق
+  - إضافة دليل ترحيل لـ `inputs.id` إلى `restoreFromStateId`
+  - إضافة دليل ترقية OSS ودليل ترحيل crew-to-flow
+  - تحديث سجل التغييرات والإصدار لـ v1.14.5a4
+
+  ## المساهمون
+
+  @akaKuruma, @greysonlalonde, @iris-clawd, @lorenzejay, @mislavivanda
+
+</Update>
+
+<Update label="9 مايو 2026">
+  ## v1.14.5a4
+
+  [عرض الإصدار على GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a4)
+
+  ## ما الذي تغير
+
+  ### الميزات
+  - تحديث قوائم LLM
+
+  ### إصلاحات الأخطاء
+  - إصلاح مشكلة الاعتماد من خلال نقل `textual` إلى `crewai-cli` وإضافة `certifi`
+
+  ### الوثائق
+  - تحديث سجل التغييرات والإصدار لـ v1.14.5a3
+
+  ## المساهمون
+
+  @cgoeppinger, @greysonlalonde
+
+</Update>
+
 <Update label="7 مايو 2026">
   ## v1.14.5a3
 

docs/ar/concepts/flows.mdx

@@ -29,6 +29,7 @@ from crewai.flow.flow import Flow, listen, start
 from dotenv import load_dotenv
 from litellm import completion
 
+load_dotenv()
 
 class ExampleFlow(Flow):
     model = "gpt-4o-mini"

docs/ar/enterprise/guides/enable-crew-studio.mdx

@@ -146,7 +146,6 @@ Crew Studio هو طريقة مبتكرة لإنشاء طواقم وكلاء ال
 
   </Step>
 
-{" "}
 <Step title="الإجابة على الأسئلة">
   أجب على أسئلة التوضيح من مساعد الطاقم لتنقيح
   متطلباتك.
@@ -161,12 +160,10 @@ Crew Studio هو طريقة مبتكرة لإنشاء طواقم وكلاء ال
 
   </Step>
 
-{" "}
 <Step title="الموافقة أو التعديل">
   وافق على الخطة أو اطلب تغييرات إذا لزم الأمر.
 </Step>
 
-{" "}
 <Step title="التنزيل أو النشر">
   نزّل الكود للتخصيص أو انشر مباشرة على المنصة.
 </Step>

docs/ar/guides/flows/inputs-id-deprecation.mdx

@@ -0,0 +1,102 @@
+---
+title: "الانتقال من inputs.id إلى restore_from_state_id"
+description: "نقل تدفقات @persist من ترطيب inputs.id المهجور إلى حقل restore_from_state_id المدعوم"
+icon: "arrow-right-arrow-left"
+---
+
+<Warning>
+  تمرير `id` داخل `inputs` لترطيب تدفق `@persist` هو **مهجور** ومقرر إزالته في إصدار مستقبلي. البديل، `restore_from_state_id`، متاح في CrewAI **v1.14.5 وما بعده** — الخطوات أدناه تنطبق بمجرد أن تقوم بالتحديث.
+</Warning>
+
+## نظرة عامة
+
+الطريقة الموثقة لترطيب تدفق `@persist` من تنفيذ سابق هي تمرير UUID لذلك التنفيذ كـ `inputs.id`. الآن، تكشف CrewAI عن حقل مخصص، `restore_from_state_id`، الذي يقوم بنفس الترطيب دون تحميل حمولة `inputs` — ودون ربط مفتاح الترطيب بهوية التنفيذ الجديد.
+
+## الانتقال
+
+إذا كنت حالياً تبدأ تدفق `@persist` باستخدام `inputs={"id": ...}`:
+
+```python
+# مهجور
+flow = CounterFlow()
+flow.kickoff(inputs={"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv"})
+```
+
+انتقل إلى `restore_from_state_id`:
+
+```python
+# مدعوم
+flow = CounterFlow()
+flow.kickoff(restore_from_state_id="abcd1234-5678-90ef-ghij-klmnopqrstuv")
+```
+
+تتمتع الوضعيتان بمعاني سلالة مختلفة:
+
+- `inputs={"id": <uuid>}` (مهجور) — **استئناف**: تكتب الكتابات تحت المعرف المقدم، مما يمدد نفس تاريخ `flow_uuid`.
+- `restore_from_state_id=<uuid>` — **تفرع**: يترطب الحالة من اللقطة، ثم يكتب تحت `state.id` جديدة. يتم الحفاظ على تاريخ التدفق المصدر.
+
+لأغلب سيناريوهات الإنتاج — إعادة تشغيل تدفق تم تهيئته من حالة سابقة — فإن التفرع هو ما تريده. راجع [إتقان حالة التدفق](/ar/guides/flows/mastering-flow-state) للحصول على النموذج الذهني الكامل.
+
+إذا كنت تبدأ تدفقك عبر واجهة برمجة تطبيقات CrewAI AMP REST، راجع [AMP](#amp) أدناه لهجرة الحمولة المعادلة.
+
+## لماذا نقوم بإهمال `inputs.id` لـ `@persist`
+
+`inputs.id` هو حالياً الطريقة الموثقة لاستئناف تدفق `@persist` من تنفيذ سابق. المشكلة هي أن نفس UUID يقوم بوظيفتين في وقت واحد:
+
+1. **يحدد أي لقطة يترطب منها `@persist`** — تحميل الحالة المحفوظة تحت ذلك UUID.
+2. **يصبح معرف تنفيذ التدفق الجديد** (`state.id` في SDK؛ يظهر كـ `flow_id` في بعض السياقات) — كل كتابة `@persist` من هذه البداية أيضاً تقع تحت نفس UUID.
+
+هذه الوظيفة المزدوجة هي السبب الجذري للمشاكل التي يصفها هذا الدليل. لأن UUID المقدم هو أيضاً معرف التنفيذ الجديد، فإن بدايتين تمرران نفس `inputs.id` ليست تنفيذين متميزين — إنهما تشتركان في معرف، وتشاركان في سجل الاستمرارية، و(على AMP) تشتركان في صف في قائمة التنفيذات. لا توجد طريقة للقول "ترطب من هذه اللقطة، ولكن سجل هذا التشغيل بشكل منفصل" دون تقسيم المسؤوليتين.
+
+`restore_from_state_id` هو هذا الانقسام. إنه يخبر `@persist` من أي لقطة يترطب، بينما يترك التنفيذ الجديد حراً لاستلام `state.id` جديدة. لم يعد مصدر الترطيب والتشغيل المسجل نفس UUID — وهو ما تريده معظم سيناريوهات الإنتاج فعلياً.
+
+## جدول إزالة
+
+من المقرر إزالة `inputs.id` لترطيب `@persist` في إصدار مستقبلي من CrewAI. لا يوجد قطع صارم فوري — تظل التدفقات الحالية تعمل — ولكن بمجرد أن تقوم بالتحديث إلى v1.14.5 أو ما بعده، يجب أن يستخدم الكود الجديد `restore_from_state_id`، ويجب أن تهاجر التدفقات الحالية في الفرصة المناسبة التالية.
+
+## AMP
+
+إذا كنت تنشر تدفقك إلى CrewAI AMP، فإن الهجرة تمتد إلى الحمولة التي تبدأ بها المرسلة إلى طاقمك المنشور، وتظهر الأعراض المرئية لإعادة استخدام `inputs.id` على لوحة معلومات النشر. تغطي القسمان الفرعيان أدناه كلاهما.
+
+### هجرة حمولة البداية
+
+إذا كنت حالياً تبدأ تدفقاً منشوراً عن طريق تضمين `id` في `inputs`:
+
+```bash
+# مهجور
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{"inputs": {"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv", "topic": "AI Agent Frameworks"}}' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+نقل UUID إلى حقل `restoreFromStateId` في المستوى الأعلى:
+
+```bash
+# مدعوم
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{
+        "inputs": {"topic": "AI Agent Frameworks"},
+        "restoreFromStateId": "abcd1234-5678-90ef-ghij-klmnopqrstuv"
+      }' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+يجلس `restoreFromStateId` بجانب `inputs` في حمولة البداية، وليس داخلها. الآن، يحمل كائن `inputs` فقط القيم التي تستهلكها تدفقك فعلياً.
+
+### ماذا يحدث عند إعادة استخدام `inputs.id`
+
+عندما تتلقى AMP بداية لتدفق يتطابق `inputs.id` الخاص به مع تنفيذ موجود، فإنه يحل إلى السجل الموجود بدلاً من إنشاء سجل جديد. من لوحة معلومات النشر سترى:
+
+- **حالة التنفيذ** — حالة التشغيل الجديد تحل محل حالة التشغيل السابق. يمكن أن تعود تنفيذات مكتملة إلى `جارية`، أو يمكن أن تتحول تشغيلات `مكتملة` إلى `خطأ` إذا فشلت البداية الجديدة — في كلتا الحالتين، لم تعد لوحة المعلومات تعكس التشغيل الأصلي.
+- **التتبع** — تتراكم تتبعات OTel عبر البدايات لأنها تشترك في نفس معرف التنفيذ؛ تتبعات التشغيل السابق إما تُستبدل بـ، أو تُخلط مع، تشغيل الجديد. لم يعد إعادة التشغيل خطوة بخطوة يتوافق مع تنفيذ واحد.
+- **قائمة التنفيذات** — البدايات التي يجب أن تظهر كصفوف منفصلة تتقلص إلى إدخال واحد، مما يخفي التاريخ.
+
+تساعد الهجرة إلى `restoreFromStateId` في الحفاظ على كل بداية كتنفيذ خاص بها — مع حالتها الخاصة، وتتبعها، وصفها في القائمة — بينما لا تزال ترطب الحالة من تشغيل سابق.
+
+<Card title="هل تحتاج مساعدة؟" icon="headset" href="mailto:support@crewai.com">
+  اتصل بفريق الدعم لدينا إذا لم تكن متأكداً من أي وضع يحتاجه تدفقك أو واجهت مشاكل أثناء الهجرة.
+</Card>

docs/ar/guides/migration/upgrading-crewai.mdx

@@ -0,0 +1,190 @@
+---
+title: "ترقية CrewAI"
+description: "كيفية ترقية CrewAI في مشروعك والتكيّف مع التغييرات الجذرية بين الإصدارات."
+icon: "arrow-up-circle"
+---
+
+## نظرة عامة
+
+تجلب إصدارات CrewAI قدرات جديدة بانتظام. يرشدك هذا الدليل خلال الخطوات العملية للحفاظ على تثبيتك محدّثًا — سواء أداة سطر الأوامر أو البيئة الافتراضية لمشروعك.
+
+إذا كنت تبدأ من الصفر، راجع [التثبيت](/ar/installation). إذا كنت قادمًا من إطار عمل آخر، راجع [الترحيل من LangGraph](/ar/guides/migration/migrating-from-langgraph).
+
+---
+
+## الشيئان اللذان قد ترغب في ترقيتهما
+
+يوجد CrewAI في مكانين على جهازك، ويتم ترقيتهما بشكل مستقل:
+
+| ماذا | كيف يُثبَّت | كيف تتم الترقية |
+|---|---|---|
+| **أداة سطر الأوامر العامة `crewai`** | `uv tool install crewai` | `uv tool install crewai --upgrade` |
+| **بيئة venv للمشروع** (حيث يعمل الكود) | `crewai install` / `uv sync` | `uv add "crewai[...]>=X.Y.Z"` ثم `crewai install` |
+
+يمكن لهما — وغالبًا ما يحدث — أن يخرجا عن التزامن. تشغيل `crewai --version` يُظهر إصدار سطر الأوامر. تشغيل `uv pip show crewai` داخل مشروعك يُظهر إصدار venv. إذا اختلفا، فهذا طبيعي؛ ما يهم بالنسبة للكود قيد التشغيل هو إصدار venv.
+
+## لماذا لا يقوم `crewai install` وحده بالترقية
+
+`crewai install` هو غلاف رفيع حول `uv sync`. يُثبّت بالضبط ما يقوله ملف `uv.lock` الحالي — وهو **لا** يرفع أي قيود إصدار.
+
+إذا كان `pyproject.toml` يقول `crewai>=1.11.1` وقد قام ملف القفل بحلّه إلى `1.11.1`، فإن تشغيل `crewai install` سيُبقيك على `1.11.1` للأبد، حتى وإن كان الإصدار `1.14.4` متاحًا.
+
+للترقية فعلًا، عليك:
+
+1. تحديث قيد الإصدار في `pyproject.toml`
+2. إعادة حلّ ملف القفل
+3. مزامنة venv
+
+`uv add` يقوم بالثلاثة في خطوة واحدة.
+
+## كيفية ترقية مشروعك
+
+```bash
+# يرفع القيد ويعيد القفل في أمر واحد
+uv add "crewai[tools]>=1.14.4"
+
+# يزامن venv (crewai install يستدعي uv sync تحت الغطاء)
+crewai install
+
+# تحقّق
+uv pip show crewai
+# → Version: 1.14.4
+```
+
+استبدل `[tools]` بأي إضافات يستخدمها مشروعك (مثلًا `[tools,anthropic]`). تحقّق من قائمة `dependencies` في `pyproject.toml` إن لم تكن متأكدًا.
+
+<Note>
+  يحدّث `uv add` كلا من `pyproject.toml` **و** `uv.lock` بشكل ذرّي. إذا قمت بتحرير `pyproject.toml` يدويًا، فإنك لا تزال بحاجة إلى تشغيل `uv lock --upgrade-package crewai` لإعادة حلّ ملف القفل قبل أن يلتقط `crewai install` الإصدار الجديد.
+</Note>
+
+## ترقية أداة سطر الأوامر العامة
+
+أداة سطر الأوامر العامة منفصلة عن مشروعك. قم بترقيتها عبر:
+
+```bash
+uv tool install crewai --upgrade
+```
+
+إذا حذّرك الـ shell بشأن `PATH` بعد الترقية، قم بتحديثه:
+
+```bash
+uv tool update-shell
+```
+
+هذا **لا** يمسّ بيئة venv الخاصة بمشروعك — لا تزال بحاجة إلى `uv add` + `crewai install` داخل المشروع.
+
+## التحقق من تزامن الاثنين
+
+```bash
+# إصدار سطر الأوامر العام
+crewai --version
+
+# إصدار venv للمشروع
+uv pip show crewai | grep Version
+```
+
+ليس من الضروري أن يتطابقا — لكن إصدار venv للمشروع هو ما يهم لسلوك التشغيل.
+
+<Note>
+  يتطلب CrewAI `Python >=3.10, <3.14`. إذا كان `uv` مثبَّتًا مقابل مفسّر أقدم، فأعد إنشاء venv للمشروع باستخدام إصدار Python مدعوم قبل تشغيل `crewai install`.
+</Note>
+
+---
+
+## التغييرات الجذرية وملاحظات الترحيل
+
+تتطلب معظم الترقيات تعديلات صغيرة فقط. المناطق أدناه هي تلك التي تنكسر بصمت أو بتتبعات مكدّس مربكة.
+
+### مسارات الاستيراد: tools و`BaseTool`
+
+الموقع الرسمي لاستيراد الـ tools هو `crewai.tools`. لا تزال المسارات القديمة تظهر في الدروس لكن يجب تحديثها.
+
+```python
+# قبل
+from crewai_tools import BaseTool
+from crewai.agents.tools import tool
+
+# بعد
+from crewai.tools import BaseTool, tool
+```
+
+كلٌ من المُزخرف `@tool` والفئة الفرعية `BaseTool` يقعان في `crewai.tools`. `AgentFinish` والرموز الأخرى الداخلية للوكيل لم تعد جزءًا من السطح العام — إذا كنت تستوردها، فانتقل إلى event listeners أو callbacks الـ `Task` بدلًا منها.
+
+### تغييرات معاملات `Agent`
+
+```python
+from crewai import Agent
+
+agent = Agent(
+    role="Researcher",
+    goal="Find authoritative sources on {topic}",
+    backstory="You are a careful, source-driven researcher.",
+    llm="gpt-4o-mini",   # اسم نموذج كسلسلة نصية أو كائن LLM
+    verbose=True,        # bool وليس مستوى عددي صحيح
+    max_iter=15,         # تغيّر الافتراضي بين الإصدارات — حدّده بشكل صريح
+    allow_delegation=False,
+)
+```
+
+- يقبل `llm` إما اسم نموذج كسلسلة نصية (يُحلَّ عبر المزوّد المهيّأ) أو كائن `LLM` للتحكم الدقيق.
+- `verbose` هو `bool` بسيط. تمرير عدد صحيح لم يعد يبدّل مستويات السجل.
+- تغيّرت افتراضات `max_iter` بين الإصدارات. إذا توقف وكيلك بصمت عن التكرار بعد أول استدعاء tool، فحدّد `max_iter` صراحةً.
+
+### معاملات `Crew`
+
+```python
+from crewai import Crew, Process
+
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,   # أو Process.hierarchical
+    memory=True,
+    cache=True,
+    embedder={"provider": "openai", "config": {"model": "text-embedding-3-small"}},
+)
+```
+
+- يتطلب `process=Process.hierarchical` إما `manager_llm=` أو `manager_agent=`. بدون أحدهما، يرفع kickoff خطأً عند التحقّق.
+- `memory=True` مع مزوّد embedding غير افتراضي يحتاج إلى قاموس `embedder` — راجع [إعداد الذاكرة وembedder](#memory-embedder-config) أدناه.
+
+### الإخراج المُهيكل لـ `Task`
+
+استخدم `output_pydantic` أو `output_json` أو `output_file` لإلزام نتيجة المهمة بشكل مكتوب الأنواع:
+
+```python
+from pydantic import BaseModel
+from crewai import Task
+
+class Article(BaseModel):
+    title: str
+    body: str
+
+write = Task(
+    description="Write an article about {topic}",
+    expected_output="A short article with a title and body",
+    agent=writer,
+    output_pydantic=Article,        # الفئة، وليس مثيلًا منها
+    output_file="output/article.md",
+)
+```
+
+`output_pydantic` يأخذ **الفئة** نفسها. تمرير `Article(title="", body="")` خطأ شائع ويفشل بخطأ تحقّق مربك.
+
+### إعداد الذاكرة وembedder {#memory-embedder-config}
+
+إذا كان `memory=True` وأنت لا تستخدم embeddings الافتراضية الخاصة بـ OpenAI، فيجب أن تمرّر `embedder`:
+
+```python
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    memory=True,
+    embedder={
+        "provider": "ollama",
+        "config": {"model": "nomic-embed-text"},
+    },
+)
+```
+
+ضع بيانات اعتماد المزوّد المعنيّة (`OPENAI_API_KEY`, `OLLAMA_HOST`, إلخ) في ملف `.env`. مسارات تخزين الذاكرة محلية بالنسبة للمشروع افتراضيًا — احذف مجلد ذاكرة المشروع إذا غيّرت embedders، لأن الأبعاد لا تختلط.

docs/ar/learn/llm-selection-guide.mdx

@@ -802,7 +802,6 @@ The tables below show a representative sample of current top-performing models a
     Begin with well-established models like **GPT-4.1**, **Claude 3.7 Sonnet**, or **Gemini 2.0 Flash** that offer good performance across multiple dimensions and have extensive real-world validation.
   </Step>
 
-{" "}
 <Step title="Identify Specialized Needs">
   Determine if your crew has specific requirements (coding, reasoning, speed)
   that would benefit from specialized models like **Claude 4 Sonnet** for
@@ -810,7 +809,6 @@ The tables below show a representative sample of current top-performing models a
   consider fast inference providers like **Groq** alongside model selection.
 </Step>
 
-{" "}
 <Step title="Implement Multi-Model Strategy">
   Use different models for different agents based on their roles.
   High-capability models for managers and complex tasks, efficient models for

docs/ar/tools/ai-ml/daytona.mdx

@@ -13,7 +13,7 @@ The Daytona sandbox tools give CrewAI agents access to isolated, ephemeral compu
 
 - **`DaytonaExecTool`** — run any shell command inside a sandbox.
 - **`DaytonaPythonTool`** — execute a block of Python source code inside a sandbox.
-- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox.
+- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox; also supports `move`, `find` (content grep), `search` (filename glob), `chmod` (permissions), `replace` (bulk find-and-replace), and `exists`.
 
 All three tools share the same sandbox lifecycle controls, so you can mix and match them while keeping state in a single persistent sandbox.
 
@@ -55,7 +55,7 @@ from crewai_tools import DaytonaPythonTool
 tool = DaytonaPythonTool()
 result = tool.run(code="print(sum(range(10)))")
 print(result)
-# {"exit_code": 0, "result": "45\n", "artifacts": None}
+# {"exit_code": 0, "result": "45\n", "artifacts": ExecutionArtifacts(stdout="45\n", charts=[])}
 ```
 
 ### Multi-step shell session (persistent)
@@ -63,17 +63,22 @@ print(result)
 ```python Code
 from crewai_tools import DaytonaExecTool, DaytonaFileTool
 
+# Create the persistent sandbox via the first tool, then attach the second
+# tool to it so both share state (installed packages, files, env vars).
 exec_tool = DaytonaExecTool(persistent=True)
-file_tool = DaytonaFileTool(persistent=True)
-
-# Install a package, then write and run a script — all in the same sandbox
 exec_tool.run(command="pip install httpx -q")
-file_tool.run(action="write", path="/workspace/fetch.py", content="import httpx; print(httpx.get('https://httpbin.org/get').status_code)")
-exec_tool.run(command="python /workspace/fetch.py")
+file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
+
+file_tool.run(
+    action="write",
+    path="workspace/script.py",
+    content="import httpx; print(f'httpx loaded, version {httpx.__version__}')",
+)
+exec_tool.run(command="python workspace/script.py")
 ```
 
 <Note>
-Each tool instance maintains its own persistent sandbox. To share **one** sandbox across two tools, create the first tool, grab its sandbox id via `tool._persistent_sandbox.id`, and pass it to the second tool via `sandbox_id=...`.
+By default, each tool with `persistent=True` lazily creates its **own** sandbox on first use. The pattern above shares a single sandbox across multiple tools by reading the first tool's `active_sandbox_id` after a `.run()` call and passing it to the others via `sandbox_id=...`. With `persistent=False` (the default), every `.run()` call gets a fresh sandbox that's deleted at the end of that call.
 </Note>
 
 ### Attach to an existing sandbox
@@ -82,7 +87,7 @@ Each tool instance maintains its own persistent sandbox. To share **one** sandbo
 from crewai_tools import DaytonaExecTool
 
 tool = DaytonaExecTool(sandbox_id="my-long-lived-sandbox")
-result = tool.run(command="ls /workspace")
+result = tool.run(command="ls workspace")
 ```
 
 ### Custom sandbox parameters
@@ -102,6 +107,41 @@ tool = DaytonaExecTool(
 )
 ```
 
+### Searching, moving, and modifying files
+
+```python Code
+from crewai_tools import DaytonaFileTool
+
+file_tool = DaytonaFileTool(persistent=True)
+
+# Find every TODO in the source tree (grep file contents recursively)
+file_tool.run(action="find", path="workspace/src", pattern="TODO:")
+
+# Find all Python files (glob match on filenames)
+file_tool.run(action="search", path="workspace", pattern="*.py")
+
+# Make a script executable
+file_tool.run(action="chmod", path="workspace/run.sh", mode="755")
+
+# Rename or move a file
+file_tool.run(
+    action="move",
+    path="workspace/draft.md",
+    destination="workspace/final.md",
+)
+
+# Bulk find-and-replace across multiple files
+file_tool.run(
+    action="replace",
+    paths=["workspace/src/a.py", "workspace/src/b.py"],
+    pattern="old_function",
+    replacement="new_function",
+)
+
+# Quick existence check before a destructive op
+file_tool.run(action="exists", path="workspace/cache.db")
+```
+
 ### Agent integration
 
 ```python Code
@@ -121,7 +161,7 @@ coder = Agent(
 )
 
 task = Task(
-    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to /workspace/fib.py, and run it.",
+    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to workspace/fib.py, and run it.",
     expected_output="The first 10 Fibonacci numbers printed to stdout.",
     agent=coder,
 )
@@ -168,12 +208,22 @@ All three tools accept these parameters at initialization:
 
 | Parameter | Type | Required | Description |
 |-----------|------|----------|-------------|
-| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`. |
-| `path` | `str` | ✓ | Absolute path inside the sandbox. |
-| `content` | `str \| None` | | Content to write or append. Required for `append`. |
+| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`, `exists`, `move`, `find`, `search`, `chmod`, `replace`. |
+| `path` | `str \| None` | ✓ for all actions except `replace` | Absolute path inside the sandbox. |
+| `content` | `str \| None` | ✓ for `append` | Content to write or append. |
 | `binary` | `bool` | | If `True`, `content` is base64 on write; returns base64 on read. |
 | `recursive` | `bool` | | For `delete`: remove directories recursively. |
-| `mode` | `str` | | For `mkdir`: octal permission string (default `"0755"`). |
+| `mode` | `str \| None` | | For `mkdir`: octal permissions for the new directory (defaults to `"0755"`). For `chmod`: octal permissions to apply to the target. |
+| `destination` | `str \| None` | ✓ for `move` | Destination path for `move`. |
+| `pattern` | `str \| None` | ✓ for `find`, `search`, `replace` | For `find`: substring matched against file CONTENTS. For `search`: glob matched against file NAMES (e.g. `*.py`). For `replace`: text to replace inside files. |
+| `replacement` | `str \| None` | ✓ for `replace` | Replacement text for `pattern`. |
+| `paths` | `list[str] \| None` | ✓ for `replace` | List of file paths in which to replace text. |
+| `owner` | `str \| None` | | For `chmod`: new file owner. |
+| `group` | `str \| None` | | For `chmod`: new file group. |
+
+<Note>
+For `chmod`, pass at least one of `mode`, `owner`, or `group` — any field left as `None` is left unchanged on the target.
+</Note>
 
 <Tip>
 For files larger than a few KB, create the file first with `action="write"` and empty content, then send the body via multiple `action="append"` calls of ~4 KB each to stay within tool-call payload limits.

docs/en/changelog.mdx

@@ -4,6 +4,137 @@ description: "Product updates, improvements, and bug fixes for CrewAI"
 icon: "clock"
 mode: "wide"
 ---
+<Update label="May 19, 2026">
+  ## v1.14.5
+
+  [View release on GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5)
+
+  ## What's Changed
+
+  ### Features
+  - Deprecate `CrewAgentExecutor`, default Crew agents to `AgentExecutor`
+  - Improve Daytona sandbox tools
+  - Add `restore_from_state_id` kickoff parameter
+  - Add highlights to `ExaSearchTool`, rename from `EXASearchTool`
+
+  ### Bug Fixes
+  - Fix memory leak in `git.py` by using `cached_property`
+  - Surface streamed tool calls when `available_functions` is absent
+  - Ensure `skills` loading events for traces
+  - Correct status endpoint path from `/{kickoff_id}/status` to `/status/{kickoff_id}`
+  - Restore missing code block in pt-BR first-flow guide
+  - Prevent `result_as_answer` from returning hook-block or error messages as final answer
+  - Preserve task outputs across async batch flush
+  - Always restore `task.output_pydantic` in finally block
+  - Handle `BaseModel` input in `convert_to_model`
+
+  ### Documentation
+  - Update changelog and version for v1.14.5
+  - Add OSS upgrade & crew-to-flow migration guide
+  - Document additional env vars for devtools
+  - Add docs for `TavilyGetResearch`
+
+  ### Refactoring
+  - Extract CLI into standalone `crewai-cli` package
+
+  ## Contributors
+
+  @NIK-TIGER-BILL, @akaKuruma, @cgoeppinger, @github-actions[bot], @greysonlalonde, @heitorado, @irfaan101, @iris-clawd, @lorenzejay, @manisrinivasan2k1, @minasami-pr, @mislavivanda, @theCyberTech, @theishangoswami, @wishhyt
+
+</Update>
+
+<Update label="May 18, 2026">
+  ## v1.14.5a7
+
+  [View release on GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a7)
+
+  ## What's Changed
+
+  ### Documentation
+  - Update changelog and version for v1.14.5a6
+
+  ### Breaking Changes
+  - Deprecate function_calling_llm field
+
+  ## Contributors
+
+  @greysonlalonde, @heitorado
+
+</Update>
+
+<Update label="May 15, 2026">
+  ## v1.14.5a6
+
+  [View release on GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a6)
+
+  ## What's Changed
+
+  ### Bug Fixes
+  - Fix streamed tool calls when available_functions is absent
+  - Bump langsmith dependency to version >=0.8.0 to address GHSA-3644-q5cj-c5c7
+  - Resolve untranslated code block placeholders in Brazilian Portuguese documentation
+
+  ### Documentation
+  - Add documentation for TavilyGetResearch
+  - Update changelog and version for v1.14.5a5
+
+  ## Contributors
+
+  @greysonlalonde, @heitorado, @iris-clawd, @lorenzejay, @manisrinivasan2k1
+
+</Update>
+
+<Update label="May 13, 2026">
+  ## v1.14.5a5
+
+  [View release on GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a5)
+
+  ## What's Changed
+
+  ### Features
+  - Deprecate CrewAgentExecutor, default Crew agents to AgentExecutor
+  - Improve Daytona sandbox tools
+
+  ### Bug Fixes
+  - Fix missing code block in pt-BR first-flow guide
+  - Log HITL pre-review and distillation failures, add learn_strict
+  - Patch urllib3 for security vulnerabilities
+  - Patch gitpython and langchain-core; ignore unpatched paramiko CVE
+  - Refresh all published workspace packages on uv lock/sync
+
+  ### Documentation
+  - Add migration guide for `inputs.id` to `restoreFromStateId`
+  - Add OSS upgrade and crew-to-flow migration guide
+  - Update changelog and version for v1.14.5a4
+
+  ## Contributors
+
+  @akaKuruma, @greysonlalonde, @iris-clawd, @lorenzejay, @mislavivanda
+
+</Update>
+
+<Update label="May 09, 2026">
+  ## v1.14.5a4
+
+  [View release on GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a4)
+
+  ## What's Changed
+
+  ### Features
+  - Update LLM listings
+
+  ### Bug Fixes
+  - Fix dependency issue by moving `textual` to `crewai-cli` and adding `certifi`
+
+  ### Documentation
+  - Update changelog and version for v1.14.5a3
+
+  ## Contributors
+
+  @cgoeppinger, @greysonlalonde
+
+</Update>
+
 <Update label="May 07, 2026">
   ## v1.14.5a3
 

docs/en/concepts/flows.mdx

@@ -29,6 +29,7 @@ from crewai.flow.flow import Flow, listen, start
 from dotenv import load_dotenv
 from litellm import completion
 
+load_dotenv()
 
 class ExampleFlow(Flow):
     model = "gpt-4o-mini"

docs/en/enterprise/guides/enable-crew-studio.mdx

@@ -146,7 +146,6 @@ Here's a typical workflow for creating a crew with Crew Studio:
 
   </Step>
 
-{" "}
 <Step title="Answer Questions">
   Respond to clarifying questions from the Crew Assistant to refine your
   requirements.
@@ -161,12 +160,10 @@ Here's a typical workflow for creating a crew with Crew Studio:
 
   </Step>
 
-{" "}
 <Step title="Approve or Modify">
   Approve the plan or request changes if necessary.
 </Step>
 
-{" "}
 <Step title="Download or Deploy">
   Download the code for customization or deploy directly to the platform.
 </Step>

docs/en/guides/flows/inputs-id-deprecation.mdx

@@ -0,0 +1,143 @@
+---
+title: "Migrating from inputs.id to restore_from_state_id"
+description: "Move @persist flows off the deprecated inputs.id hydration onto the supported restore_from_state_id field"
+icon: "arrow-right-arrow-left"
+---
+
+<Warning>
+  Passing `id` inside `inputs` to hydrate a `@persist` flow is **deprecated** and
+  scheduled for removal in a future release. The replacement, `restore_from_state_id`,
+  is available in CrewAI **v1.14.5 and later** — the steps below apply once you
+  upgrade.
+</Warning>
+
+## Overview
+
+The documented way to hydrate a `@persist` flow from a previous execution is to pass
+that execution's UUID as `inputs.id`. CrewAI now exposes a dedicated field,
+`restore_from_state_id`, that performs the same hydration without overloading the
+`inputs` payload — and without coupling the hydration key to the new execution's
+identity.
+
+## Migration
+
+If you currently kickoff a `@persist` flow with `inputs={"id": ...}`:
+
+```python
+# Deprecated
+flow = CounterFlow()
+flow.kickoff(inputs={"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv"})
+```
+
+Switch to `restore_from_state_id`:
+
+```python
+# Supported
+flow = CounterFlow()
+flow.kickoff(restore_from_state_id="abcd1234-5678-90ef-ghij-klmnopqrstuv")
+```
+
+The two modes have different lineage semantics:
+
+- `inputs={"id": <uuid>}` (deprecated) — **resume**: writes land under the supplied
+  id, extending the same `flow_uuid` history.
+- `restore_from_state_id=<uuid>` — **fork**: hydrates state from the snapshot, then
+  writes under a fresh `state.id`. The source flow's history is preserved.
+
+For most production scenarios — re-running a flow seeded from a previous state — fork
+is what you want. See [Mastering Flow State](/en/guides/flows/mastering-flow-state)
+for the full mental model.
+
+If you kickoff your flow over the CrewAI AMP REST API, see [AMP](#amp) below for the
+equivalent payload migration.
+
+## Why we are deprecating `inputs.id` for `@persist`
+
+`inputs.id` is currently the documented way to resume a `@persist` flow from a
+previous execution. The problem is that the same UUID does two jobs at once:
+
+1. **It selects which snapshot `@persist` hydrates from** — load the state saved
+   under that UUID.
+2. **It becomes the new execution's Flow Execution ID** (`state.id` in the SDK;
+   surfaced as `flow_id` in some contexts) — every `@persist` write from this
+   kickoff also lands under that same UUID.
+
+This dual role is the root cause of the issues this guide describes. Because the
+supplied UUID is also the new execution's id, two kickoffs that pass the same
+`inputs.id` are not two distinct executions — they share an id, share a persistence
+record, and (on AMP) share a row in the executions list. There is no way to say
+"hydrate from this snapshot, but record this run separately" without splitting the
+two responsibilities.
+
+`restore_from_state_id` is that split. It tells `@persist` which snapshot to hydrate
+from, while leaving the new execution free to receive a fresh `state.id`. The
+hydration source and the recorded run are no longer the same UUID — which is what
+most production scenarios actually want.
+
+## Removal timeline
+
+`inputs.id` for `@persist` hydration is scheduled for removal in a future release of
+CrewAI. There is no immediate hard cut-off — existing flows continue to work — but
+once you upgrade to v1.14.5 or later, new code should use `restore_from_state_id`, and
+existing flows should migrate at the next convenient opportunity.
+
+## AMP
+
+If you deploy your flow to CrewAI AMP, the migration extends to the kickoff payload
+sent to your deployed crew, and the visible symptoms of reusing `inputs.id` show up
+on the deployment dashboard. The two subsections below cover both.
+
+### Migrating the kickoff payload
+
+If you currently kickoff a deployed flow by embedding `id` in `inputs`:
+
+```bash
+# Deprecated
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{"inputs": {"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv", "topic": "AI Agent Frameworks"}}' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+Move the UUID to the top-level `restoreFromStateId` field:
+
+```bash
+# Supported
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{
+        "inputs": {"topic": "AI Agent Frameworks"},
+        "restoreFromStateId": "abcd1234-5678-90ef-ghij-klmnopqrstuv"
+      }' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+`restoreFromStateId` sits next to `inputs` in the kickoff payload, not inside it. The
+`inputs` object now only carries values your flow actually consumes.
+
+### What happens when `inputs.id` is reused
+
+When AMP receives a kickoff for a flow whose `inputs.id` matches an existing
+execution, it resolves to the existing record rather than creating a new one. From
+the deployment dashboard you'll see:
+
+- **Execution status** — the new run's status overwrites the previous run's. A
+  finished execution can flip back to `running`, or a `completed` run can flip to
+  `error` if the new kickoff fails — either way the dashboard no longer reflects
+  the original run.
+- **Traces** — OTel traces stack across kickoffs because they share the same
+  execution id; the previous run's traces are either replaced by, or mixed with,
+  the new run's. A step-by-step replay no longer corresponds to a single execution.
+- **Executions list** — kickoffs that should appear as separate rows collapse into
+  a single entry, hiding history.
+
+Migrating to `restoreFromStateId` keeps every kickoff as its own execution — with
+its own status, traces, and row in the list — while still hydrating state from a
+previous run.
+
+<Card title="Need Help?" icon="headset" href="mailto:support@crewai.com">
+  Contact our support team if you're unsure which mode your flow needs or hit issues
+  during the migration.
+</Card>

docs/en/guides/flows/mastering-flow-state.mdx

@@ -313,9 +313,9 @@ flow1 = PersistentCounterFlow()
 result1 = flow1.kickoff()
 print(f"First run result: {result1}")
 
-# Second run - state is automatically loaded
+# Second run - pass the ID to load the persisted state
 flow2 = PersistentCounterFlow()
-result2 = flow2.kickoff()
+result2 = flow2.kickoff(inputs={"id": flow1.state.id})
 print(f"Second run result: {result2}")  # Will be higher due to persisted state
 ```
 

docs/en/guides/migration/upgrading-crewai.mdx

@@ -0,0 +1,190 @@
+---
+title: "Upgrading CrewAI"
+description: "How to upgrade CrewAI in your project and adapt to breaking changes between versions."
+icon: "arrow-up-circle"
+---
+
+## Overview
+
+CrewAI releases ship new capabilities regularly. This guide walks you through the practical steps to keep your installation up to date — both the CLI and your project's virtual environment.
+
+If you're starting fresh, see [Installation](/en/installation). If you're coming from another framework, see [Migrating from LangGraph](/en/guides/migration/migrating-from-langgraph).
+
+---
+
+## The Two Things You Might Want to Upgrade
+
+CrewAI lives in two places on your machine, and they upgrade independently:
+
+| What | How it's installed | How to upgrade |
+|---|---|---|
+| The **global `crewai` CLI** | `uv tool install crewai` | `uv tool install crewai --upgrade` |
+| The **project venv** (what your code runs) | `crewai install` / `uv sync` | `uv add "crewai[...]>=X.Y.Z"` then `crewai install` |
+
+These can — and often do — get out of sync. Running `crewai --version` tells you the CLI version. Running `uv pip show crewai` inside your project tells you the venv version. If they differ, that's normal; what matters for your running code is the venv version.
+
+## Why `crewai install` Alone Doesn't Upgrade
+
+`crewai install` is a thin wrapper around `uv sync`. It installs exactly what the current `uv.lock` file says — it does **not** bump any version constraints.
+
+If your `pyproject.toml` says `crewai>=1.11.1` and the lock file resolved to `1.11.1`, running `crewai install` will keep you on `1.11.1` forever, even if `1.14.4` is available.
+
+To actually upgrade, you need to:
+
+1. Update the version constraint in `pyproject.toml`
+2. Re-solve the lock file
+3. Sync the venv
+
+`uv add` does all three in one shot.
+
+## How to Upgrade Your Project
+
+```bash
+# Bump the constraint and re-lock in one command
+uv add "crewai[tools]>=1.14.4"
+
+# Sync the venv (crewai install calls uv sync under the hood)
+crewai install
+
+# Verify
+uv pip show crewai
+# → Version: 1.14.4
+```
+
+Replace `[tools]` with whatever extras your project uses (e.g. `[tools,anthropic]`). Check your `pyproject.toml` `dependencies` list if you're unsure.
+
+<Note>
+  `uv add` updates both `pyproject.toml` **and** `uv.lock` atomically. If you edit `pyproject.toml` manually, you still need to run `uv lock --upgrade-package crewai` to re-solve the lock file before `crewai install` will pick up the new version.
+</Note>
+
+## Upgrading the Global CLI
+
+The global CLI is separate from your project. Upgrade it with:
+
+```bash
+uv tool install crewai --upgrade
+```
+
+If your shell warns about `PATH` after the upgrade, refresh it:
+
+```bash
+uv tool update-shell
+```
+
+This does **not** touch your project's venv — you still need `uv add` + `crewai install` inside the project.
+
+## Verify Both Are in Sync
+
+```bash
+# Global CLI version
+crewai --version
+
+# Project venv version
+uv pip show crewai | grep Version
+```
+
+They don't need to match — but your project venv version is what matters for runtime behavior.
+
+<Note>
+  CrewAI requires `Python >=3.10, <3.14`. If `uv` was installed against an older interpreter, recreate the project venv with a supported Python before running `crewai install`.
+</Note>
+
+---
+
+## Breaking Changes & Migration Notes
+
+Most upgrades only require small adjustments. The areas below are the ones that break silently or with confusing tracebacks.
+
+### Import paths: tools and `BaseTool`
+
+The canonical import location for tools is `crewai.tools`. Older paths still surface in tutorials but should be updated.
+
+```python
+# Before
+from crewai_tools import BaseTool
+from crewai.agents.tools import tool
+
+# After
+from crewai.tools import BaseTool, tool
+```
+
+The `@tool` decorator and `BaseTool` subclass both live in `crewai.tools`. `AgentFinish` and other internal-agent symbols are no longer part of the public surface — if you were importing them, switch to event listeners or `Task` callbacks instead.
+
+### `Agent` parameter changes
+
+```python
+from crewai import Agent
+
+agent = Agent(
+    role="Researcher",
+    goal="Find authoritative sources on {topic}",
+    backstory="You are a careful, source-driven researcher.",
+    llm="gpt-4o-mini",   # string model name OR an LLM object
+    verbose=True,        # bool, not an int level
+    max_iter=15,         # default has changed across versions — set explicitly
+    allow_delegation=False,
+)
+```
+
+- `llm` accepts either a string model name (resolved via the configured provider) or an `LLM` object for fine-grained control.
+- `verbose` is a plain `bool`. Passing an integer no longer toggles log levels.
+- `max_iter` defaults have shifted between releases. If your agent silently stops looping after the first tool call, set `max_iter` explicitly.
+
+### `Crew` parameters
+
+```python
+from crewai import Crew, Process
+
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,   # or Process.hierarchical
+    memory=True,
+    cache=True,
+    embedder={"provider": "openai", "config": {"model": "text-embedding-3-small"}},
+)
+```
+
+- `process=Process.hierarchical` requires either `manager_llm=` or `manager_agent=`. Without one, kickoff raises at validation time.
+- `memory=True` with a non-default embedding provider needs an `embedder` dict — see [Memory & embedder config](#memory-embedder-config) below.
+
+### `Task` structured output
+
+Use `output_pydantic`, `output_json`, or `output_file` to coerce a task's result into a typed shape:
+
+```python
+from pydantic import BaseModel
+from crewai import Task
+
+class Article(BaseModel):
+    title: str
+    body: str
+
+write = Task(
+    description="Write an article about {topic}",
+    expected_output="A short article with a title and body",
+    agent=writer,
+    output_pydantic=Article,        # the class, NOT an instance
+    output_file="output/article.md",
+)
+```
+
+`output_pydantic` takes the **class** itself. Passing `Article(title="", body="")` is a common mistake and fails with a confusing validation error.
+
+### Memory & embedder config {#memory-embedder-config}
+
+If `memory=True` and you're not using the default OpenAI embeddings, you must pass an `embedder`:
+
+```python
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    memory=True,
+    embedder={
+        "provider": "ollama",
+        "config": {"model": "nomic-embed-text"},
+    },
+)
+```
+
+Set the relevant provider credentials (`OPENAI_API_KEY`, `OLLAMA_HOST`, etc.) in your `.env` file. Memory storage paths are project-local by default — delete the project's memory directory if you change embedders, since dimensions don't mix.

docs/en/installation.mdx

@@ -106,6 +106,9 @@ If you haven't installed `uv` yet, follow **step 1** to quickly get it set up on
         ```shell
         uv tool install crewai --upgrade
         ```
+      <Note>
+        This upgrades the **global `crewai` CLI tool** only. To upgrade the `crewai` version inside your project's virtual environment, see [Upgrading CrewAI in a project](/en/guides/migration/upgrading-crewai).
+      </Note>
       <Check>Installation successful! You're ready to create your first crew! 🎉</Check>
     </Step>
 

docs/en/learn/llm-selection-guide.mdx

@@ -805,7 +805,6 @@ The tables below show a representative sample of current top-performing models a
     Begin with well-established models like **GPT-4.1**, **Claude 3.7 Sonnet**, or **Gemini 2.0 Flash** that offer good performance across multiple dimensions and have extensive real-world validation.
   </Step>
 
-{" "}
 <Step title="Identify Specialized Needs">
   Determine if your crew has specific requirements (coding, reasoning, speed)
   that would benefit from specialized models like **Claude 4 Sonnet** for
@@ -813,7 +812,6 @@ The tables below show a representative sample of current top-performing models a
   consider fast inference providers like **Groq** alongside model selection.
 </Step>
 
-{" "}
 <Step title="Implement Multi-Model Strategy">
   Use different models for different agents based on their roles.
   High-capability models for managers and complex tasks, efficient models for

docs/en/tools/ai-ml/daytona.mdx

@@ -13,7 +13,7 @@ The Daytona sandbox tools give CrewAI agents access to isolated, ephemeral compu
 
 - **`DaytonaExecTool`** — run any shell command inside a sandbox.
 - **`DaytonaPythonTool`** — execute a block of Python source code inside a sandbox.
-- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox.
+- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox; also supports `move`, `find` (content grep), `search` (filename glob), `chmod` (permissions), `replace` (bulk find-and-replace), and `exists`.
 
 All three tools share the same sandbox lifecycle controls, so you can mix and match them while keeping state in a single persistent sandbox.
 
@@ -55,7 +55,7 @@ from crewai_tools import DaytonaPythonTool
 tool = DaytonaPythonTool()
 result = tool.run(code="print(sum(range(10)))")
 print(result)
-# {"exit_code": 0, "result": "45\n", "artifacts": None}
+# {"exit_code": 0, "result": "45\n", "artifacts": ExecutionArtifacts(stdout="45\n", charts=[])}
 ```
 
 ### Multi-step shell session (persistent)
@@ -63,17 +63,22 @@ print(result)
 ```python Code
 from crewai_tools import DaytonaExecTool, DaytonaFileTool
 
+# Create the persistent sandbox via the first tool, then attach the second
+# tool to it so both share state (installed packages, files, env vars).
 exec_tool = DaytonaExecTool(persistent=True)
-file_tool = DaytonaFileTool(persistent=True)
-
-# Install a package, then write and run a script — all in the same sandbox
 exec_tool.run(command="pip install httpx -q")
-file_tool.run(action="write", path="/workspace/fetch.py", content="import httpx; print(httpx.get('https://httpbin.org/get').status_code)")
-exec_tool.run(command="python /workspace/fetch.py")
+file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
+
+file_tool.run(
+    action="write",
+    path="workspace/script.py",
+    content="import httpx; print(f'httpx loaded, version {httpx.__version__}')",
+)
+exec_tool.run(command="python workspace/script.py")
 ```
 
 <Note>
-Each tool instance maintains its own persistent sandbox. To share **one** sandbox across two tools, create the first tool, grab its sandbox id via `tool._persistent_sandbox.id`, and pass it to the second tool via `sandbox_id=...`.
+By default, each tool with `persistent=True` lazily creates its **own** sandbox on first use. The pattern above shares a single sandbox across multiple tools by reading the first tool's `active_sandbox_id` after a `.run()` call and passing it to the others via `sandbox_id=...`. With `persistent=False` (the default), every `.run()` call gets a fresh sandbox that's deleted at the end of that call.
 </Note>
 
 ### Attach to an existing sandbox
@@ -82,7 +87,7 @@ Each tool instance maintains its own persistent sandbox. To share **one** sandbo
 from crewai_tools import DaytonaExecTool
 
 tool = DaytonaExecTool(sandbox_id="my-long-lived-sandbox")
-result = tool.run(command="ls /workspace")
+result = tool.run(command="ls workspace")
 ```
 
 ### Custom sandbox parameters
@@ -102,6 +107,41 @@ tool = DaytonaExecTool(
 )
 ```
 
+### Searching, moving, and modifying files
+
+```python Code
+from crewai_tools import DaytonaFileTool
+
+file_tool = DaytonaFileTool(persistent=True)
+
+# Find every TODO in the source tree (grep file contents recursively)
+file_tool.run(action="find", path="workspace/src", pattern="TODO:")
+
+# Find all Python files (glob match on filenames)
+file_tool.run(action="search", path="workspace", pattern="*.py")
+
+# Make a script executable
+file_tool.run(action="chmod", path="workspace/run.sh", mode="755")
+
+# Rename or move a file
+file_tool.run(
+    action="move",
+    path="workspace/draft.md",
+    destination="workspace/final.md",
+)
+
+# Bulk find-and-replace across multiple files
+file_tool.run(
+    action="replace",
+    paths=["workspace/src/a.py", "workspace/src/b.py"],
+    pattern="old_function",
+    replacement="new_function",
+)
+
+# Quick existence check before a destructive op
+file_tool.run(action="exists", path="workspace/cache.db")
+```
+
 ### Agent integration
 
 ```python Code
@@ -121,7 +161,7 @@ coder = Agent(
 )
 
 task = Task(
-    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to /workspace/fib.py, and run it.",
+    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to workspace/fib.py, and run it.",
     expected_output="The first 10 Fibonacci numbers printed to stdout.",
     agent=coder,
 )
@@ -168,12 +208,22 @@ All three tools accept these parameters at initialization:
 
 | Parameter | Type | Required | Description |
 |-----------|------|----------|-------------|
-| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`. |
-| `path` | `str` | ✓ | Absolute path inside the sandbox. |
-| `content` | `str \| None` | | Content to write or append. Required for `append`. |
+| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`, `exists`, `move`, `find`, `search`, `chmod`, `replace`. |
+| `path` | `str \| None` | ✓ for all actions except `replace` | Absolute path inside the sandbox. |
+| `content` | `str \| None` | ✓ for `append` | Content to write or append. |
 | `binary` | `bool` | | If `True`, `content` is base64 on write; returns base64 on read. |
 | `recursive` | `bool` | | For `delete`: remove directories recursively. |
-| `mode` | `str` | | For `mkdir`: octal permission string (default `"0755"`). |
+| `mode` | `str \| None` | | For `mkdir`: octal permissions for the new directory (defaults to `"0755"`). For `chmod`: octal permissions to apply to the target. |
+| `destination` | `str \| None` | ✓ for `move` | Destination path for `move`. |
+| `pattern` | `str \| None` | ✓ for `find`, `search`, `replace` | For `find`: substring matched against file CONTENTS. For `search`: glob matched against file NAMES (e.g. `*.py`). For `replace`: text to replace inside files. |
+| `replacement` | `str \| None` | ✓ for `replace` | Replacement text for `pattern`. |
+| `paths` | `list[str] \| None` | ✓ for `replace` | List of file paths in which to replace text. |
+| `owner` | `str \| None` | | For `chmod`: new file owner. |
+| `group` | `str \| None` | | For `chmod`: new file group. |
+
+<Note>
+For `chmod`, pass at least one of `mode`, `owner`, or `group` — any field left as `None` is left unchanged on the target.
+</Note>
 
 <Tip>
 For files larger than a few KB, create the file first with `action="write"` and empty content, then send the body via multiple `action="append"` calls of ~4 KB each to stay within tool-call payload limits.

docs/en/tools/search-research/overview.mdx

@@ -54,6 +54,14 @@ These tools enable your agents to search the web, research topics, and find info
     Extract structured content from web pages using the Tavily API.
   </Card>
 
+  <Card title="Tavily Research Tool" icon="flask" href="/en/tools/search-research/tavilyresearchtool">
+    Run multi-step research tasks and get cited reports using the Tavily Research API.
+  </Card>
+
+  <Card title="Tavily Get Research Tool" icon="clipboard-list" href="/en/tools/search-research/tavilygetresearchtool">
+    Retrieve the status and results of an existing Tavily research task.
+  </Card>
+
   <Card title="Arxiv Paper Tool" icon="box-archive" href="/en/tools/search-research/arxivpapertool">
     Search arXiv and optionally download PDFs.
   </Card>
@@ -76,7 +84,15 @@ These tools enable your agents to search the web, research topics, and find info
 - **Academic Research**: Find scholarly articles and technical papers
 
 ```python
-from crewai_tools import SerperDevTool, GitHubSearchTool, YoutubeVideoSearchTool, TavilySearchTool, TavilyExtractorTool
+from crewai_tools import (
+    GitHubSearchTool,
+    SerperDevTool,
+    TavilyExtractorTool,
+    TavilyGetResearchTool,
+    TavilyResearchTool,
+    TavilySearchTool,
+    YoutubeVideoSearchTool,
+)
 
 # Create research tools
 web_search = SerperDevTool()
@@ -84,11 +100,21 @@ code_search = GitHubSearchTool()
 video_research = YoutubeVideoSearchTool()
 tavily_search = TavilySearchTool()
 content_extractor = TavilyExtractorTool()
+tavily_research = TavilyResearchTool()
+tavily_get_research = TavilyGetResearchTool()
 
 # Add to your agent
 agent = Agent(
     role="Research Analyst",
-    tools=[web_search, code_search, video_research, tavily_search, content_extractor],
+    tools=[
+        web_search,
+        code_search,
+        video_research,
+        tavily_search,
+        content_extractor,
+        tavily_research,
+        tavily_get_research,
+    ],
     goal="Gather comprehensive information on any topic"
 )
 ```

docs/en/tools/search-research/tavilygetresearchtool.mdx

@@ -0,0 +1,85 @@
+---
+title: "Tavily Get Research Tool"
+description: "Retrieve the status and results of an existing Tavily research task"
+icon: "clipboard-list"
+mode: "wide"
+---
+
+The `TavilyGetResearchTool` lets CrewAI agents check an existing Tavily research task by `request_id`. Use it when a research task was started earlier and you need to retrieve its current status or final results.
+
+If you need to start a new research job, use the [Tavily Research Tool](/en/tools/search-research/tavilyresearchtool). This tool is specifically for looking up an existing Tavily research request after you already have its `request_id`.
+
+## Installation
+
+To use the `TavilyGetResearchTool`, install the `tavily-python` library alongside `crewai-tools`:
+
+```shell
+uv add 'crewai[tools]' tavily-python
+```
+
+## Environment Variables
+
+Set your Tavily API key:
+
+```bash
+export TAVILY_API_KEY='your_tavily_api_key'
+```
+
+Get an API key at [https://app.tavily.com/](https://app.tavily.com/) (sign up, then create a key).
+
+## Example Usage
+
+```python
+from crewai_tools import TavilyGetResearchTool
+
+tavily_get_research_tool = TavilyGetResearchTool()
+
+status_result = tavily_get_research_tool.run(
+    request_id="your-research-request-id"
+)
+
+print(status_result)
+```
+
+## Common Workflow
+
+Use `TavilyGetResearchTool` when your application or another service has already created a Tavily research task and saved its `request_id`.
+
+Typical cases include:
+
+- Polling for completion after kicking off research in a background job.
+- Looking up the latest status of a long-running research task.
+- Fetching final research output from a previously created Tavily request.
+
+## Configuration Options
+
+The `TavilyGetResearchTool` accepts the following argument when calling the `run` method:
+
+- `request_id` (str): **Required.** The existing Tavily research request ID to retrieve.
+
+## Async Usage
+
+Use `_arun` when your application is already running inside an async event loop:
+
+```python
+from crewai_tools import TavilyGetResearchTool
+
+tavily_get_research_tool = TavilyGetResearchTool()
+
+status_result = await tavily_get_research_tool._arun(
+    request_id="your-research-request-id"
+)
+```
+
+## Features
+
+- **Research status retrieval**: Fetch the current status of an existing Tavily research task.
+- **Result retrieval**: Return available research output once Tavily has completed the task.
+- **Sync and async**: Use either `_run`/`run` or `_arun` depending on your application's runtime.
+- **JSON output**: Returns Tavily responses as formatted JSON strings.
+
+## Response Format
+
+The tool returns a JSON string containing the current research task status and any available results from Tavily. The exact response shape depends on the task state returned by Tavily, so incomplete tasks may return status information before the final research output is available.
+
+Refer to the [Tavily API documentation](https://docs.tavily.com/) for full details on the Research API.

docs/ko/changelog.mdx

@@ -4,6 +4,137 @@ description: "CrewAI의 제품 업데이트, 개선 사항 및 버그 수정"
 icon: "clock"
 mode: "wide"
 ---
+<Update label="2026년 5월 19일">
+  ## v1.14.5
+
+  [GitHub 릴리스 보기](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5)
+
+  ## 변경 사항
+
+  ### 기능
+  - `CrewAgentExecutor` 사용 중단, 기본 Crew 에이전트를 `AgentExecutor`로 설정
+  - Daytona 샌드박스 도구 개선
+  - `restore_from_state_id` 시작 매개변수 추가
+  - `ExaSearchTool`에 하이라이트 추가, 이름을 `EXASearchTool`에서 변경
+
+  ### 버그 수정
+  - `git.py`에서 `cached_property`를 사용하여 메모리 누수 수정
+  - `available_functions`가 없을 때 스트리밍 도구 호출 표시
+  - 추적을 위한 `skills` 로딩 이벤트 보장
+  - 상태 엔드포인트 경로를 `/{kickoff_id}/status`에서 `/status/{kickoff_id}`로 수정
+  - pt-BR 첫 흐름 가이드에서 누락된 코드 블록 복원
+  - `result_as_answer`가 후크 블록이나 오류 메시지를 최종 답변으로 반환하지 않도록 방지
+  - 비동기 배치 플러시 간 작업 출력 보존
+  - 항상 finally 블록에서 `task.output_pydantic` 복원
+  - `convert_to_model`에서 `BaseModel` 입력 처리
+
+  ### 문서화
+  - v1.14.5에 대한 변경 로그 및 버전 업데이트
+  - OSS 업그레이드 및 Crew-투-흐름 마이그레이션 가이드 추가
+  - 개발 도구를 위한 추가 환경 변수 문서화
+  - `TavilyGetResearch`에 대한 문서 추가
+
+  ### 리팩토링
+  - CLI를 독립형 `crewai-cli` 패키지로 추출
+
+  ## 기여자
+
+  @NIK-TIGER-BILL, @akaKuruma, @cgoeppinger, @github-actions[bot], @greysonlalonde, @heitorado, @irfaan101, @iris-clawd, @lorenzejay, @manisrinivasan2k1, @minasami-pr, @mislavivanda, @theCyberTech, @theishangoswami, @wishhyt
+
+</Update>
+
+<Update label="2026년 5월 18일">
+  ## v1.14.5a7
+
+  [GitHub 릴리스 보기](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a7)
+
+  ## 변경 사항
+
+  ### 문서
+  - v1.14.5a6의 변경 로그 및 버전 업데이트
+
+  ### 주요 변경 사항
+  - function_calling_llm 필드 사용 중단
+
+  ## 기여자
+
+  @greysonlalonde, @heitorado
+
+</Update>
+
+<Update label="2026년 5월 15일">
+  ## v1.14.5a6
+
+  [GitHub 릴리스 보기](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a6)
+
+  ## 변경 사항
+
+  ### 버그 수정
+  - available_functions가 없을 때 스트리밍 도구 호출 수정
+  - GHSA-3644-q5cj-c5c7 문제를 해결하기 위해 langsmith 의존성을 버전 >=0.8.0으로 업데이트
+  - 브라질 포르투갈어 문서에서 번역되지 않은 코드 블록 자리 표시자 해결
+
+  ### 문서
+  - TavilyGetResearch에 대한 문서 추가
+  - v1.14.5a5에 대한 변경 로그 및 버전 업데이트
+
+  ## 기여자
+
+  @greysonlalonde, @heitorado, @iris-clawd, @lorenzejay, @manisrinivasan2k1
+
+</Update>
+
+<Update label="2026년 5월 13일">
+  ## v1.14.5a5
+
+  [GitHub 릴리스 보기](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a5)
+
+  ## 변경 사항
+
+  ### 기능
+  - CrewAgentExecutor 사용 중단, 기본 Crew 에이전트를 AgentExecutor로 설정
+  - Daytona 샌드박스 도구 개선
+
+  ### 버그 수정
+  - pt-BR 첫 번째 흐름 가이드에서 누락된 코드 블록 수정
+  - HITL 사전 검토 및 증류 실패 로그 기록, learn_strict 추가
+  - 보안 취약점을 위한 urllib3 패치
+  - gitpython 및 langchain-core 패치; 패치되지 않은 paramiko CVE 무시
+  - uv 잠금/동기화 시 모든 게시된 작업공간 패키지 새로 고침
+
+  ### 문서
+  - `inputs.id`에서 `restoreFromStateId`로의 마이그레이션 가이드 추가
+  - OSS 업그레이드 및 crew-to-flow 마이그레이션 가이드 추가
+  - v1.14.5a4의 변경 로그 및 버전 업데이트
+
+  ## 기여자
+
+  @akaKuruma, @greysonlalonde, @iris-clawd, @lorenzejay, @mislavivanda
+
+</Update>
+
+<Update label="2026년 5월 9일">
+  ## v1.14.5a4
+
+  [GitHub 릴리스 보기](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a4)
+
+  ## 변경 사항
+
+  ### 기능
+  - LLM 목록 업데이트
+
+  ### 버그 수정
+  - `textual`을 `crewai-cli`로 이동하고 `certifi`를 추가하여 의존성 문제 수정
+
+  ### 문서
+  - v1.14.5a3의 변경 로그 및 버전 업데이트
+
+  ## 기여자
+
+  @cgoeppinger, @greysonlalonde
+
+</Update>
+
 <Update label="2026년 5월 7일">
   ## v1.14.5a3
 

docs/ko/concepts/flows.mdx

@@ -29,6 +29,7 @@ from crewai.flow.flow import Flow, listen, start
 from dotenv import load_dotenv
 from litellm import completion
 
+load_dotenv()
 
 class ExampleFlow(Flow):
     model = "gpt-4o-mini"

docs/ko/enterprise/guides/enable-crew-studio.mdx

@@ -145,7 +145,6 @@ LLM 연결과 기본 설정을 구성했다면 이제 Crew Studio 사용을 시
 
   </Step>
 
-{" "}
 <Step title="질문에 답하기">
   crew assistant가 요구 사항을 구체화할 수 있도록 하는 추가 질문에 답변하세요.
 </Step>
@@ -159,12 +158,10 @@ LLM 연결과 기본 설정을 구성했다면 이제 Crew Studio 사용을 시
 
   </Step>
 
-{" "}
 <Step title="승인 또는 수정">
   계획을 승인하거나 필요하다면 변경을 요청하세요.
 </Step>
 
-{" "}
 <Step title="다운로드 또는 배포">
   사용자화를 위해 코드를 다운로드하거나 플랫폼에 직접 배포하세요.
 </Step>

docs/ko/guides/flows/inputs-id-deprecation.mdx

@@ -0,0 +1,125 @@
+---
+title: "inputs.id에서 restore_from_state_id로 마이그레이션"
+description: "더 이상 지원되지 않는 inputs.id 하이드레이션에서 지원되는 restore_from_state_id 필드로 @persist 흐름을 이동"
+icon: "arrow-right-arrow-left"
+---
+
+<Warning>
+  `inputs` 내에서 `id`를 전달하여 `@persist` 흐름을 하이드레이트하는 것은 **더 이상 지원되지 않으며**
+  향후 릴리스에서 제거될 예정입니다. 대체품인 `restore_from_state_id`는 CrewAI **v1.14.5 이상**에서 사용할 수 있으며,
+  아래 단계는 업그레이드 후 적용됩니다.
+</Warning>
+
+## 개요
+
+이전 실행에서 `@persist` 흐름을 하이드레이트하는 문서화된 방법은
+해당 실행의 UUID를 `inputs.id`로 전달하는 것입니다. CrewAI는 이제
+`inputs` 페이로드를 과부하하지 않고 동일한 하이드레이션을 수행하는 전용 필드인
+`restore_from_state_id`를 제공합니다 — 그리고 하이드레이션 키를 새로운 실행의
+정체성과 결합하지 않습니다.
+
+## 마이그레이션
+
+현재 `inputs={"id": ...}`로 `@persist` 흐름을 시작하는 경우:
+
+```python
+# 더 이상 지원되지 않음
+flow = CounterFlow()
+flow.kickoff(inputs={"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv"})
+```
+
+`restore_from_state_id`로 전환하십시오:
+
+```python
+# 지원됨
+flow = CounterFlow()
+flow.kickoff(restore_from_state_id="abcd1234-5678-90ef-ghij-klmnopqrstuv")
+```
+
+두 모드는 서로 다른 계보 의미론을 가지고 있습니다:
+
+- `inputs={"id": <uuid>}` (더 이상 지원되지 않음) — **재개**: 제공된
+  id 아래에 기록이 작성되어 동일한 `flow_uuid` 이력이 확장됩니다.
+- `restore_from_state_id=<uuid>` — **분기**: 스냅샷에서 상태를 하이드레이트한 후
+  새로운 `state.id` 아래에 기록합니다. 원본 흐름의 이력은 보존됩니다.
+
+대부분의 프로덕션 시나리오에서는 — 이전 상태에서 시드된 흐름을 다시 실행하는 경우 — 분기가
+필요합니다. 전체 정신 모델은 [Flow State 마스터링](/ko/guides/flows/mastering-flow-state)을 참조하십시오.
+
+CrewAI AMP REST API를 통해 흐름을 시작하는 경우, 아래 [AMP](#amp)에서
+동일한 페이로드 마이그레이션을 참조하십시오.
+
+## 왜 `@persist`에 대해 `inputs.id`를 더 이상 지원하지 않습니까?
+
+`inputs.id`는 현재 이전 실행에서 `@persist` 흐름을 재개하는 문서화된 방법입니다. 문제는
+동일한 UUID가 두 가지 작업을 동시에 수행한다는 것입니다:
+
+1. **어떤 스냅샷에서 `@persist`가 하이드레이트되는지를 선택합니다** — 해당 UUID 아래에 저장된 상태를 로드합니다.
+2. **새 실행의 흐름 실행 ID가 됩니다** (`state.id`는 SDK에서; 일부 컨텍스트에서는 `flow_id`로 표시됨) — 이
+   시작에서의 모든 `@persist` 기록도 동일한 UUID 아래에 작성됩니다.
+
+이 이중 역할이 이 가이드에서 설명하는 문제의 근본 원인입니다. 제공된 UUID가 새 실행의 id이기도 하므로,
+동일한 `inputs.id`를 전달하는 두 번의 시작은 두 개의 별도 실행이 아닙니다 — 그들은 id를 공유하고,
+지속성 기록을 공유하며, (AMP에서) 실행 목록에서 행을 공유합니다. "이 스냅샷에서 하이드레이트하지만,
+이 실행을 별도로 기록하십시오"라고 말할 방법이 없습니다.
+
+`restore_from_state_id`가 그 분리입니다. 이는 `@persist`에 어떤 스냅샷에서 하이드레이트할지를 알려주며,
+새 실행이 새로운 `state.id`를 받을 수 있도록 합니다. 하이드레이션 소스와 기록된 실행은 더 이상 동일한 UUID가 아닙니다 — 이는 대부분의 프로덕션 시나리오에서 실제로 원하는 것입니다.
+
+## 제거 일정
+
+`@persist` 하이드레이션을 위한 `inputs.id`는 CrewAI의 향후 릴리스에서 제거될 예정입니다. 즉각적인 강제 종료는 없으며 — 기존 흐름은 계속 작동합니다 — 하지만 v1.14.5 이상으로 업그레이드하면,
+새 코드에서는 `restore_from_state_id`를 사용해야 하며, 기존 흐름은 다음 편리한 기회에 마이그레이션해야 합니다.
+
+## AMP
+
+흐름을 CrewAI AMP에 배포하는 경우, 마이그레이션은 배포된 팀에 전송되는 시작 페이로드로 확장되며,
+`inputs.id`를 재사용하는 가시적인 증상은 배포 대시보드에 나타납니다. 아래 두 개의 하위 섹션이 이를 다룹니다.
+
+### 시작 페이로드 마이그레이션
+
+현재 `inputs`에 `id`를 포함하여 배포된 흐름을 시작하는 경우:
+
+```bash
+# 더 이상 지원되지 않음
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{"inputs": {"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv", "topic": "AI Agent Frameworks"}}' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+UUID를 최상위 `restoreFromStateId` 필드로 이동하십시오:
+
+```bash
+# 지원됨
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{
+        "inputs": {"topic": "AI Agent Frameworks"},
+        "restoreFromStateId": "abcd1234-5678-90ef-ghij-klmnopqrstuv"
+      }' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+`restoreFromStateId`는 시작 페이로드에서 `inputs` 옆에 위치하며, 내부에 있지 않습니다. 
+`inputs` 객체는 이제 흐름이 실제로 소비하는 값만 포함합니다.
+
+### `inputs.id`가 재사용될 때 발생하는 일
+
+AMP가 기존 실행과 `inputs.id`가 일치하는 흐름의 시작을 수신하면,
+새로운 기록을 생성하는 대신 기존 기록으로 해결됩니다. 배포 대시보드에서 다음을 확인할 수 있습니다:
+
+- **실행 상태** — 새로운 실행의 상태가 이전 실행의 상태를 덮어씁니다. 완료된 실행은
+  다시 `실행 중`으로 전환되거나, `완료`된 실행은 새로운 시작이 실패할 경우 `오류`로 전환될 수 있습니다 — 어쨌든 대시보드는 더 이상
+  원래 실행을 반영하지 않습니다.
+- **추적** — OTel 추적이 시작 간에 쌓이기 때문에 동일한 실행 id를 공유합니다; 이전 실행의 추적은
+  새로운 실행의 추적과 교체되거나 혼합됩니다. 단계별 재생은 더 이상 단일 실행에 해당하지 않습니다.
+- **실행 목록** — 별도의 행으로 나타나야 할 시작이 단일 항목으로 축소되어 이력을 숨깁니다.
+
+`restoreFromStateId`로 마이그레이션하면 모든 시작이 자체 실행으로 유지됩니다 — 각자의 상태, 추적 및 목록의 행을 가지며 — 여전히 이전 실행에서 상태를 하이드레이트합니다.
+
+<Card title="도움이 필요하신가요?" icon="headset" href="mailto:support@crewai.com">
+  흐름이 어떤 모드가 필요한지 확실하지 않거나 마이그레이션 중 문제가 발생하면 지원 팀에 문의하십시오.
+</Card>

docs/ko/guides/migration/upgrading-crewai.mdx

@@ -0,0 +1,190 @@
+---
+title: "CrewAI 업그레이드"
+description: "프로젝트에서 CrewAI를 업그레이드하고 버전 간 브레이킹 체인지에 적응하는 방법."
+icon: "arrow-up-circle"
+---
+
+## 개요
+
+CrewAI 릴리스는 정기적으로 새로운 기능을 제공합니다. 이 가이드는 CLI와 프로젝트의 가상 환경을 모두 최신 상태로 유지하기 위한 실용적인 단계를 안내합니다.
+
+새로 시작한다면 [설치](/ko/installation)를 참고하세요. 다른 프레임워크에서 옮겨오는 경우라면 [LangGraph에서 마이그레이션](/ko/guides/migration/migrating-from-langgraph)을 참고하세요.
+
+---
+
+## 업그레이드할 수 있는 두 가지
+
+CrewAI는 사용자의 머신에 두 곳에 존재하며, 각각 독립적으로 업그레이드됩니다:
+
+| 무엇 | 설치 방법 | 업그레이드 방법 |
+|---|---|---|
+| **전역 `crewai` CLI** | `uv tool install crewai` | `uv tool install crewai --upgrade` |
+| **프로젝트 venv** (코드가 실행되는 곳) | `crewai install` / `uv sync` | `uv add "crewai[...]>=X.Y.Z"` 후 `crewai install` |
+
+이 둘은 — 그리고 자주 — 동기화가 어긋날 수 있습니다. `crewai --version`은 CLI 버전을 알려줍니다. 프로젝트 안에서 `uv pip show crewai`를 실행하면 venv 버전을 알려줍니다. 둘이 다른 것은 정상이며, 실행 중인 코드에 중요한 것은 venv 버전입니다.
+
+## 왜 `crewai install`만으로는 업그레이드되지 않는가
+
+`crewai install`은 `uv sync`를 감싼 얇은 래퍼입니다. 현재 `uv.lock` 파일이 지시하는 것 그대로를 설치할 뿐이며 — 어떤 버전 제약도 올리지 **않습니다**.
+
+`pyproject.toml`이 `crewai>=1.11.1`이라 적혀 있고 lock 파일이 `1.11.1`로 해소되었다면, `crewai install`을 실행해도 `1.14.4`가 사용 가능하더라도 영원히 `1.11.1`에 머무릅니다.
+
+실제로 업그레이드하려면 다음을 해야 합니다:
+
+1. `pyproject.toml`의 버전 제약 업데이트
+2. lock 파일 재해소
+3. venv 동기화
+
+`uv add`는 이 세 가지를 한 번에 처리합니다.
+
+## 프로젝트 업그레이드 방법
+
+```bash
+# 제약을 올리고 lock을 다시 만드는 한 번의 명령
+uv add "crewai[tools]>=1.14.4"
+
+# venv 동기화 (crewai install은 내부적으로 uv sync를 호출)
+crewai install
+
+# 확인
+uv pip show crewai
+# → Version: 1.14.4
+```
+
+`[tools]`를 프로젝트에서 사용하는 extras로 바꾸세요 (예: `[tools,anthropic]`). 잘 모르겠다면 `pyproject.toml`의 `dependencies` 목록을 확인하세요.
+
+<Note>
+  `uv add`는 `pyproject.toml`과 `uv.lock`을 **둘 다** 원자적으로 업데이트합니다. `pyproject.toml`을 수동으로 편집하는 경우, `crewai install`이 새 버전을 가져가도록 하기 전에 `uv lock --upgrade-package crewai`를 실행해 lock 파일을 다시 해소해야 합니다.
+</Note>
+
+## 전역 CLI 업그레이드
+
+전역 CLI는 프로젝트와 분리되어 있습니다. 다음 명령으로 업그레이드하세요:
+
+```bash
+uv tool install crewai --upgrade
+```
+
+업그레이드 후 셸이 `PATH`에 대해 경고하면 새로고침하세요:
+
+```bash
+uv tool update-shell
+```
+
+이 명령은 프로젝트의 venv를 **건드리지 않습니다** — 프로젝트 내부에서는 여전히 `uv add` + `crewai install`이 필요합니다.
+
+## 둘이 동기화되었는지 확인
+
+```bash
+# 전역 CLI 버전
+crewai --version
+
+# 프로젝트 venv 버전
+uv pip show crewai | grep Version
+```
+
+둘이 일치할 필요는 없지만 — 런타임 동작에 중요한 것은 프로젝트 venv 버전입니다.
+
+<Note>
+  CrewAI는 `Python >=3.10, <3.14`를 요구합니다. `uv`가 더 오래된 인터프리터로 설치되어 있다면, `crewai install`을 실행하기 전에 지원되는 Python으로 프로젝트 venv를 다시 만드세요.
+</Note>
+
+---
+
+## 브레이킹 체인지 및 마이그레이션 노트
+
+대부분의 업그레이드는 작은 조정만 필요합니다. 아래 항목들은 조용히 깨지거나 헷갈리는 트레이스백을 내는 영역들입니다.
+
+### Import 경로: tools와 `BaseTool`
+
+tools의 정식 import 위치는 `crewai.tools`입니다. 옛 경로들이 아직 튜토리얼에 등장하지만 업데이트해야 합니다.
+
+```python
+# 이전
+from crewai_tools import BaseTool
+from crewai.agents.tools import tool
+
+# 이후
+from crewai.tools import BaseTool, tool
+```
+
+`@tool` 데코레이터와 `BaseTool` 서브클래스는 모두 `crewai.tools`에 있습니다. `AgentFinish` 등 내부 에이전트 심볼들은 더 이상 공개 표면이 아닙니다 — import 중이었다면 event listener나 `Task` 콜백으로 전환하세요.
+
+### `Agent` 파라미터 변경
+
+```python
+from crewai import Agent
+
+agent = Agent(
+    role="Researcher",
+    goal="Find authoritative sources on {topic}",
+    backstory="You are a careful, source-driven researcher.",
+    llm="gpt-4o-mini",   # 모델명 문자열 또는 LLM 객체
+    verbose=True,        # 정수 레벨이 아닌 bool
+    max_iter=15,         # 버전마다 기본값이 바뀌었음 — 명시적으로 지정
+    allow_delegation=False,
+)
+```
+
+- `llm`은 문자열 모델명(설정된 provider를 통해 해소)이나 세밀한 제어를 위한 `LLM` 객체를 받습니다.
+- `verbose`는 일반 `bool`입니다. 정수를 전달해도 더 이상 로그 레벨을 토글하지 않습니다.
+- `max_iter`의 기본값은 릴리스 사이에 변경되었습니다. 첫 tool 호출 후 에이전트가 조용히 반복을 멈춘다면 `max_iter`를 명시적으로 지정하세요.
+
+### `Crew` 파라미터
+
+```python
+from crewai import Crew, Process
+
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,   # 또는 Process.hierarchical
+    memory=True,
+    cache=True,
+    embedder={"provider": "openai", "config": {"model": "text-embedding-3-small"}},
+)
+```
+
+- `process=Process.hierarchical`은 `manager_llm=` 또는 `manager_agent=` 중 하나가 필요합니다. 둘 다 없으면 kickoff 시 검증 단계에서 오류가 발생합니다.
+- 기본이 아닌 임베딩 provider와 함께 `memory=True`를 쓰려면 `embedder` dict가 필요합니다 — 아래의 [메모리와 embedder 설정](#memory-embedder-config)을 참고하세요.
+
+### `Task` 구조화된 출력
+
+`output_pydantic`, `output_json`, 또는 `output_file`을 사용해 task 결과를 타입이 지정된 형태로 강제할 수 있습니다:
+
+```python
+from pydantic import BaseModel
+from crewai import Task
+
+class Article(BaseModel):
+    title: str
+    body: str
+
+write = Task(
+    description="Write an article about {topic}",
+    expected_output="A short article with a title and body",
+    agent=writer,
+    output_pydantic=Article,        # 인스턴스가 아닌 클래스
+    output_file="output/article.md",
+)
+```
+
+`output_pydantic`은 **클래스** 자체를 받습니다. `Article(title="", body="")`을 전달하는 것은 흔한 실수이며 헷갈리는 검증 오류로 실패합니다.
+
+### 메모리와 embedder 설정 {#memory-embedder-config}
+
+`memory=True`이고 OpenAI의 기본 임베딩을 사용하지 않는다면, `embedder`를 반드시 전달해야 합니다:
+
+```python
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    memory=True,
+    embedder={
+        "provider": "ollama",
+        "config": {"model": "nomic-embed-text"},
+    },
+)
+```
+
+해당 provider의 자격 증명(`OPENAI_API_KEY`, `OLLAMA_HOST` 등)을 `.env` 파일에 설정하세요. 메모리 저장 경로는 기본적으로 프로젝트-로컬입니다 — embedder를 바꾸면 차원이 호환되지 않으므로 프로젝트의 메모리 디렉터리를 삭제하세요.

docs/ko/learn/llm-selection-guide.mdx

@@ -797,7 +797,6 @@ LLM 선택을 최적화하고자 하는 팀을 위해 **CrewAI AMP 플랫폼**
     여러 차원에서 우수한 성능을 제공하며 실제 환경에서 광범위하게 검증된 **GPT-4.1**, **Claude 3.7 Sonnet**, **Gemini 2.0 Flash**와 같은 잘 알려진 모델부터 시작하십시오.
   </Step>
 
-{" "}
 <Step title="특화된 요구 사항 식별">
   crew에 코드 작성, reasoning, 속도 등 특정 요구가 있는지 확인하고, 이러한
   요구에 부합하는 **Claude 4 Sonnet**(개발용) 또는 **o3**(복잡한 분석용)과 같은
@@ -805,7 +804,6 @@ LLM 선택을 최적화하고자 하는 팀을 위해 **CrewAI AMP 플랫폼**
   더불어 **Groq**와 같은 빠른 추론 제공자를 고려할 수 있습니다.
 </Step>
 
-{" "}
 <Step title="다중 모델 전략 구현">
   각 에이전트의 역할에 따라 다양한 모델을 사용하세요. 관리자와 복잡한 작업에는
   고성능 모델을, 일상적 운영에는 효율적인 모델을 적용합니다.

docs/ko/tools/ai-ml/daytona.mdx

@@ -13,7 +13,7 @@ The Daytona sandbox tools give CrewAI agents access to isolated, ephemeral compu
 
 - **`DaytonaExecTool`** — run any shell command inside a sandbox.
 - **`DaytonaPythonTool`** — execute a block of Python source code inside a sandbox.
-- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox.
+- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox; also supports `move`, `find` (content grep), `search` (filename glob), `chmod` (permissions), `replace` (bulk find-and-replace), and `exists`.
 
 All three tools share the same sandbox lifecycle controls, so you can mix and match them while keeping state in a single persistent sandbox.
 
@@ -55,7 +55,7 @@ from crewai_tools import DaytonaPythonTool
 tool = DaytonaPythonTool()
 result = tool.run(code="print(sum(range(10)))")
 print(result)
-# {"exit_code": 0, "result": "45\n", "artifacts": None}
+# {"exit_code": 0, "result": "45\n", "artifacts": ExecutionArtifacts(stdout="45\n", charts=[])}
 ```
 
 ### Multi-step shell session (persistent)
@@ -63,17 +63,22 @@ print(result)
 ```python Code
 from crewai_tools import DaytonaExecTool, DaytonaFileTool
 
+# Create the persistent sandbox via the first tool, then attach the second
+# tool to it so both share state (installed packages, files, env vars).
 exec_tool = DaytonaExecTool(persistent=True)
-file_tool = DaytonaFileTool(persistent=True)
-
-# Install a package, then write and run a script — all in the same sandbox
 exec_tool.run(command="pip install httpx -q")
-file_tool.run(action="write", path="/workspace/fetch.py", content="import httpx; print(httpx.get('https://httpbin.org/get').status_code)")
-exec_tool.run(command="python /workspace/fetch.py")
+file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
+
+file_tool.run(
+    action="write",
+    path="workspace/script.py",
+    content="import httpx; print(f'httpx loaded, version {httpx.__version__}')",
+)
+exec_tool.run(command="python workspace/script.py")
 ```
 
 <Note>
-Each tool instance maintains its own persistent sandbox. To share **one** sandbox across two tools, create the first tool, grab its sandbox id via `tool._persistent_sandbox.id`, and pass it to the second tool via `sandbox_id=...`.
+By default, each tool with `persistent=True` lazily creates its **own** sandbox on first use. The pattern above shares a single sandbox across multiple tools by reading the first tool's `active_sandbox_id` after a `.run()` call and passing it to the others via `sandbox_id=...`. With `persistent=False` (the default), every `.run()` call gets a fresh sandbox that's deleted at the end of that call.
 </Note>
 
 ### Attach to an existing sandbox
@@ -82,7 +87,7 @@ Each tool instance maintains its own persistent sandbox. To share **one** sandbo
 from crewai_tools import DaytonaExecTool
 
 tool = DaytonaExecTool(sandbox_id="my-long-lived-sandbox")
-result = tool.run(command="ls /workspace")
+result = tool.run(command="ls workspace")
 ```
 
 ### Custom sandbox parameters
@@ -102,6 +107,41 @@ tool = DaytonaExecTool(
 )
 ```
 
+### Searching, moving, and modifying files
+
+```python Code
+from crewai_tools import DaytonaFileTool
+
+file_tool = DaytonaFileTool(persistent=True)
+
+# Find every TODO in the source tree (grep file contents recursively)
+file_tool.run(action="find", path="workspace/src", pattern="TODO:")
+
+# Find all Python files (glob match on filenames)
+file_tool.run(action="search", path="workspace", pattern="*.py")
+
+# Make a script executable
+file_tool.run(action="chmod", path="workspace/run.sh", mode="755")
+
+# Rename or move a file
+file_tool.run(
+    action="move",
+    path="workspace/draft.md",
+    destination="workspace/final.md",
+)
+
+# Bulk find-and-replace across multiple files
+file_tool.run(
+    action="replace",
+    paths=["workspace/src/a.py", "workspace/src/b.py"],
+    pattern="old_function",
+    replacement="new_function",
+)
+
+# Quick existence check before a destructive op
+file_tool.run(action="exists", path="workspace/cache.db")
+```
+
 ### Agent integration
 
 ```python Code
@@ -121,7 +161,7 @@ coder = Agent(
 )
 
 task = Task(
-    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to /workspace/fib.py, and run it.",
+    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to workspace/fib.py, and run it.",
     expected_output="The first 10 Fibonacci numbers printed to stdout.",
     agent=coder,
 )
@@ -168,12 +208,22 @@ All three tools accept these parameters at initialization:
 
 | Parameter | Type | Required | Description |
 |-----------|------|----------|-------------|
-| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`. |
-| `path` | `str` | ✓ | Absolute path inside the sandbox. |
-| `content` | `str \| None` | | Content to write or append. Required for `append`. |
+| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`, `exists`, `move`, `find`, `search`, `chmod`, `replace`. |
+| `path` | `str \| None` | ✓ for all actions except `replace` | Absolute path inside the sandbox. |
+| `content` | `str \| None` | ✓ for `append` | Content to write or append. |
 | `binary` | `bool` | | If `True`, `content` is base64 on write; returns base64 on read. |
 | `recursive` | `bool` | | For `delete`: remove directories recursively. |
-| `mode` | `str` | | For `mkdir`: octal permission string (default `"0755"`). |
+| `mode` | `str \| None` | | For `mkdir`: octal permissions for the new directory (defaults to `"0755"`). For `chmod`: octal permissions to apply to the target. |
+| `destination` | `str \| None` | ✓ for `move` | Destination path for `move`. |
+| `pattern` | `str \| None` | ✓ for `find`, `search`, `replace` | For `find`: substring matched against file CONTENTS. For `search`: glob matched against file NAMES (e.g. `*.py`). For `replace`: text to replace inside files. |
+| `replacement` | `str \| None` | ✓ for `replace` | Replacement text for `pattern`. |
+| `paths` | `list[str] \| None` | ✓ for `replace` | List of file paths in which to replace text. |
+| `owner` | `str \| None` | | For `chmod`: new file owner. |
+| `group` | `str \| None` | | For `chmod`: new file group. |
+
+<Note>
+For `chmod`, pass at least one of `mode`, `owner`, or `group` — any field left as `None` is left unchanged on the target.
+</Note>
 
 <Tip>
 For files larger than a few KB, create the file first with `action="write"` and empty content, then send the body via multiple `action="append"` calls of ~4 KB each to stay within tool-call payload limits.

docs/pt-BR/changelog.mdx

@@ -4,6 +4,137 @@ description: "Atualizações de produto, melhorias e correções do CrewAI"
 icon: "clock"
 mode: "wide"
 ---
+<Update label="19 mai 2026">
+  ## v1.14.5
+
+  [Ver release no GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5)
+
+  ## O que Mudou
+
+  ### Recursos
+  - Deprecar `CrewAgentExecutor`, definir agentes Crew como `AgentExecutor`
+  - Melhorar ferramentas do sandbox Daytona
+  - Adicionar parâmetro de início `restore_from_state_id`
+  - Adicionar destaques ao `ExaSearchTool`, renomeando de `EXASearchTool`
+
+  ### Correções de Bugs
+  - Corrigir vazamento de memória em `git.py` usando `cached_property`
+  - Exibir chamadas de ferramentas transmitidas quando `available_functions` está ausente
+  - Garantir eventos de carregamento de `skills` para rastros
+  - Corrigir caminho do endpoint de status de `/{kickoff_id}/status` para `/status/{kickoff_id}`
+  - Restaurar bloco de código ausente no guia de primeiro fluxo em pt-BR
+  - Impedir que `result_as_answer` retorne mensagens de bloqueio de hook ou de erro como resposta final
+  - Preservar saídas de tarefas durante o descarregamento assíncrono em lote
+  - Sempre restaurar `task.output_pydantic` no bloco finally
+  - Lidar com entrada de `BaseModel` em `convert_to_model`
+
+  ### Documentação
+  - Atualizar changelog e versão para v1.14.5
+  - Adicionar guia de migração de atualização OSS & crew-to-flow
+  - Documentar variáveis de ambiente adicionais para devtools
+  - Adicionar documentação para `TavilyGetResearch`
+
+  ### Refatoração
+  - Extrair CLI para o pacote autônomo `crewai-cli`
+
+  ## Contribuidores
+
+  @NIK-TIGER-BILL, @akaKuruma, @cgoeppinger, @github-actions[bot], @greysonlalonde, @heitorado, @irfaan101, @iris-clawd, @lorenzejay, @manisrinivasan2k1, @minasami-pr, @mislavivanda, @theCyberTech, @theishangoswami, @wishhyt
+
+</Update>
+
+<Update label="18 mai 2026">
+  ## v1.14.5a7
+
+  [Ver release no GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a7)
+
+  ## O que Mudou
+
+  ### Documentação
+  - Atualizar changelog e versão para v1.14.5a6
+
+  ### Mudanças Quebradoras
+  - Depreciar o campo function_calling_llm
+
+  ## Contributors
+
+  @greysonlalonde, @heitorado
+
+</Update>
+
+<Update label="15 mai 2026">
+  ## v1.14.5a6
+
+  [Ver release no GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a6)
+
+  ## O que mudou
+
+  ### Correções de Bugs
+  - Corrigir chamadas de ferramentas transmitidas quando available_functions está ausente
+  - Atualizar a dependência langsmith para a versão >=0.8.0 para resolver GHSA-3644-q5cj-c5c7
+  - Resolver espaços reservados de blocos de código não traduzidos na documentação em português brasileiro
+
+  ### Documentação
+  - Adicionar documentação para TavilyGetResearch
+  - Atualizar changelog e versão para v1.14.5a5
+
+  ## Contributors
+
+  @greysonlalonde, @heitorado, @iris-clawd, @lorenzejay, @manisrinivasan2k1
+
+</Update>
+
+<Update label="13 mai 2026">
+  ## v1.14.5a5
+
+  [Ver release no GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a5)
+
+  ## O Que Mudou
+
+  ### Recursos
+  - Deprecar CrewAgentExecutor, definir agentes Crew como AgentExecutor
+  - Melhorar ferramentas de sandbox Daytona
+
+  ### Correções de Bugs
+  - Corrigir bloco de código ausente no guia de primeiro fluxo em pt-BR
+  - Registrar falhas de pré-revisão e destilação HITL, adicionar learn_strict
+  - Corrigir urllib3 para vulnerabilidades de segurança
+  - Corrigir gitpython e langchain-core; ignorar CVE paramiko não corrigido
+  - Atualizar todos os pacotes de workspace publicados no bloqueio/sincronização uv
+
+  ### Documentação
+  - Adicionar guia de migração de `inputs.id` para `restoreFromStateId`
+  - Adicionar guia de atualização OSS e migração de crew para flow
+  - Atualizar changelog e versão para v1.14.5a4
+
+  ## Contribuidores
+
+  @akaKuruma, @greysonlalonde, @iris-clawd, @lorenzejay, @mislavivanda
+
+</Update>
+
+<Update label="09 mai 2026">
+  ## v1.14.5a4
+
+  [Ver release no GitHub](https://github.com/crewAIInc/crewAI/releases/tag/1.14.5a4)
+
+  ## O que Mudou
+
+  ### Funcionalidades
+  - Atualizar listagens de LLM
+
+  ### Correções de Bugs
+  - Corrigir problema de dependência movendo `textual` para `crewai-cli` e adicionando `certifi`
+
+  ### Documentação
+  - Atualizar changelog e versão para v1.14.5a3
+
+  ## Contribuidores
+
+  @cgoeppinger, @greysonlalonde
+
+</Update>
+
 <Update label="07 mai 2026">
   ## v1.14.5a3
 

docs/pt-BR/concepts/flows.mdx

@@ -24,7 +24,63 @@ Os flows permitem que você crie fluxos de trabalho estruturados e orientados po
 Vamos criar um Flow simples no qual você usará a OpenAI para gerar uma cidade aleatória em uma tarefa e, em seguida, usará essa cidade para gerar uma curiosidade em outra tarefa.
 
 ```python Code
-# (O código não é traduzido)
+
+from crewai.flow.flow import Flow, listen, start
+from dotenv import load_dotenv
+from litellm import completion
+
+load_dotenv()
+
+class ExampleFlow(Flow):
+    model = "gpt-4o-mini"
+
+    @start()
+    def generate_city(self):
+        print("Starting flow")
+        # Cada estado do flow recebe automaticamente um ID único
+        print(f"Flow State ID: {self.state['id']}")
+
+        response = completion(
+            model=self.model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": "Return the name of a random city in the world.",
+                },
+            ],
+        )
+
+        random_city = response["choices"][0]["message"]["content"]
+        # Armazena a cidade no nosso estado
+        self.state["city"] = random_city
+        print(f"Random City: {random_city}")
+
+        return random_city
+
+    @listen(generate_city)
+    def generate_fun_fact(self, random_city):
+        response = completion(
+            model=self.model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": f"Tell me a fun fact about {random_city}",
+                },
+            ],
+        )
+
+        fun_fact = response["choices"][0]["message"]["content"]
+        # Armazena a curiosidade no nosso estado
+        self.state["fun_fact"] = fun_fact
+        return fun_fact
+
+
+
+flow = ExampleFlow()
+flow.plot()
+result = flow.kickoff()
+
+print(f"Generated fun fact: {result}")
 ```
 
 Na ilustração acima, criamos um Flow simples que gera uma cidade aleatória usando a OpenAI e depois cria uma curiosidade sobre essa cidade. O Flow consiste em duas tarefas: `generate_city` e `generate_fun_fact`. A tarefa `generate_city` é o ponto de início do Flow, enquanto a tarefa `generate_fun_fact` fica escutando o resultado da tarefa `generate_city`.
@@ -56,12 +112,16 @@ O decorador `@listen()` pode ser usado de várias formas:
 1. **Escutando um Método pelo Nome**: Você pode passar o nome do método ao qual deseja escutar como string. Quando esse método concluir, o método ouvinte será chamado.
 
    ```python Code
-   # (O código não é traduzido)
+   @listen("generate_city")
+   def generate_fun_fact(self, random_city):
+       # Implementação
    ```
 
 2. **Escutando um Método Diretamente**: Você pode passar o próprio método. Quando esse método concluir, o método ouvinte será chamado.
    ```python Code
-   # (O código não é traduzido)
+   @listen(generate_city)
+   def generate_fun_fact(self, random_city):
+       # Implementação
    ```
 
 ### Saída de um Flow
@@ -76,7 +136,24 @@ Veja como acessar a saída final:
 
 <CodeGroup>
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, listen, start
+
+class OutputExampleFlow(Flow):
+    @start()
+    def first_method(self):
+        return "Output from first_method"
+
+    @listen(first_method)
+    def second_method(self, first_output):
+        return f"Second method received: {first_output}"
+
+
+flow = OutputExampleFlow()
+flow.plot("my_flow_plot")
+final_output = flow.kickoff()
+
+print("---- Final Output ----")
+print(final_output)
 ```
 
 ```text Output
@@ -97,8 +174,34 @@ Além de recuperar a saída final, você pode acessar e atualizar o estado dentr
 Veja um exemplo de como atualizar e acessar o estado:
 
 <CodeGroup>
+
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, listen, start
+from pydantic import BaseModel
+
+class ExampleState(BaseModel):
+    counter: int = 0
+    message: str = ""
+
+class StateExampleFlow(Flow[ExampleState]):
+
+    @start()
+    def first_method(self):
+        self.state.message = "Hello from first_method"
+        self.state.counter += 1
+
+    @listen(first_method)
+    def second_method(self):
+        self.state.message += " - updated by second_method"
+        self.state.counter += 1
+        return self.state.message
+
+flow = StateExampleFlow()
+flow.plot("my_flow_plot")
+final_output = flow.kickoff()
+print(f"Final Output: {final_output}")
+print("Final State:")
+print(flow.state)
 ```
 
 ```text Output
@@ -128,7 +231,33 @@ Essa abordagem oferece flexibilidade, permitindo que o desenvolvedor adicione ou
 Mesmo com estados não estruturados, os flows do CrewAI geram e mantêm automaticamente um identificador único (UUID) para cada instância de estado.
 
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, listen, start
+
+class UnstructuredExampleFlow(Flow):
+
+    @start()
+    def first_method(self):
+        # O estado inclui automaticamente um campo 'id'
+        print(f"State ID: {self.state['id']}")
+        self.state['counter'] = 0
+        self.state['message'] = "Hello from structured flow"
+
+    @listen(first_method)
+    def second_method(self):
+        self.state['counter'] += 1
+        self.state['message'] += " - updated"
+
+    @listen(second_method)
+    def third_method(self):
+        self.state['counter'] += 1
+        self.state['message'] += " - updated again"
+
+        print(f"State after third_method: {self.state}")
+
+
+flow = UnstructuredExampleFlow()
+flow.plot("my_flow_plot")
+flow.kickoff()
 ```
 
 ![Flow Visual image](/images/crewai-flow-3.png)
@@ -148,7 +277,39 @@ Ao usar modelos como o `BaseModel` da Pydantic, os desenvolvedores podem definir
 Cada estado nos flows do CrewAI recebe automaticamente um identificador único (UUID) para ajudar no rastreamento e gerenciamento. Esse ID é gerado e mantido automaticamente pelo sistema de flows.
 
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, listen, start
+from pydantic import BaseModel
+
+
+class ExampleState(BaseModel):
+    # Nota: o campo 'id' é adicionado automaticamente a todos os estados
+    counter: int = 0
+    message: str = ""
+
+
+class StructuredExampleFlow(Flow[ExampleState]):
+
+    @start()
+    def first_method(self):
+        # Acesse o ID gerado automaticamente, se necessário
+        print(f"State ID: {self.state.id}")
+        self.state.message = "Hello from structured flow"
+
+    @listen(first_method)
+    def second_method(self):
+        self.state.counter += 1
+        self.state.message += " - updated"
+
+    @listen(second_method)
+    def third_method(self):
+        self.state.counter += 1
+        self.state.message += " - updated again"
+
+        print(f"State after third_method: {self.state}")
+
+
+flow = StructuredExampleFlow()
+flow.kickoff()
 ```
 
 ![Flow Visual image](/images/crewai-flow-3.png)
@@ -182,7 +343,19 @@ O decorador @persist permite a persistência automática do estado nos flows do
 Quando aplicado no nível da classe, o decorador @persist garante a persistência automática de todos os estados dos métodos do flow:
 
 ```python
-# (O código não é traduzido)
+@persist  # Usa SQLiteFlowPersistence por padrão
+class MyFlow(Flow[MyState]):
+    @start()
+    def initialize_flow(self):
+        # Este método terá seu estado persistido automaticamente
+        self.state.counter = 1
+        print("Initialized flow. State ID:", self.state.id)
+
+    @listen(initialize_flow)
+    def next_step(self):
+        # O estado (incluindo self.state.id) é recarregado automaticamente
+        self.state.counter += 1
+        print("Flow state is persisted. Counter:", self.state.counter)
 ```
 
 ### Persistência no Nível de Método
@@ -190,7 +363,14 @@ Quando aplicado no nível da classe, o decorador @persist garante a persistênci
 Para um controle mais granular, você pode aplicar @persist em métodos específicos:
 
 ```python
-# (O código não é traduzido)
+class AnotherFlow(Flow[dict]):
+    @persist  # Persiste apenas o estado deste método
+    @start()
+    def begin(self):
+        if "runs" not in self.state:
+            self.state["runs"] = 0
+        self.state["runs"] += 1
+        print("Method-level persisted runs:", self.state["runs"])
 ```
 
 ### Forking de Estado Persistido
@@ -282,8 +462,29 @@ A arquitetura de persistência enfatiza precisão técnica e opções de persona
 A função `or_` nos flows permite escutar múltiplos métodos e acionar o método ouvinte quando qualquer um dos métodos especificados gerar uma saída.
 
 <CodeGroup>
+
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, listen, or_, start
+
+class OrExampleFlow(Flow):
+
+    @start()
+    def start_method(self):
+        return "Hello from the start method"
+
+    @listen(start_method)
+    def second_method(self):
+        return "Hello from the second method"
+
+    @listen(or_(start_method, second_method))
+    def logger(self, result):
+        print(f"Logger: {result}")
+
+
+
+flow = OrExampleFlow()
+flow.plot("my_flow_plot")
+flow.kickoff()
 ```
 
 ```text Output
@@ -302,8 +503,28 @@ A função `or_` serve para escutar vários métodos e disparar o método ouvint
 A função `and_` nos flows permite escutar múltiplos métodos e acionar o método ouvinte apenas quando todos os métodos especificados emitirem uma saída.
 
 <CodeGroup>
+
 ```python Code
-# (O código não é traduzido)
+from crewai.flow.flow import Flow, and_, listen, start
+
+class AndExampleFlow(Flow):
+
+    @start()
+    def start_method(self):
+        self.state["greeting"] = "Hello from the start method"
+
+    @listen(start_method)
+    def second_method(self):
+        self.state["joke"] = "What do computers eat? Microchips."
+
+    @listen(and_(start_method, second_method))
+    def logger(self):
+        print("---- Logger ----")
+        print(self.state)
+
+flow = AndExampleFlow()
+flow.plot()
+flow.kickoff()
 ```
 
 ```text Output
@@ -323,8 +544,42 @@ O decorador `@router()` nos flows permite definir lógica de roteamento condicio
 Você pode especificar diferentes rotas conforme a saída do método, permitindo controlar o fluxo de execução de forma dinâmica.
 
 <CodeGroup>
+
 ```python Code
-# (O código não é traduzido)
+import random
+from crewai.flow.flow import Flow, listen, router, start
+from pydantic import BaseModel
+
+class ExampleState(BaseModel):
+    success_flag: bool = False
+
+class RouterFlow(Flow[ExampleState]):
+
+    @start()
+    def start_method(self):
+        print("Starting the structured flow")
+        random_boolean = random.choice([True, False])
+        self.state.success_flag = random_boolean
+
+    @router(start_method)
+    def second_method(self):
+        if self.state.success_flag:
+            return "success"
+        else:
+            return "failed"
+
+    @listen("success")
+    def third_method(self):
+        print("Third method running")
+
+    @listen("failed")
+    def fourth_method(self):
+        print("Fourth method running")
+
+
+flow = RouterFlow()
+flow.plot("my_flow_plot")
+flow.kickoff()
 ```
 
 ```text Output
@@ -401,7 +656,105 @@ Para um guia completo sobre feedback humano em flows, incluindo feedback assínc
 Os agentes podem ser integrados facilmente aos seus flows, oferecendo uma alternativa leve às crews completas quando você precisar executar tarefas simples e focadas. Veja um exemplo de como utilizar um agente em um flow para realizar uma pesquisa de mercado:
 
 ```python
-# (O código não é traduzido)
+import asyncio
+from typing import Any, Dict, List
+
+from crewai_tools import SerperDevTool
+from pydantic import BaseModel, Field
+
+from crewai.agent import Agent
+from crewai.flow.flow import Flow, listen, start
+
+
+# Define um formato de saída estruturado
+class MarketAnalysis(BaseModel):
+    key_trends: List[str] = Field(description="List of identified market trends")
+    market_size: str = Field(description="Estimated market size")
+    competitors: List[str] = Field(description="Major competitors in the space")
+
+
+# Define o estado do flow
+class MarketResearchState(BaseModel):
+    product: str = ""
+    analysis: MarketAnalysis | None = None
+
+
+# Cria uma classe de flow
+class MarketResearchFlow(Flow[MarketResearchState]):
+    @start()
+    def initialize_research(self) -> Dict[str, Any]:
+        print(f"Starting market research for {self.state.product}")
+        return {"product": self.state.product}
+
+    @listen(initialize_research)
+    async def analyze_market(self) -> Dict[str, Any]:
+        # Cria um agente para pesquisa de mercado
+        analyst = Agent(
+            role="Market Research Analyst",
+            goal=f"Analyze the market for {self.state.product}",
+            backstory="You are an experienced market analyst with expertise in "
+            "identifying market trends and opportunities.",
+            tools=[SerperDevTool()],
+            verbose=True,
+        )
+
+        # Define a consulta de pesquisa
+        query = f"""
+        Research the market for {self.state.product}. Include:
+        1. Key market trends
+        2. Market size
+        3. Major competitors
+
+        Format your response according to the specified structure.
+        """
+
+        # Executa a análise com formato de saída estruturado
+        result = await analyst.kickoff_async(query, response_format=MarketAnalysis)
+        if result.pydantic:
+            print("result", result.pydantic)
+        else:
+            print("result", result)
+
+        # Retorna a análise para atualizar o estado
+        return {"analysis": result.pydantic}
+
+    @listen(analyze_market)
+    def present_results(self, analysis) -> None:
+        print("\nMarket Analysis Results")
+        print("=====================")
+
+        if isinstance(analysis, dict):
+            # Se recebemos um dict com a chave 'analysis', extrai o objeto de análise real
+            market_analysis = analysis.get("analysis")
+        else:
+            market_analysis = analysis
+
+        if market_analysis and isinstance(market_analysis, MarketAnalysis):
+            print("\nKey Market Trends:")
+            for trend in market_analysis.key_trends:
+                print(f"- {trend}")
+
+            print(f"\nMarket Size: {market_analysis.market_size}")
+
+            print("\nMajor Competitors:")
+            for competitor in market_analysis.competitors:
+                print(f"- {competitor}")
+        else:
+            print("No structured analysis data available.")
+            print("Raw analysis:", analysis)
+
+
+# Exemplo de uso
+async def run_flow():
+    flow = MarketResearchFlow()
+    flow.plot("MarketResearchFlowPlot")
+    result = await flow.kickoff_async(inputs={"product": "AI-powered chatbots"})
+    return result
+
+
+# Executa o flow
+if __name__ == "__main__":
+    asyncio.run(run_flow())
 ```
 
 ![Flow Visual image](/images/crewai-flow-7.png)
@@ -463,7 +816,50 @@ No arquivo `main.py`, você cria seu flow e conecta as crews. É possível defin
 Veja um exemplo de como conectar a `poem_crew` no arquivo `main.py`:
 
 ```python Code
-# (O código não é traduzido)
+#!/usr/bin/env python
+from random import randint
+
+from pydantic import BaseModel
+from crewai.flow.flow import Flow, listen, start
+from .crews.poem_crew.poem_crew import PoemCrew
+
+class PoemState(BaseModel):
+    sentence_count: int = 1
+    poem: str = ""
+
+class PoemFlow(Flow[PoemState]):
+
+    @start()
+    def generate_sentence_count(self):
+        print("Generating sentence count")
+        self.state.sentence_count = randint(1, 5)
+
+    @listen(generate_sentence_count)
+    def generate_poem(self):
+        print("Generating poem")
+        result = PoemCrew().crew().kickoff(inputs={"sentence_count": self.state.sentence_count})
+
+        print("Poem generated", result.raw)
+        self.state.poem = result.raw
+
+    @listen(generate_poem)
+    def save_poem(self):
+        print("Saving poem")
+        with open("poem.txt", "w") as f:
+            f.write(self.state.poem)
+
+def kickoff():
+    poem_flow = PoemFlow()
+    poem_flow.kickoff()
+
+
+def plot():
+    poem_flow = PoemFlow()
+    poem_flow.plot("PoemFlowPlot")
+
+if __name__ == "__main__":
+    kickoff()
+    plot()
 ```
 
 Neste exemplo, a classe `PoemFlow` define um fluxo que gera a quantidade de frases, usa a `PoemCrew` para gerar um poema e, depois, salva o poema em um arquivo. O flow inicia com o método `kickoff()`, e o gráfico é gerado pelo método `plot()`.
@@ -515,7 +911,8 @@ O CrewAI oferece duas formas práticas de gerar plots dos seus flows:
 Se estiver trabalhando diretamente com uma instância do flow, basta chamar o método `plot()` do objeto. Isso criará um arquivo HTML com o plot interativo do seu flow.
 
 ```python Code
-# (O código não é traduzido)
+# Considerando que você já tem uma instância do flow
+flow.plot("my_flow_plot")
 ```
 
 Esse comando gera um arquivo chamado `my_flow_plot.html` no diretório atual. Abra esse arquivo em um navegador para visualizar o plot interativo.

docs/pt-BR/enterprise/guides/enable-crew-studio.mdx

@@ -146,7 +146,6 @@ Veja um fluxo de trabalho típico para criação de um crew com o Crew Studio:
 
   </Step>
 
-{" "}
 <Step title="Responder Perguntas">
   Responda às perguntas de esclarecimento do Crew Assistant para refinar seus
   requisitos.
@@ -161,12 +160,10 @@ Veja um fluxo de trabalho típico para criação de um crew com o Crew Studio:
 
   </Step>
 
-{" "}
 <Step title="Aprovar ou Modificar">
   Aprove o plano ou solicite alterações, se necessário.
 </Step>
 
-{" "}
 <Step title="Baixar ou Fazer Deploy">
   Baixe o código para personalização ou faça o deploy diretamente na plataforma.
 </Step>

docs/pt-BR/guides/flows/first-flow.mdx

@@ -266,7 +266,165 @@ Nosso flow irá:
 Vamos criar nosso flow no arquivo `main.py`:
 
 ```python
-# [CÓDIGO NÃO TRADUZIDO, MANTER COMO ESTÁ]
+#!/usr/bin/env python
+import json
+import os
+from typing import List, Dict
+from pydantic import BaseModel, Field
+from crewai import LLM
+from crewai.flow.flow import Flow, listen, start
+from guide_creator_flow.crews.content_crew.content_crew import ContentCrew
+
+# Definir nossos modelos para dados estruturados
+class Section(BaseModel):
+    title: str = Field(description="Title of the section")
+    description: str = Field(description="Brief description of what the section should cover")
+
+class GuideOutline(BaseModel):
+    title: str = Field(description="Title of the guide")
+    introduction: str = Field(description="Introduction to the topic")
+    target_audience: str = Field(description="Description of the target audience")
+    sections: List[Section] = Field(description="List of sections in the guide")
+    conclusion: str = Field(description="Conclusion or summary of the guide")
+
+# Definir o estado do nosso flow
+class GuideCreatorState(BaseModel):
+    topic: str = ""
+    audience_level: str = ""
+    guide_outline: GuideOutline = None
+    sections_content: Dict[str, str] = {}
+
+class GuideCreatorFlow(Flow[GuideCreatorState]):
+    """Flow para criar um guia abrangente sobre qualquer tópico"""
+
+    @start()
+    def get_user_input(self):
+        """Obter entrada do usuário sobre o tópico e público do guia"""
+        print("\n=== Create Your Comprehensive Guide ===\n")
+
+        # Obter entrada do usuário
+        self.state.topic = input("What topic would you like to create a guide for? ")
+
+        # Obter nível do público com validação
+        while True:
+            audience = input("Who is your target audience? (beginner/intermediate/advanced) ").lower()
+            if audience in ["beginner", "intermediate", "advanced"]:
+                self.state.audience_level = audience
+                break
+            print("Please enter 'beginner', 'intermediate', or 'advanced'")
+
+        print(f"\nCreating a guide on {self.state.topic} for {self.state.audience_level} audience...\n")
+        return self.state
+
+    @listen(get_user_input)
+    def create_guide_outline(self, state):
+        """Criar um esboço estruturado para o guia usando uma chamada direta ao LLM"""
+        print("Creating guide outline...")
+
+        # Inicializar o LLM
+        llm = LLM(model="openai/gpt-4o-mini", response_format=GuideOutline)
+
+        # Criar as mensagens para o esboço
+        messages = [
+            {"role": "system", "content": "You are a helpful assistant designed to output JSON."},
+            {"role": "user", "content": f"""
+            Create a detailed outline for a comprehensive guide on "{state.topic}" for {state.audience_level} level learners.
+
+            The outline should include:
+            1. A compelling title for the guide
+            2. An introduction to the topic
+            3. 4-6 main sections that cover the most important aspects of the topic
+            4. A conclusion or summary
+
+            For each section, provide a clear title and a brief description of what it should cover.
+            """}
+        ]
+
+        # Fazer a chamada ao LLM com formato de resposta JSON
+        response = llm.call(messages=messages)
+
+        # Analisar a resposta JSON
+        outline_dict = json.loads(response)
+        self.state.guide_outline = GuideOutline(**outline_dict)
+
+        # Garantir que o diretório de saída exista antes de salvar
+        os.makedirs("output", exist_ok=True)
+
+        # Salvar o esboço em um arquivo
+        with open("output/guide_outline.json", "w") as f:
+            json.dump(outline_dict, f, indent=2)
+
+        print(f"Guide outline created with {len(self.state.guide_outline.sections)} sections")
+        return self.state.guide_outline
+
+    @listen(create_guide_outline)
+    def write_and_compile_guide(self, outline):
+        """Escrever todas as seções e compilar o guia"""
+        print("Writing guide sections and compiling...")
+        completed_sections = []
+
+        # Processar seções uma por uma para manter o fluxo de contexto
+        for section in outline.sections:
+            print(f"Processing section: {section.title}")
+
+            # Construir contexto a partir das seções anteriores
+            previous_sections_text = ""
+            if completed_sections:
+                previous_sections_text = "# Previously Written Sections\n\n"
+                for title in completed_sections:
+                    previous_sections_text += f"## {title}\n\n"
+                    previous_sections_text += self.state.sections_content.get(title, "") + "\n\n"
+            else:
+                previous_sections_text = "No previous sections written yet."
+
+            # Executar a crew de conteúdo para esta seção
+            result = ContentCrew().crew().kickoff(inputs={
+                "section_title": section.title,
+                "section_description": section.description,
+                "audience_level": self.state.audience_level,
+                "previous_sections": previous_sections_text,
+                "draft_content": ""
+            })
+
+            # Armazenar o conteúdo
+            self.state.sections_content[section.title] = result.raw
+            completed_sections.append(section.title)
+            print(f"Section completed: {section.title}")
+
+        # Compilar o guia final
+        guide_content = f"# {outline.title}\n\n"
+        guide_content += f"## Introduction\n\n{outline.introduction}\n\n"
+
+        # Adicionar cada seção em ordem
+        for section in outline.sections:
+            section_content = self.state.sections_content.get(section.title, "")
+            guide_content += f"\n\n{section_content}\n\n"
+
+        # Adicionar conclusão
+        guide_content += f"## Conclusion\n\n{outline.conclusion}\n\n"
+
+        # Salvar o guia
+        with open("output/complete_guide.md", "w") as f:
+            f.write(guide_content)
+
+        print("\nComplete guide compiled and saved to output/complete_guide.md")
+        return "Guide creation completed successfully"
+
+def kickoff():
+    """Executar o flow criador de guias"""
+    GuideCreatorFlow().kickoff()
+    print("\n=== Flow Complete ===")
+    print("Your comprehensive guide is ready in the output directory.")
+    print("Open output/complete_guide.md to view it.")
+
+def plot():
+    """Gerar uma visualização do flow"""
+    flow = GuideCreatorFlow()
+    flow.plot("guide_creator_flow")
+    print("Flow visualization saved to guide_creator_flow.html")
+
+if __name__ == "__main__":
+    kickoff()
 ```
 
 Vamos analisar o que está acontecendo neste flow:

docs/pt-BR/guides/flows/inputs-id-deprecation.mdx

@@ -0,0 +1,142 @@
+---
+title: "Migrando de inputs.id para restore_from_state_id"
+description: "Mover fluxos @persist da hidratação obsoleta inputs.id para o campo suportado restore_from_state_id"
+icon: "arrow-right-arrow-left"
+---
+
+<Warning>
+  Passar `id` dentro de `inputs` para hidratar um fluxo `@persist` é **obsoleto** e
+  programado para remoção em uma versão futura. A substituição, `restore_from_state_id`,
+  está disponível no CrewAI **v1.14.5 e posterior** — os passos abaixo se aplicam uma vez que você
+  faça a atualização.
+</Warning>
+
+## Visão Geral
+
+A maneira documentada de hidratar um fluxo `@persist` de uma execução anterior é passar
+o UUID dessa execução como `inputs.id`. O CrewAI agora expõe um campo dedicado,
+`restore_from_state_id`, que realiza a mesma hidratação sem sobrecarregar a
+carga útil de `inputs` — e sem acoplar a chave de hidratação à identidade da nova execução.
+
+## Migração
+
+Se você atualmente inicia um fluxo `@persist` com `inputs={"id": ...}`:
+
+```python
+# Obsoleto
+flow = CounterFlow()
+flow.kickoff(inputs={"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv"})
+```
+
+Mude para `restore_from_state_id`:
+
+```python
+# Suportado
+flow = CounterFlow()
+flow.kickoff(restore_from_state_id="abcd1234-5678-90ef-ghij-klmnopqrstuv")
+```
+
+Os dois modos têm semânticas de linhagem diferentes:
+
+- `inputs={"id": <uuid>}` (obsoleto) — **retomar**: as gravações são feitas sob o id fornecido,
+  estendendo a mesma história de `flow_uuid`.
+- `restore_from_state_id=<uuid>` — **dividir**: hidrata o estado a partir de um snapshot, então
+  grava sob um novo `state.id`. A história do fluxo de origem é preservada.
+
+Para a maioria dos cenários de produção — reexecutar um fluxo hidratado de um estado anterior — criar um fork
+é o que você deseja. Veja [Dominando o Estado do Fluxo](/pt-BR/guides/flows/mastering-flow-state)
+para o modelo mental completo.
+
+Se você iniciar seu fluxo pela API REST do CrewAI AMP, veja [AMP](#amp) abaixo para a
+migração equivalente da carga útil.
+
+## Por que estamos descontinuando `inputs.id` para `@persist`?
+
+`inputs.id` é atualmente a maneira documentada de retomar um fluxo `@persist` de uma
+execução anterior. O problema é que o mesmo UUID faz duas funções ao mesmo tempo:
+
+1. **Seleciona qual snapshot o `@persist` usa para hidratar** — carrega o estado salvo
+   sob aquele UUID.
+2. **Torna-se o ID de Execução do Fluxo da nova execução** (`state.id` no SDK;
+   apresentado como `flow_id` em alguns contextos) — cada gravação `@persist` a partir desta
+   inicialização também cai sob aquele mesmo UUID.
+
+Esse papel duplo é a causa raiz dos problemas que este guia descreve. Como o
+UUID fornecido também é o id da nova execução, duas inicializações que passam o mesmo
+`inputs.id` não são duas execuções distintas — elas compartilham um id, compartilham um registro
+de persistência e (no AMP) compartilham uma linha na lista de execuções. Não há como dizer
+"hidratar a partir deste snapshot, mas registrar esta execução separadamente" sem dividir as
+duas responsabilidades.
+
+`restore_from_state_id` é essa divisão. Ele informa ao `@persist` de qual snapshot hidratar,
+enquanto deixa a nova execução livre para receber um novo `state.id`. A
+fonte de hidratação e a execução registrada não são mais o mesmo UUID — que é o que
+a maioria dos cenários de produção realmente deseja.
+
+## Cronograma de remoção
+
+`inputs.id` para hidratação `@persist` está programado para remoção em uma versão futura do
+CrewAI. Não há um corte imediato — fluxos existentes continuam a funcionar — mas
+uma vez que você atualize para v1.14.5 ou posterior, novo código deve usar `restore_from_state_id`, e
+fluxos existentes devem migrar na próxima oportunidade conveniente.
+
+## AMP
+
+Se você implantar seu fluxo no CrewAI AMP, a migração se estende à carga útil de inicialização
+enviada para sua Crew implantada, e os sintomas visíveis de reutilização de `inputs.id` aparecem
+no painel de controle de implantação. As duas subseções abaixo cobrem ambos.
+
+### Migrando a carga útil de inicialização
+
+Se você atualmente inicia um fluxo implantado incorporando `id` em `inputs`:
+
+```bash
+# Obsoleto
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{"inputs": {"id": "abcd1234-5678-90ef-ghij-klmnopqrstuv", "topic": "AI Agent Frameworks"}}' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+Mova o UUID para o campo `restoreFromStateId` de nível superior:
+
+```bash
+# Suportado
+curl -X POST \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_CREW_TOKEN" \
+  -d '{
+        "inputs": {"topic": "AI Agent Frameworks"},
+        "restoreFromStateId": "abcd1234-5678-90ef-ghij-klmnopqrstuv"
+      }' \
+  https://your-crew-url.crewai.com/kickoff
+```
+
+`restoreFromStateId` fica ao lado de `inputs` na carga útil de inicialização, não dentro dela. O
+objeto `inputs` agora carrega apenas valores que seu fluxo realmente consome.
+
+### O que acontece quando `inputs.id` é reutilizado
+
+Quando o AMP recebe um kickoff para um fluxo cujo `inputs.id` corresponde a uma execução
+existente, ele resolve para o registro existente em vez de criar um novo. A partir
+do painel de controle de implantação, você verá:
+
+- **Status da execução** — o status da nova execução sobrescreve o status da execução anterior. Uma
+  execução finalizada pode voltar para `running`, ou uma execução `completed` pode mudar para
+  `error` se a nova inicialização falhar — de qualquer forma, o painel não reflete mais
+  a execução original.
+- **Rastros** — Os OTel traces se acumulam entre as inicializações porque compartilham o mesmo
+  id de execução; os traces da execução anterior são substituídos ou misturados
+  com os da nova execução. Uma reprodução passo a passo não corresponde mais a uma única execução.
+- **Lista de execuções** — kickoffs que deveriam aparecer como linhas separadas colapsam em
+  uma única entrada, ocultando o histórico.
+
+Migrar para `restoreFromStateId` mantém cada kickoff como sua própria execução — com
+seu próprio status, traces e entrada na lista — enquanto ainda hidrata o estado de uma
+execução anterior.
+
+<Card title="Precisa de Ajuda?" icon="headset" href="mailto:support@crewai.com">
+  Entre em contato com nossa equipe de suporte se você não tiver certeza de qual modo seu fluxo precisa ou se encontrar problemas
+  durante a migração.
+</Card>

docs/pt-BR/guides/flows/mastering-flow-state.mdx

@@ -63,7 +63,60 @@ Com estado não estruturado:
 Veja um exemplo simples de gerenciamento de estado não estruturado:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+
+class UnstructuredStateFlow(Flow):
+    @start()
+    def initialize_data(self):
+        print("Initializing flow data")
+        # Adiciona pares chave-valor ao estado
+        self.state["user_name"] = "Alex"
+        self.state["preferences"] = {
+            "theme": "dark",
+            "language": "English"
+        }
+        self.state["items"] = []
+
+        # O estado do flow recebe automaticamente um ID único
+        print(f"Flow ID: {self.state['id']}")
+
+        return "Initialized"
+
+    @listen(initialize_data)
+    def process_data(self, previous_result):
+        print(f"Previous step returned: {previous_result}")
+
+        # Acessa e modifica o estado
+        user = self.state["user_name"]
+        print(f"Processing data for {user}")
+
+        # Adiciona itens a uma lista no estado
+        self.state["items"].append("item1")
+        self.state["items"].append("item2")
+
+        # Adiciona um novo par chave-valor
+        self.state["processed"] = True
+
+        return "Processed"
+
+    @listen(process_data)
+    def generate_summary(self, previous_result):
+        # Acessa múltiplos valores do estado
+        user = self.state["user_name"]
+        theme = self.state["preferences"]["theme"]
+        items = self.state["items"]
+        processed = self.state.get("processed", False)
+
+        summary = f"User {user} has {len(items)} items with {theme} theme. "
+        summary += "Data is processed." if processed else "Data is not processed."
+
+        return summary
+
+# Executa o flow
+flow = UnstructuredStateFlow()
+result = flow.kickoff()
+print(f"Final result: {result}")
+print(f"Final state: {flow.state}")
 ```
 
 ### Quando Usar Estado Não Estruturado
@@ -94,7 +147,63 @@ Ao utilizar estado estruturado:
 Veja como implementar o gerenciamento de estado estruturado:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+from pydantic import BaseModel, Field
+from typing import List, Dict, Optional
+
+# Define o modelo de estado
+class UserPreferences(BaseModel):
+    theme: str = "light"
+    language: str = "English"
+
+class AppState(BaseModel):
+    user_name: str = ""
+    preferences: UserPreferences = UserPreferences()
+    items: List[str] = []
+    processed: bool = False
+    completion_percentage: float = 0.0
+
+# Cria um flow com estado tipado
+class StructuredStateFlow(Flow[AppState]):
+    @start()
+    def initialize_data(self):
+        print("Initializing flow data")
+        # Define valores do estado (com checagem de tipo)
+        self.state.user_name = "Taylor"
+        self.state.preferences.theme = "dark"
+
+        # O campo ID está disponível automaticamente
+        print(f"Flow ID: {self.state.id}")
+
+        return "Initialized"
+
+    @listen(initialize_data)
+    def process_data(self, previous_result):
+        print(f"Processing data for {self.state.user_name}")
+
+        # Modifica o estado (com checagem de tipo)
+        self.state.items.append("item1")
+        self.state.items.append("item2")
+        self.state.processed = True
+        self.state.completion_percentage = 50.0
+
+        return "Processed"
+
+    @listen(process_data)
+    def generate_summary(self, previous_result):
+        # Acessa o estado (com autocompletar)
+        summary = f"User {self.state.user_name} has {len(self.state.items)} items "
+        summary += f"with {self.state.preferences.theme} theme. "
+        summary += "Data is processed." if self.state.processed else "Data is not processed."
+        summary += f" Completion: {self.state.completion_percentage}%"
+
+        return summary
+
+# Executa o flow
+flow = StructuredStateFlow()
+result = flow.kickoff()
+print(f"Final result: {result}")
+print(f"Final state: {flow.state}")
 ```
 
 ### Benefícios do Estado Estruturado
@@ -138,7 +247,29 @@ Independente de você usar estado estruturado ou não estruturado, é possível
 Métodos do flow podem retornar valores que serão passados como argumento para métodos listeners:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+
+class DataPassingFlow(Flow):
+    @start()
+    def generate_data(self):
+        # Este valor de retorno será passado para os métodos listeners
+        return "Generated data"
+
+    @listen(generate_data)
+    def process_data(self, data_from_previous_step):
+        print(f"Received: {data_from_previous_step}")
+        # Você pode modificar os dados e repassá-los adiante
+        processed_data = f"{data_from_previous_step} - processed"
+        # Também atualiza o estado
+        self.state["last_processed"] = processed_data
+        return processed_data
+
+    @listen(process_data)
+    def finalize_data(self, processed_data):
+        print(f"Received processed data: {processed_data}")
+        # Acessa tanto os dados passados quanto o estado
+        last_processed = self.state.get("last_processed", "")
+        return f"Final: {processed_data} (from state: {last_processed})"
 ```
 
 Esse padrão permite combinar passagem de dados direta com atualizações de estado para obter máxima flexibilidade.
@@ -156,7 +287,36 @@ O decorador `@persist()` automatiza a persistência de estado, salvando o estado
 Ao aplicar em nível de classe, `@persist()` salva o estado após cada execução de método:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+from crewai.flow.persistence import persist
+from pydantic import BaseModel
+
+class CounterState(BaseModel):
+    value: int = 0
+
+@persist()  # Aplica à classe inteira do flow
+class PersistentCounterFlow(Flow[CounterState]):
+    @start()
+    def increment(self):
+        self.state.value += 1
+        print(f"Incremented to {self.state.value}")
+        return self.state.value
+
+    @listen(increment)
+    def double(self, value):
+        self.state.value = value * 2
+        print(f"Doubled to {self.state.value}")
+        return self.state.value
+
+# Primeira execução
+flow1 = PersistentCounterFlow()
+result1 = flow1.kickoff()
+print(f"First run result: {result1}")
+
+# Segunda execução - passa o ID para carregar o estado persistido
+flow2 = PersistentCounterFlow()
+result2 = flow2.kickoff(inputs={"id": flow1.state.id})
+print(f"Second run result: {result2}")  # Será maior devido ao estado persistido
 ```
 
 #### Persistência em Nível de Método
@@ -164,7 +324,26 @@ Ao aplicar em nível de classe, `@persist()` salva o estado após cada execuçã
 Para mais controle, você pode aplicar `@persist()` em métodos específicos:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+from crewai.flow.persistence import persist
+
+class SelectivePersistFlow(Flow):
+    @start()
+    def first_step(self):
+        self.state["count"] = 1
+        return "First step"
+
+    @persist()  # Persiste apenas após este método
+    @listen(first_step)
+    def important_step(self, prev_result):
+        self.state["count"] += 1
+        self.state["important_data"] = "This will be persisted"
+        return "Important step completed"
+
+    @listen(important_step)
+    def final_step(self, prev_result):
+        self.state["count"] += 1
+        return f"Complete with count {self.state['count']}"
 ```
 
 #### Forking de Estado Persistido
@@ -216,7 +395,45 @@ Notas sobre o comportamento:
 Você pode usar o estado para implementar lógicas condicionais complexas em seus flows:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, router, start
+from pydantic import BaseModel
+
+class PaymentState(BaseModel):
+    amount: float = 0.0
+    is_approved: bool = False
+    retry_count: int = 0
+
+class PaymentFlow(Flow[PaymentState]):
+    @start()
+    def process_payment(self):
+        # Simula o processamento do pagamento
+        self.state.amount = 100.0
+        self.state.is_approved = self.state.amount < 1000
+        return "Payment processed"
+
+    @router(process_payment)
+    def check_approval(self, previous_result):
+        if self.state.is_approved:
+            return "approved"
+        elif self.state.retry_count < 3:
+            return "retry"
+        else:
+            return "rejected"
+
+    @listen("approved")
+    def handle_approval(self):
+        return f"Payment of ${self.state.amount} approved!"
+
+    @listen("retry")
+    def handle_retry(self):
+        self.state.retry_count += 1
+        print(f"Retrying payment (attempt {self.state.retry_count})...")
+        # Aqui poderia ser implementada a lógica de retry
+        return "Retry initiated"
+
+    @listen("rejected")
+    def handle_rejection(self):
+        return f"Payment of ${self.state.amount} rejected after {self.state.retry_count} retries."
 ```
 
 ### Manipulações Complexas de Estado
@@ -224,7 +441,60 @@ Você pode usar o estado para implementar lógicas condicionais complexas em seu
 Para transformar estados complexos, você pode criar métodos dedicados:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+from pydantic import BaseModel
+from typing import List, Dict
+
+class UserData(BaseModel):
+    name: str
+    active: bool = True
+    login_count: int = 0
+
+class ComplexState(BaseModel):
+    users: Dict[str, UserData] = {}
+    active_user_count: int = 0
+
+class TransformationFlow(Flow[ComplexState]):
+    @start()
+    def initialize(self):
+        # Adiciona alguns usuários
+        self.add_user("alice", "Alice")
+        self.add_user("bob", "Bob")
+        self.add_user("charlie", "Charlie")
+        return "Initialized"
+
+    @listen(initialize)
+    def process_users(self, _):
+        # Incrementa contagens de login
+        for user_id in self.state.users:
+            self.increment_login(user_id)
+
+        # Desativa um usuário
+        self.deactivate_user("bob")
+
+        # Atualiza a contagem de ativos
+        self.update_active_count()
+
+        return f"Processed {len(self.state.users)} users"
+
+    # Métodos auxiliares para transformações de estado
+    def add_user(self, user_id: str, name: str):
+        self.state.users[user_id] = UserData(name=name)
+        self.update_active_count()
+
+    def increment_login(self, user_id: str):
+        if user_id in self.state.users:
+            self.state.users[user_id].login_count += 1
+
+    def deactivate_user(self, user_id: str):
+        if user_id in self.state.users:
+            self.state.users[user_id].active = False
+            self.update_active_count()
+
+    def update_active_count(self):
+        self.state.active_user_count = sum(
+            1 for user in self.state.users.values() if user.active
+        )
 ```
 
 Esse padrão de criar métodos auxiliares mantém seus métodos de flow limpos, enquanto permite manipulações complexas de estado.
@@ -238,7 +508,71 @@ Um dos padrões mais poderosos na CrewAI é combinar o gerenciamento de estado d
 Você pode usar o estado do flow para parametrizar crews:
 
 ```python
-# código não traduzido
+from crewai.flow.flow import Flow, listen, start
+from crewai import Agent, Crew, Process, Task
+from pydantic import BaseModel
+
+class ResearchState(BaseModel):
+    topic: str = ""
+    depth: str = "medium"
+    results: str = ""
+
+class ResearchFlow(Flow[ResearchState]):
+    @start()
+    def get_parameters(self):
+        # Em uma aplicação real, isso pode vir da entrada do usuário
+        self.state.topic = "Artificial Intelligence Ethics"
+        self.state.depth = "deep"
+        return "Parameters set"
+
+    @listen(get_parameters)
+    def execute_research(self, _):
+        # Cria os agentes
+        researcher = Agent(
+            role="Research Specialist",
+            goal=f"Research {self.state.topic} in {self.state.depth} detail",
+            backstory="You are an expert researcher with a talent for finding accurate information."
+        )
+
+        writer = Agent(
+            role="Content Writer",
+            goal="Transform research into clear, engaging content",
+            backstory="You excel at communicating complex ideas clearly and concisely."
+        )
+
+        # Cria as tarefas
+        research_task = Task(
+            description=f"Research {self.state.topic} with {self.state.depth} analysis",
+            expected_output="Comprehensive research notes in markdown format",
+            agent=researcher
+        )
+
+        writing_task = Task(
+            description=f"Create a summary on {self.state.topic} based on the research",
+            expected_output="Well-written article in markdown format",
+            agent=writer,
+            context=[research_task]
+        )
+
+        # Cria e executa a crew
+        research_crew = Crew(
+            agents=[researcher, writer],
+            tasks=[research_task, writing_task],
+            process=Process.sequential,
+            verbose=True
+        )
+
+        # Executa a crew e armazena o resultado no estado
+        result = research_crew.kickoff()
+        self.state.results = result.raw
+
+        return "Research completed"
+
+    @listen(execute_research)
+    def summarize_results(self, _):
+        # Acessa os resultados armazenados
+        result_length = len(self.state.results)
+        return f"Research on {self.state.topic} completed with {result_length} characters of results."
 ```
 
 ### Manipulando Saídas de Crews no Estado
@@ -246,7 +580,21 @@ Você pode usar o estado do flow para parametrizar crews:
 Quando um crew finaliza, é possível processar sua saída e armazená-la no estado do flow:
 
 ```python
-# código não traduzido
+@listen(execute_crew)
+def process_crew_results(self, _):
+    # Faz parsing dos resultados brutos (assumindo saída em JSON)
+    import json
+    try:
+        results_dict = json.loads(self.state.raw_results)
+        self.state.processed_results = {
+            "title": results_dict.get("title", ""),
+            "main_points": results_dict.get("main_points", []),
+            "conclusion": results_dict.get("conclusion", "")
+        }
+        return "Results processed successfully"
+    except json.JSONDecodeError:
+        self.state.error = "Failed to parse crew results as JSON"
+        return "Error processing results"
 ```
 
 ## Boas Práticas para Gerenciamento de Estado
@@ -256,7 +604,19 @@ Quando um crew finaliza, é possível processar sua saída e armazená-la no est
 Projete seu estado para conter somente o necessário:
 
 ```python
-# Exemplo não traduzido
+# Abrangente demais
+class BloatedState(BaseModel):
+    user_data: Dict = {}
+    system_settings: Dict = {}
+    temporary_calculations: List = []
+    debug_info: Dict = {}
+    # ...muitos outros campos
+
+# Melhor: estado focado
+class FocusedState(BaseModel):
+    user_id: str
+    preferences: Dict[str, str]
+    completion_status: Dict[str, bool]
 ```
 
 ### 2. Use Estado Estruturado em Flows Complexos
@@ -264,7 +624,23 @@ Projete seu estado para conter somente o necessário:
 À medida que seus flows evoluem em complexidade, o estado estruturado se torna cada vez mais valioso:
 
 ```python
-# Exemplo não traduzido
+# Flow simples pode usar estado não estruturado
+class SimpleGreetingFlow(Flow):
+    @start()
+    def greet(self):
+        self.state["name"] = "World"
+        return f"Hello, {self.state['name']}!"
+
+# Flow complexo se beneficia de estado estruturado
+class UserRegistrationState(BaseModel):
+    username: str
+    email: str
+    verification_status: bool = False
+    registration_date: datetime = Field(default_factory=datetime.now)
+    last_login: Optional[datetime] = None
+
+class RegistrationFlow(Flow[UserRegistrationState]):
+    # Métodos com acesso ao estado fortemente tipado
 ```
 
 ### 3. Documente Transições de Estado
@@ -272,7 +648,18 @@ Projete seu estado para conter somente o necessário:
 Para flows complexos, documente como o estado muda ao longo da execução:
 
 ```python
-# Exemplo não traduzido
+@start()
+def initialize_order(self):
+    """
+    Initialize order state with empty values.
+
+    State before: {}
+    State after: {order_id: str, items: [], status: 'new'}
+    """
+    self.state.order_id = str(uuid.uuid4())
+    self.state.items = []
+    self.state.status = "new"
+    return "Order initialized"
 ```
 
 ### 4. Trate Erros de Estado de Forma Elegante
@@ -280,7 +667,18 @@ Para flows complexos, documente como o estado muda ao longo da execução:
 Implemente tratamento de erros ao acessar o estado:
 
 ```python
-# Exemplo não traduzido
+@listen(previous_step)
+def process_data(self, _):
+    try:
+        # Tenta acessar um valor que pode não existir
+        user_preference = self.state.preferences.get("theme", "default")
+    except (AttributeError, KeyError):
+        # Trata o erro de forma elegante
+        self.state.errors = self.state.get("errors", [])
+        self.state.errors.append("Failed to access preferences")
+        user_preference = "default"
+
+    return f"Used preference: {user_preference}"
 ```
 
 ### 5. Use o Estado Para Acompanhar o Progresso
@@ -288,7 +686,30 @@ Implemente tratamento de erros ao acessar o estado:
 Aproveite o estado para monitorar o progresso em flows de longa duração:
 
 ```python
-# Exemplo não traduzido
+class ProgressTrackingFlow(Flow):
+    @start()
+    def initialize(self):
+        self.state["total_steps"] = 3
+        self.state["current_step"] = 0
+        self.state["progress"] = 0.0
+        self.update_progress()
+        return "Initialized"
+
+    def update_progress(self):
+        """Helper method to calculate and update progress"""
+        if self.state.get("total_steps", 0) > 0:
+            self.state["progress"] = (self.state.get("current_step", 0) /
+                                    self.state["total_steps"]) * 100
+            print(f"Progress: {self.state['progress']:.1f}%")
+
+    @listen(initialize)
+    def step_one(self, _):
+        # Realiza o trabalho...
+        self.state["current_step"] = 1
+        self.update_progress()
+        return "Step 1 complete"
+
+    # Etapas adicionais...
 ```
 
 ### 6. Prefira Operações Imutáveis Quando Possível
@@ -296,7 +717,22 @@ Aproveite o estado para monitorar o progresso em flows de longa duração:
 Especialmente com estado estruturado, prefira operações imutáveis para maior clareza:
 
 ```python
-# Exemplo não traduzido
+# Em vez de modificar listas no local:
+self.state.items.append(new_item)  # Operação mutável
+
+# Considere criar um novo estado:
+from pydantic import BaseModel
+from typing import List
+
+class ItemState(BaseModel):
+    items: List[str] = []
+
+class ImmutableFlow(Flow[ItemState]):
+    @start()
+    def add_item(self):
+        # Cria uma nova lista com o item adicionado
+        self.state.items = [*self.state.items, "new item"]
+        return "Item added"
 ```
 
 ## Depurando o Estado do Flow
@@ -306,7 +742,24 @@ Especialmente com estado estruturado, prefira operações imutáveis para maior
 Ao desenvolver, adicione logs para acompanhar mudanças no estado:
 
 ```python
-# Exemplo não traduzido
+import logging
+logging.basicConfig(level=logging.INFO)
+
+class LoggingFlow(Flow):
+    def log_state(self, step_name):
+        logging.info(f"State after {step_name}: {self.state}")
+
+    @start()
+    def initialize(self):
+        self.state["counter"] = 0
+        self.log_state("initialize")
+        return "Initialized"
+
+    @listen(initialize)
+    def increment(self, _):
+        self.state["counter"] += 1
+        self.log_state("increment")
+        return f"Incremented to {self.state['counter']}"
 ```
 
 ### Visualizando o Estado
@@ -314,7 +767,30 @@ Ao desenvolver, adicione logs para acompanhar mudanças no estado:
 Você pode adicionar métodos para visualizar seu estado durante o debug:
 
 ```python
-# Exemplo não traduzido
+def visualize_state(self):
+    """Create a simple visualization of the current state"""
+    import json
+    from rich.console import Console
+    from rich.panel import Panel
+
+    console = Console()
+
+    if hasattr(self.state, "model_dump"):
+        # Pydantic v2
+        state_dict = self.state.model_dump()
+    elif hasattr(self.state, "dict"):
+        # Pydantic v1
+        state_dict = self.state.dict()
+    else:
+        # Estado não estruturado
+        state_dict = dict(self.state)
+
+    # Remove o id para uma saída mais limpa
+    if "id" in state_dict:
+        state_dict.pop("id")
+
+    state_json = json.dumps(state_dict, indent=2, default=str)
+    console.print(Panel(state_json, title="Current Flow State"))
 ```
 
 ## Conclusão

docs/pt-BR/guides/migration/upgrading-crewai.mdx

@@ -0,0 +1,190 @@
+---
+title: "Atualizando o CrewAI"
+description: "Como atualizar o CrewAI no seu projeto e adaptar-se a breaking changes entre versões."
+icon: "arrow-up-circle"
+---
+
+## Visão Geral
+
+Os lançamentos do CrewAI trazem novos recursos regularmente. Este guia mostra os passos práticos para manter sua instalação atualizada — tanto a CLI quanto o ambiente virtual do seu projeto.
+
+Se você está começando do zero, veja [Instalação](/pt-BR/installation). Se está vindo de outro framework, veja [Migrando do LangGraph](/pt-BR/guides/migration/migrating-from-langgraph).
+
+---
+
+## As Duas Coisas Que Você Pode Querer Atualizar
+
+O CrewAI vive em dois lugares na sua máquina, e cada um se atualiza de forma independente:
+
+| O quê | Como é instalado | Como atualizar |
+|---|---|---|
+| A **CLI global `crewai`** | `uv tool install crewai` | `uv tool install crewai --upgrade` |
+| O **venv do projeto** (onde seu código roda) | `crewai install` / `uv sync` | `uv add "crewai[...]>=X.Y.Z"` e depois `crewai install` |
+
+Esses dois podem — e frequentemente ficam — fora de sincronia. Rodar `crewai --version` mostra a versão da CLI. Rodar `uv pip show crewai` dentro do seu projeto mostra a versão do venv. Se forem diferentes, isso é normal; o que importa para o código em execução é a versão do venv.
+
+## Por Que `crewai install` Sozinho Não Atualiza
+
+`crewai install` é um wrapper fino em torno de `uv sync`. Ele instala exatamente o que o arquivo `uv.lock` atual diz — ele **não** muda nenhuma restrição de versão.
+
+Se seu `pyproject.toml` diz `crewai>=1.11.1` e o lock file resolveu para `1.11.1`, executar `crewai install` vai te manter em `1.11.1` para sempre, mesmo que `1.14.4` esteja disponível.
+
+Para realmente atualizar, você precisa:
+
+1. Atualizar a restrição de versão em `pyproject.toml`
+2. Re-resolver o lock file
+3. Sincronizar o venv
+
+`uv add` faz os três de uma vez só.
+
+## Como Atualizar Seu Projeto
+
+```bash
+# Aumenta a restrição e re-resolve o lock em um único comando
+uv add "crewai[tools]>=1.14.4"
+
+# Sincroniza o venv (crewai install chama uv sync por baixo dos panos)
+crewai install
+
+# Verifica
+uv pip show crewai
+# → Version: 1.14.4
+```
+
+Substitua `[tools]` por quaisquer extras que seu projeto utilize (ex.: `[tools,anthropic]`). Verifique a lista de `dependencies` do seu `pyproject.toml` se estiver em dúvida.
+
+<Note>
+  `uv add` atualiza tanto `pyproject.toml` **quanto** `uv.lock` atomicamente. Se você editar `pyproject.toml` manualmente, ainda precisa rodar `uv lock --upgrade-package crewai` para re-resolver o lock file antes que `crewai install` pegue a nova versão.
+</Note>
+
+## Atualizando a CLI Global
+
+A CLI global é separada do seu projeto. Atualize com:
+
+```bash
+uv tool install crewai --upgrade
+```
+
+Se seu shell avisar sobre o `PATH` após a atualização, recarregue-o:
+
+```bash
+uv tool update-shell
+```
+
+Isso **não** mexe no venv do seu projeto — você ainda precisa de `uv add` + `crewai install` dentro do projeto.
+
+## Verifique Se Ambos Estão em Sincronia
+
+```bash
+# Versão da CLI global
+crewai --version
+
+# Versão do venv do projeto
+uv pip show crewai | grep Version
+```
+
+Eles não precisam coincidir — mas a versão do venv do projeto é o que importa para o comportamento em runtime.
+
+<Note>
+  CrewAI requer `Python >=3.10, <3.14`. Se o `uv` foi instalado contra um interpretador mais antigo, recrie o venv do projeto com uma versão suportada do Python antes de rodar `crewai install`.
+</Note>
+
+---
+
+## Breaking Changes e Notas de Migração
+
+A maioria das atualizações requer apenas pequenos ajustes. As áreas abaixo são as que quebram silenciosamente ou com tracebacks confusos.
+
+### Caminhos de import: tools e `BaseTool`
+
+O caminho canônico para tools é `crewai.tools`. Caminhos antigos ainda aparecem em tutoriais, mas devem ser atualizados.
+
+```python
+# Antes
+from crewai_tools import BaseTool
+from crewai.agents.tools import tool
+
+# Depois
+from crewai.tools import BaseTool, tool
+```
+
+O decorador `@tool` e a subclasse `BaseTool` ambos vivem em `crewai.tools`. `AgentFinish` e outros símbolos internos do agente não fazem mais parte da superfície pública — se você os estava importando, mude para event listeners ou callbacks de `Task`.
+
+### Mudanças de parâmetros em `Agent`
+
+```python
+from crewai import Agent
+
+agent = Agent(
+    role="Researcher",
+    goal="Find authoritative sources on {topic}",
+    backstory="You are a careful, source-driven researcher.",
+    llm="gpt-4o-mini",   # nome do modelo como string OU um objeto LLM
+    verbose=True,        # bool, não um nível inteiro
+    max_iter=15,         # default mudou entre versões — defina explicitamente
+    allow_delegation=False,
+)
+```
+
+- `llm` aceita tanto um nome de modelo como string (resolvido pelo provedor configurado) quanto um objeto `LLM` para controle granular.
+- `verbose` é um `bool` puro. Passar um inteiro não alterna mais níveis de log.
+- Os defaults de `max_iter` mudaram entre releases. Se seu agente para silenciosamente de iterar após a primeira chamada de tool, defina `max_iter` explicitamente.
+
+### Parâmetros de `Crew`
+
+```python
+from crewai import Crew, Process
+
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,   # ou Process.hierarchical
+    memory=True,
+    cache=True,
+    embedder={"provider": "openai", "config": {"model": "text-embedding-3-small"}},
+)
+```
+
+- `process=Process.hierarchical` requer ou `manager_llm=` ou `manager_agent=`. Sem um deles, o kickoff lança erro na validação.
+- `memory=True` com um provedor de embedding não-default precisa de um dicionário `embedder` — veja [Configuração de memória e embedder](#memory-embedder-config) abaixo.
+
+### Saída estruturada de `Task`
+
+Use `output_pydantic`, `output_json` ou `output_file` para forçar o resultado de uma task em um formato tipado:
+
+```python
+from pydantic import BaseModel
+from crewai import Task
+
+class Article(BaseModel):
+    title: str
+    body: str
+
+write = Task(
+    description="Write an article about {topic}",
+    expected_output="A short article with a title and body",
+    agent=writer,
+    output_pydantic=Article,        # a classe, NÃO uma instância
+    output_file="output/article.md",
+)
+```
+
+`output_pydantic` recebe a **classe** em si. Passar `Article(title="", body="")` é um erro comum e falha com um erro de validação confuso.
+
+### Configuração de memória e embedder {#memory-embedder-config}
+
+Se `memory=True` e você não está usando os embeddings padrão da OpenAI, é preciso passar um `embedder`:
+
+```python
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    memory=True,
+    embedder={
+        "provider": "ollama",
+        "config": {"model": "nomic-embed-text"},
+    },
+)
+```
+
+Defina as credenciais do provedor relevante (`OPENAI_API_KEY`, `OLLAMA_HOST`, etc.) no seu arquivo `.env`. Os caminhos de armazenamento de memória são locais ao projeto por default — apague o diretório de memória do projeto se trocar de embedder, já que dimensões diferentes não se misturam.

docs/pt-BR/learn/llm-selection-guide.mdx

@@ -797,7 +797,6 @@ As tabelas abaixo mostram uma amostra dos modelos de maior destaque em cada cate
     Inicie com opções consagradas como **GPT-4.1**, **Claude 3.7 Sonnet** ou **Gemini 2.0 Flash**, que oferecem bom desempenho e ampla validação.
   </Step>
 
-{" "}
 <Step title="Identifique Demandas Especializadas">
   Descubra se sua crew possui requisitos específicos (código, raciocínio,
   velocidade) que justifiquem modelos como **Claude 4 Sonnet** para
@@ -805,7 +804,6 @@ As tabelas abaixo mostram uma amostra dos modelos de maior destaque em cada cate
   velocidade, considere Groq aliado à seleção do modelo.
 </Step>
 
-{" "}
 <Step title="Implemente Estratégia Multi-Modelo">
   Use modelos diferentes para agentes distintos conforme o papel. Modelos de
   alta capacidade para managers e tarefas complexas, eficientes para rotinas.

docs/pt-BR/tools/ai-ml/daytona.mdx

@@ -13,7 +13,7 @@ The Daytona sandbox tools give CrewAI agents access to isolated, ephemeral compu
 
 - **`DaytonaExecTool`** — run any shell command inside a sandbox.
 - **`DaytonaPythonTool`** — execute a block of Python source code inside a sandbox.
-- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox.
+- **`DaytonaFileTool`** — read, write, append, list, delete, and inspect files inside a sandbox; also supports `move`, `find` (content grep), `search` (filename glob), `chmod` (permissions), `replace` (bulk find-and-replace), and `exists`.
 
 All three tools share the same sandbox lifecycle controls, so you can mix and match them while keeping state in a single persistent sandbox.
 
@@ -55,7 +55,7 @@ from crewai_tools import DaytonaPythonTool
 tool = DaytonaPythonTool()
 result = tool.run(code="print(sum(range(10)))")
 print(result)
-# {"exit_code": 0, "result": "45\n", "artifacts": None}
+# {"exit_code": 0, "result": "45\n", "artifacts": ExecutionArtifacts(stdout="45\n", charts=[])}
 ```
 
 ### Multi-step shell session (persistent)
@@ -63,17 +63,22 @@ print(result)
 ```python Code
 from crewai_tools import DaytonaExecTool, DaytonaFileTool
 
+# Create the persistent sandbox via the first tool, then attach the second
+# tool to it so both share state (installed packages, files, env vars).
 exec_tool = DaytonaExecTool(persistent=True)
-file_tool = DaytonaFileTool(persistent=True)
-
-# Install a package, then write and run a script — all in the same sandbox
 exec_tool.run(command="pip install httpx -q")
-file_tool.run(action="write", path="/workspace/fetch.py", content="import httpx; print(httpx.get('https://httpbin.org/get').status_code)")
-exec_tool.run(command="python /workspace/fetch.py")
+file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
+
+file_tool.run(
+    action="write",
+    path="workspace/script.py",
+    content="import httpx; print(f'httpx loaded, version {httpx.__version__}')",
+)
+exec_tool.run(command="python workspace/script.py")
 ```
 
 <Note>
-Each tool instance maintains its own persistent sandbox. To share **one** sandbox across two tools, create the first tool, grab its sandbox id via `tool._persistent_sandbox.id`, and pass it to the second tool via `sandbox_id=...`.
+By default, each tool with `persistent=True` lazily creates its **own** sandbox on first use. The pattern above shares a single sandbox across multiple tools by reading the first tool's `active_sandbox_id` after a `.run()` call and passing it to the others via `sandbox_id=...`. With `persistent=False` (the default), every `.run()` call gets a fresh sandbox that's deleted at the end of that call.
 </Note>
 
 ### Attach to an existing sandbox
@@ -82,7 +87,7 @@ Each tool instance maintains its own persistent sandbox. To share **one** sandbo
 from crewai_tools import DaytonaExecTool
 
 tool = DaytonaExecTool(sandbox_id="my-long-lived-sandbox")
-result = tool.run(command="ls /workspace")
+result = tool.run(command="ls workspace")
 ```
 
 ### Custom sandbox parameters
@@ -102,6 +107,41 @@ tool = DaytonaExecTool(
 )
 ```
 
+### Searching, moving, and modifying files
+
+```python Code
+from crewai_tools import DaytonaFileTool
+
+file_tool = DaytonaFileTool(persistent=True)
+
+# Find every TODO in the source tree (grep file contents recursively)
+file_tool.run(action="find", path="workspace/src", pattern="TODO:")
+
+# Find all Python files (glob match on filenames)
+file_tool.run(action="search", path="workspace", pattern="*.py")
+
+# Make a script executable
+file_tool.run(action="chmod", path="workspace/run.sh", mode="755")
+
+# Rename or move a file
+file_tool.run(
+    action="move",
+    path="workspace/draft.md",
+    destination="workspace/final.md",
+)
+
+# Bulk find-and-replace across multiple files
+file_tool.run(
+    action="replace",
+    paths=["workspace/src/a.py", "workspace/src/b.py"],
+    pattern="old_function",
+    replacement="new_function",
+)
+
+# Quick existence check before a destructive op
+file_tool.run(action="exists", path="workspace/cache.db")
+```
+
 ### Agent integration
 
 ```python Code
@@ -121,7 +161,7 @@ coder = Agent(
 )
 
 task = Task(
-    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to /workspace/fib.py, and run it.",
+    description="Write a Python script that prints the first 10 Fibonacci numbers, save it to workspace/fib.py, and run it.",
     expected_output="The first 10 Fibonacci numbers printed to stdout.",
     agent=coder,
 )
@@ -168,12 +208,22 @@ All three tools accept these parameters at initialization:
 
 | Parameter | Type | Required | Description |
 |-----------|------|----------|-------------|
-| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`. |
-| `path` | `str` | ✓ | Absolute path inside the sandbox. |
-| `content` | `str \| None` | | Content to write or append. Required for `append`. |
+| `action` | `str` | ✓ | One of: `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`, `exists`, `move`, `find`, `search`, `chmod`, `replace`. |
+| `path` | `str \| None` | ✓ for all actions except `replace` | Absolute path inside the sandbox. |
+| `content` | `str \| None` | ✓ for `append` | Content to write or append. |
 | `binary` | `bool` | | If `True`, `content` is base64 on write; returns base64 on read. |
 | `recursive` | `bool` | | For `delete`: remove directories recursively. |
-| `mode` | `str` | | For `mkdir`: octal permission string (default `"0755"`). |
+| `mode` | `str \| None` | | For `mkdir`: octal permissions for the new directory (defaults to `"0755"`). For `chmod`: octal permissions to apply to the target. |
+| `destination` | `str \| None` | ✓ for `move` | Destination path for `move`. |
+| `pattern` | `str \| None` | ✓ for `find`, `search`, `replace` | For `find`: substring matched against file CONTENTS. For `search`: glob matched against file NAMES (e.g. `*.py`). For `replace`: text to replace inside files. |
+| `replacement` | `str \| None` | ✓ for `replace` | Replacement text for `pattern`. |
+| `paths` | `list[str] \| None` | ✓ for `replace` | List of file paths in which to replace text. |
+| `owner` | `str \| None` | | For `chmod`: new file owner. |
+| `group` | `str \| None` | | For `chmod`: new file group. |
+
+<Note>
+For `chmod`, pass at least one of `mode`, `owner`, or `group` — any field left as `None` is left unchanged on the target.
+</Note>
 
 <Tip>
 For files larger than a few KB, create the file first with `action="write"` and empty content, then send the body via multiple `action="append"` calls of ~4 KB each to stay within tool-call payload limits.

lib/cli/pyproject.toml

@@ -8,7 +8,7 @@ authors = [
 ]
 requires-python = ">=3.10, <3.14"
 dependencies = [
-    "crewai-core==1.14.5a3",
+    "crewai-core==1.14.5",
     "click~=8.1.7",
     "pydantic>=2.11.9,<2.13",
     "pydantic-settings~=2.10.1",
@@ -22,6 +22,8 @@ dependencies = [
     "packaging>=23.0",
     "python-dotenv>=1.2.2,<2",
     "uv~=0.11.6",
+    "textual>=7.5.0",
+    "certifi",
 ]
 
 [project.urls]

lib/cli/src/crewai_cli/__init__.py

@@ -1 +1 @@
-__version__ = "1.14.5a3"
+__version__ = "1.14.5"

lib/cli/src/crewai_cli/cli.py

@@ -26,6 +26,7 @@ from crewai_cli.replay_from_task import replay_task_command
 from crewai_cli.reset_memories_command import reset_memories_command
 from crewai_cli.run_crew import run_crew
 from crewai_cli.settings.main import SettingsCommand
+from crewai_cli.skills.main import SkillCommand
 from crewai_cli.task_outputs import load_task_outputs
 from crewai_cli.tools.main import ToolCommand
 from crewai_cli.train_crew import train_crew
@@ -546,6 +547,56 @@ def tool_publish(is_public: bool, force: bool) -> None:
     tool_cmd.publish(is_public, force)
 
 
+@crewai.group()
+def skill() -> None:
+    """Skill Repository related commands."""
+
+
+@skill.command(name="create")
+@click.argument("name")
+@click.option(
+    "--no-project",
+    "in_project",
+    is_flag=True,
+    default=True,
+    flag_value=False,
+    help="Create skill in current dir instead of ./skills/",
+)
+def skill_create(name: str, in_project: bool) -> None:
+    skill_cmd = SkillCommand()
+    skill_cmd.create(name, in_project=in_project)
+
+
+@skill.command(name="install")
+@click.argument("ref")
+def skill_install(ref: str) -> None:
+    skill_cmd = SkillCommand()
+    skill_cmd.install(ref)
+
+
+@skill.command(name="publish")
+@click.option(
+    "--force",
+    is_flag=True,
+    default=False,
+    show_default=True,
+    help="Skip git-state validation.",
+)
+@click.option("--public", "is_public", flag_value=True, default=False)
+@click.option("--private", "is_public", flag_value=False)
+@click.option("--org", default=None, help="Organisation slug (overrides settings).")
+def skill_publish(is_public: bool, org: str | None, force: bool) -> None:
+    skill_cmd = SkillCommand()
+    skill_cmd.publish(is_public, org=org, force=force)
+
+
+@skill.command(name="list")
+def skill_list() -> None:
+    """List locally installed skills."""
+    skill_cmd = SkillCommand()
+    skill_cmd.list_cached()
+
+
 @crewai.group()
 def template() -> None:
     """Browse and install project templates."""

lib/cli/src/crewai_cli/constants.py

@@ -132,19 +132,44 @@ PROVIDERS: list[str] = [
 
 MODELS: dict[str, list[str]] = {
     "openai": [
-        "gpt-4",
+        "gpt-5.5",
+        "gpt-5.5-pro",
+        "gpt-5.4",
+        "gpt-5.4-pro",
+        "gpt-5.4-mini",
+        "gpt-5.4-nano",
+        "gpt-5.2",
+        "gpt-5.2-pro",
+        "gpt-5.1",
+        "gpt-5",
+        "gpt-5-pro",
+        "gpt-5-mini",
+        "gpt-5-nano",
         "gpt-4.1",
-        "gpt-4.1-mini-2025-04-14",
-        "gpt-4.1-nano-2025-04-14",
+        "gpt-4.1-mini",
+        "gpt-4.1-nano",
         "gpt-4o",
         "gpt-4o-mini",
+        "o4-mini",
+        "o3",
+        "o3-mini",
+        "o1",
         "o1-mini",
         "o1-preview",
+        "gpt-4",
+        "gpt-4.1-mini-2025-04-14",
+        "gpt-4.1-nano-2025-04-14",
     ],
     "anthropic": [
+        "claude-opus-4-6",
+        "claude-sonnet-4-6",
+        "claude-haiku-4-5-20251001",
+        "claude-3-7-sonnet-20250219",
+        "claude-3-5-sonnet-20241022",
+        "claude-3-5-haiku-20241022",
         "claude-3-5-sonnet-20240620",
-        "claude-3-sonnet-20240229",
         "claude-3-opus-20240229",
+        "claude-3-sonnet-20240229",
         "claude-3-haiku-20240307",
     ],
     "gemini": [

lib/cli/src/crewai_cli/deploy/main.py

@@ -1,5 +1,6 @@
 from typing import Any
 
+from crewai_core.plus_api import CreateCrewPayload
 from rich.console import Console
 
 from crewai_cli import git
@@ -161,7 +162,7 @@ class DeployCommand(BaseCommand, PlusAPIMixin):
         self,
         env_vars: dict[str, str],
         remote_repo_url: str,
-    ) -> dict[str, Any]:
+    ) -> CreateCrewPayload:
         """
         Create the payload for crew creation.
 
@@ -172,6 +173,8 @@ class DeployCommand(BaseCommand, PlusAPIMixin):
         Returns:
             Dict[str, Any]: The payload for crew creation.
         """
+        if not self.project_name:
+            raise ValueError("project_name is required to create a deployment payload")
         return {
             "deploy": {
                 "name": self.project_name,

lib/cli/src/crewai_cli/git.py

@@ -1,4 +1,4 @@
-from functools import lru_cache
+from functools import cached_property
 import subprocess
 
 
@@ -9,7 +9,7 @@ class Repository:
         if not self.is_git_installed():
             raise ValueError("Git is not installed or not found in your PATH.")
 
-        if not self.is_git_repo():
+        if not self.is_git_repo:
             raise ValueError(f"{self.path} is not a Git repository.")
 
         self.fetch()
@@ -40,13 +40,9 @@ class Repository:
             encoding="utf-8",
         ).strip()
 
-    @lru_cache(maxsize=None)  # noqa: B019
+    @cached_property
     def is_git_repo(self) -> bool:
-        """Check if the current directory is a git repository.
-
-        Notes:
-          - TODO: This method is cached to avoid redundant checks, but using lru_cache on methods can lead to memory leaks
-        """
+        """Check if the current directory is a git repository."""
         try:
             subprocess.check_output(
                 ["git", "rev-parse", "--is-inside-work-tree"],  # noqa: S607

lib/cli/src/crewai_cli/skills/main.py

@@ -0,0 +1,415 @@
+"""Skill Repository CLI commands for CrewAI."""
+
+from __future__ import annotations
+
+import base64
+import io
+import json
+import os
+from pathlib import Path
+import tarfile
+import zipfile
+
+from rich.console import Console
+from rich.table import Table
+
+from crewai_cli.command import BaseCommand, PlusAPIMixin
+from crewai_cli.config import Settings
+from crewai_cli.constants import DEFAULT_CREWAI_ENTERPRISE_URL
+
+
+console = Console()
+
+_SKILL_MD_TEMPLATE = """\
+---
+name: {name}
+version: 0.1.0
+description: |
+  A short description of what this skill does.
+---
+
+## Instructions
+
+Describe the skill behaviour here. This section is shown to the agent at activation time.
+"""
+
+
+class SkillCommand(BaseCommand, PlusAPIMixin):
+    """Skill Repository related operations for CrewAI projects."""
+
+    def __init__(self) -> None:
+        BaseCommand.__init__(self)
+        PlusAPIMixin.__init__(self, telemetry=self._telemetry)
+
+    # ------------------------------------------------------------------
+    # create
+    # ------------------------------------------------------------------
+
+    def create(self, name: str, in_project: bool = True) -> None:
+        """Scaffold a new skill directory.
+
+        If pyproject.toml is present (crew project), creates ./skills/{name}/.
+        Otherwise creates ./{name}/.
+        """
+        if in_project and os.path.isfile("pyproject.toml"):
+            skill_dir = Path("skills") / name
+        else:
+            skill_dir = Path(name)
+
+        if skill_dir.exists():
+            console.print(f"[red]Directory {skill_dir} already exists.[/red]")
+            raise SystemExit(1)
+
+        skill_dir.mkdir(parents=True)
+        (skill_dir / "scripts").mkdir()
+        (skill_dir / "references").mkdir()
+        (skill_dir / "assets").mkdir()
+
+        skill_md = skill_dir / "SKILL.md"
+        skill_md.write_text(_SKILL_MD_TEMPLATE.format(name=name))
+
+        console.print(
+            f"[green]Created skill [bold]{name}[/bold] at [bold]{skill_dir}[/bold].[/green]"
+        )
+        console.print(f"Edit [bold]{skill_md}[/bold] to define the skill instructions.")
+
+    # ------------------------------------------------------------------
+    # install
+    # ------------------------------------------------------------------
+
+    def install(self, ref: str) -> None:
+        """Download and install a registry skill.
+
+        Format: @org/name
+
+        Inside a crew project (pyproject.toml present): installs to ./skills/{name}/
+        Outside a project: installs to ~/.crewai/skills/{org}/{name}/
+        """
+        if not ref.startswith("@"):
+            console.print(
+                "[red]Invalid skill reference. Use the format @org/name.[/red]"
+            )
+            raise SystemExit(1)
+
+        without_at = ref[1:]
+        if without_at.count("/") != 1:
+            console.print(
+                "[red]Invalid skill reference. Use the format @org/name.[/red]"
+            )
+            raise SystemExit(1)
+
+        org, name = without_at.split("/", 1)
+        if (
+            not org
+            or not name
+            or org.startswith(".")
+            or name.startswith(".")
+            or len(Path(org).parts) != 1
+            or len(Path(name).parts) != 1
+        ):
+            console.print(
+                "[red]Invalid skill reference: org and name must be single, "
+                "non-empty path segments (no slashes, no '..').[/red]"
+            )
+            raise SystemExit(1)
+
+        self._print_current_organization()
+        console.print(f"[bold blue]Downloading skill {ref}...[/bold blue]")
+
+        get_response = self.plus_api_client.get_skill(org, name)
+
+        if get_response.status_code == 404:
+            console.print(
+                f"[red]Skill {ref} not found. Ensure it has been published and you have access.[/red]"
+            )
+            raise SystemExit(1)
+        if get_response.status_code != 200:
+            console.print(
+                f"[red]Failed to download skill {ref}: {get_response.status_code}[/red]"
+            )
+            raise SystemExit(1)
+
+        data = get_response.json()
+        version = data.get("latest_version") or data.get("version")
+
+        download_url = data.get("download_url")
+        if download_url:
+            import httpx
+
+            dl_response = httpx.get(download_url, follow_redirects=True)
+            dl_response.raise_for_status()
+            archive_bytes = dl_response.content
+        else:
+            encoded = data.get("file", "")
+            if "," in encoded:
+                encoded = encoded.split(",", 1)[1]
+            archive_bytes = base64.b64decode(encoded)
+
+        in_project = os.path.isfile("pyproject.toml")
+        if in_project:
+            dest = Path("skills") / name
+            dest.mkdir(parents=True, exist_ok=True)
+            self._unpack_archive(archive_bytes, dest)
+            console.print(
+                f"[green]Installed [bold]{ref}[/bold]{' (' + version + ')' if version else ''} to [bold]{dest}[/bold].[/green]"
+            )
+        else:
+            try:
+                from crewai.skills.cache import SkillCacheManager
+
+                cache = SkillCacheManager()
+                cache.store(org, name, version, archive_bytes)
+            except ImportError:
+                # Fallback if SDK not installed — write directly
+                cache_dir = Path.home() / ".crewai" / "skills" / org / name
+                if cache_dir.exists():
+                    import shutil
+
+                    shutil.rmtree(cache_dir)
+                cache_dir.mkdir(parents=True, exist_ok=True)
+                self._unpack_archive(archive_bytes, cache_dir)
+                # Write metadata so `crewai skill list` can discover it
+                from datetime import datetime, timezone
+
+                meta = {
+                    "org": org,
+                    "name": name,
+                    "version": version,
+                    "installed_at": datetime.now(tz=timezone.utc).isoformat(),
+                }
+                (cache_dir / ".crewai_meta.json").write_text(json.dumps(meta, indent=2))
+            console.print(
+                f"[green]Installed [bold]{ref}[/bold]{' (' + version + ')' if version else ''} to global cache.[/green]"
+            )
+
+    # ------------------------------------------------------------------
+    # publish
+    # ------------------------------------------------------------------
+
+    def publish(self, is_public: bool, org: str | None, force: bool = False) -> None:
+        """Publish the skill in the current directory to the registry."""
+        skill_md = Path("SKILL.md")
+        if not skill_md.exists():
+            console.print(
+                "[red]No SKILL.md found in current directory. "
+                "Run this command from inside a skill directory.[/red]"
+            )
+            raise SystemExit(1)
+
+        # Parse frontmatter to extract name + version
+        try:
+            frontmatter = self._parse_frontmatter(skill_md.read_text())
+        except ValueError as exc:
+            console.print(f"[red]Failed to parse SKILL.md frontmatter: {exc}[/red]")
+            raise SystemExit(1) from exc
+
+        name = frontmatter.get("name")
+        version = frontmatter.get("version")
+        description = frontmatter.get("description")
+
+        if not name:
+            console.print(
+                "[red]SKILL.md frontmatter must include a 'name' field.[/red]"
+            )
+            raise SystemExit(1)
+
+        if not version:
+            console.print(
+                "[red]SKILL.md frontmatter must include a 'version' field before publishing.[/red]"
+            )
+            raise SystemExit(1)
+
+        settings = Settings()
+        effective_org = org or settings.org_name
+        if not effective_org:
+            console.print(
+                "[red]No organisation set. Run `crewai org switch <org_id>` first, "
+                "or pass --org.[/red]"
+            )
+            raise SystemExit(1)
+
+        self._print_current_organization()
+        console.print(
+            f"[bold blue]Publishing skill [bold]{name}[/bold] v{version} to {effective_org}...[/bold blue]"
+        )
+
+        archive_bytes = self._build_skill_tarball()
+        encoded_file = "data:application/x-gzip;base64," + base64.b64encode(
+            archive_bytes
+        ).decode("utf-8")
+
+        response = self.plus_api_client.publish_skill(
+            org=effective_org,
+            name=name,
+            version=version,
+            is_public=is_public,
+            description=description,
+            encoded_file=encoded_file,
+        )
+
+        self._validate_response(response)
+
+        base_url = settings.enterprise_base_url or DEFAULT_CREWAI_ENTERPRISE_URL
+        console.print(
+            f"[green]Published [bold]{effective_org}/{name}[/bold] v{version}.\n\n"
+            "Security checks are running in the background. "
+            "Your skill will be available once checks complete.\n"
+            f"Monitor status at: {base_url}/crewai_plus/skills/{effective_org}/{name}[/green]"
+        )
+
+    # ------------------------------------------------------------------
+    # list_cached
+    # ------------------------------------------------------------------
+
+    def list_cached(self) -> None:
+        """Show locally installed skills."""
+        table = Table(title="Installed Skills", show_lines=True)
+        table.add_column("Source", style="dim")
+        table.add_column("Ref")
+        table.add_column("Version")
+        table.add_column("Path")
+
+        # Project-local ./skills/
+        local_skills_dir = Path("skills")
+        if local_skills_dir.is_dir():
+            for skill_dir in sorted(local_skills_dir.iterdir()):
+                if skill_dir.is_dir() and (skill_dir / "SKILL.md").exists():
+                    version = self._read_version(skill_dir / "SKILL.md")
+                    table.add_row(
+                        "project",
+                        skill_dir.name,
+                        version or "-",
+                        str(skill_dir),
+                    )
+
+        # Global cache
+        cache_root = Path.home() / ".crewai" / "skills"
+        if cache_root.exists():
+            for org_dir in sorted(cache_root.iterdir()):
+                if not org_dir.is_dir():
+                    continue
+                for skill_dir in sorted(org_dir.iterdir()):
+                    meta_file = skill_dir / ".crewai_meta.json"
+                    if meta_file.exists():
+                        try:
+                            meta = json.loads(meta_file.read_text())
+                            table.add_row(
+                                "cache",
+                                f"@{meta['org']}/{meta['name']}",
+                                meta.get("version") or "-",
+                                str(skill_dir),
+                            )
+                        except (json.JSONDecodeError, KeyError):
+                            console.print(
+                                f"[yellow]Warning: skipping malformed cache entry at {meta_file}[/yellow]"
+                            )
+
+        console.print(table)
+
+    # ------------------------------------------------------------------
+    # internal helpers
+    # ------------------------------------------------------------------
+
+    def _print_current_organization(self) -> None:
+        settings = Settings()
+        if settings.org_uuid:
+            console.print(
+                f"Current organization: {settings.org_name} ({settings.org_uuid})",
+                style="bold blue",
+            )
+        else:
+            console.print(
+                "No organization currently set. We recommend setting one before using: "
+                "`crewai org switch <org_id>` command.",
+                style="yellow",
+            )
+
+    def _unpack_archive(self, archive_bytes: bytes, dest: Path) -> None:
+        """Unpack a .tar.gz or .zip archive into dest."""
+        # Try tar first, then zip
+        try:
+            with tarfile.open(fileobj=io.BytesIO(archive_bytes), mode="r:gz") as tf:
+                try:
+                    tf.extractall(dest, filter="data")
+                except TypeError:
+                    _safe_extractall(tf, dest)
+            return
+        except tarfile.TarError:
+            pass
+
+        # Fallback: zip
+        with zipfile.ZipFile(io.BytesIO(archive_bytes)) as zf:
+            _safe_extract_zip(zf, dest)
+
+    def _build_skill_tarball(self) -> bytes:
+        """Build an in-memory .tar.gz of SKILL.md + scripts/ + references/ + assets/."""
+        buf = io.BytesIO()
+        with tarfile.open(fileobj=buf, mode="w:gz") as tf:
+            tf.add("SKILL.md")
+            for folder in ("scripts", "references", "assets"):
+                folder_path = Path(folder)
+                if folder_path.is_dir():
+                    for fpath in sorted(folder_path.rglob("*")):
+                        if fpath.is_file():
+                            tf.add(str(fpath))
+        return buf.getvalue()
+
+    def _parse_frontmatter(self, content: str) -> dict[str, str]:
+        """Extract YAML frontmatter fields from a SKILL.md string.
+
+        Reuses crewai.skills.parser when available, with a minimal
+        fallback for environments where the full SDK isn't installed.
+        """
+        try:
+            from crewai.skills.parser import parse_frontmatter
+
+            fm_dict, _ = parse_frontmatter(content)
+            return fm_dict
+        except ImportError:
+            pass
+
+        # Fallback: minimal YAML parsing without SDK dependency
+        import re
+
+        match = re.match(r"^---\n(.*?)\n---", content, re.DOTALL)
+        if not match:
+            raise ValueError("No YAML frontmatter block found")
+        try:
+            import yaml
+
+            return yaml.safe_load(match.group(1)) or {}
+        except ImportError:
+            result: dict[str, str] = {}
+            for line in match.group(1).splitlines():
+                if ":" in line:
+                    key, _, value = line.partition(":")
+                    result[key.strip()] = value.strip()
+            return result
+
+    def _read_version(self, skill_md: Path) -> str | None:
+        """Read the version field from a SKILL.md file, or None."""
+        try:
+            fm = self._parse_frontmatter(skill_md.read_text())
+            return fm.get("version")
+        except Exception:
+            return None
+
+
+def _safe_extractall(tf: tarfile.TarFile, dest: Path) -> None:
+    """Path-traversal-safe extraction for Python < 3.12."""
+    dest_resolved = dest.resolve()
+    for member in tf.getmembers():
+        member_path = (dest / member.name).resolve()
+        if not member_path.is_relative_to(dest_resolved):
+            raise ValueError(f"Blocked path traversal attempt: {member.name!r}")
+    tf.extractall(dest)  # noqa: S202
+
+
+def _safe_extract_zip(zf: zipfile.ZipFile, dest: Path) -> None:
+    """Path-traversal-safe ZIP extraction."""
+    dest_resolved = dest.resolve()
+    for member in zf.namelist():
+        member_path = (dest / member).resolve()
+        if not member_path.is_relative_to(dest_resolved):
+            raise ValueError(f"Blocked path traversal attempt: {member!r}")
+    zf.extractall(dest)  # noqa: S202

lib/cli/tests/skills/test_main.py

@@ -0,0 +1,205 @@
+"""Tests for SkillCommand CLI."""
+
+from __future__ import annotations
+
+import io
+import os
+import tempfile
+import zipfile
+from contextlib import contextmanager
+from datetime import datetime, timedelta
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from crewai_cli.shared.token_manager import TokenManager
+
+
+@contextmanager
+def in_temp_dir():
+    original = os.getcwd()
+    with tempfile.TemporaryDirectory() as td:
+        os.chdir(td)
+        try:
+            yield td
+        finally:
+            os.chdir(original)
+
+
+@pytest.fixture
+def skill_command():
+    with tempfile.TemporaryDirectory() as temp_dir:
+        with patch.object(
+            TokenManager, "_get_secure_storage_path", return_value=Path(temp_dir)
+        ):
+            TokenManager().save_tokens(
+                "test-token", (datetime.now() + timedelta(seconds=36000)).timestamp()
+            )
+            from crewai_cli.skills.main import SkillCommand
+            cmd = SkillCommand()
+            yield cmd
+
+
+# ---------------------------------------------------------------------------
+# create
+# ---------------------------------------------------------------------------
+
+class TestSkillCreate:
+    def test_create_in_project(self, skill_command, tmp_path):
+        with in_temp_dir():
+            # Simulate being inside a project
+            Path("pyproject.toml").write_text("[tool.poetry]\nname = 'test'\n")
+            skill_command.create("my-skill")
+            assert Path("skills/my-skill/SKILL.md").exists()
+            assert Path("skills/my-skill/scripts").is_dir()
+            assert Path("skills/my-skill/references").is_dir()
+            assert Path("skills/my-skill/assets").is_dir()
+
+    def test_create_outside_project(self, skill_command, tmp_path):
+        with in_temp_dir():
+            skill_command.create("standalone-skill", in_project=False)
+            assert Path("standalone-skill/SKILL.md").exists()
+
+    def test_create_adds_name_to_skill_md(self, skill_command):
+        with in_temp_dir():
+            skill_command.create("hello-world", in_project=False)
+            content = Path("hello-world/SKILL.md").read_text()
+            assert "name: hello-world" in content
+            assert "version: 0.1.0" in content
+
+    def test_create_fails_if_dir_exists(self, skill_command):
+        with in_temp_dir():
+            Path("existing-skill").mkdir()
+            with pytest.raises(SystemExit):
+                skill_command.create("existing-skill", in_project=False)
+
+
+# ---------------------------------------------------------------------------
+# install
+# ---------------------------------------------------------------------------
+
+class TestSkillInstall:
+    def _zip_skill(self, name: str) -> bytes:
+        buf = io.BytesIO()
+        with zipfile.ZipFile(buf, "w") as zf:
+            zf.writestr("SKILL.md", f"---\nname: {name}\ndescription: Test.\n---\nInstructions.")
+        return buf.getvalue()
+
+    def test_install_invalid_ref_no_at(self, skill_command):
+        with pytest.raises(SystemExit):
+            skill_command.install("acme/my-skill")
+
+    def test_install_invalid_ref_no_slash(self, skill_command):
+        with pytest.raises(SystemExit):
+            skill_command.install("@acmeskill")
+
+    def test_install_404(self, skill_command):
+        mock_resp = MagicMock()
+        mock_resp.status_code = 404
+        skill_command.plus_api_client.get_skill = MagicMock(return_value=mock_resp)
+
+        with pytest.raises(SystemExit):
+            skill_command.install("@acme/ghost")
+
+    def test_install_in_project(self, skill_command):
+        import base64
+        archive = self._zip_skill("my-skill")
+        encoded = "data:application/zip;base64," + base64.b64encode(archive).decode()
+
+        mock_resp = MagicMock()
+        mock_resp.status_code = 200
+        mock_resp.json.return_value = {"file": encoded, "version": "1.0.0"}
+        skill_command.plus_api_client.get_skill = MagicMock(return_value=mock_resp)
+
+        with in_temp_dir():
+            Path("pyproject.toml").write_text("[tool]\n")
+            skill_command.install("@acme/my-skill")
+            assert Path("skills/my-skill/SKILL.md").exists()
+
+
+# ---------------------------------------------------------------------------
+# publish
+# ---------------------------------------------------------------------------
+
+class TestSkillPublish:
+    def test_publish_no_skill_md(self, skill_command):
+        with in_temp_dir():
+            with pytest.raises(SystemExit):
+                skill_command.publish(is_public=True, org="acme")
+
+    def test_publish_missing_version(self, skill_command):
+        with in_temp_dir():
+            Path("SKILL.md").write_text(
+                "---\nname: my-skill\ndescription: Test.\n---\nInstructions."
+            )
+            with pytest.raises(SystemExit):
+                skill_command.publish(is_public=True, org="acme")
+
+    def test_publish_missing_name(self, skill_command):
+        with in_temp_dir():
+            Path("SKILL.md").write_text(
+                "---\ndescription: Test.\nversion: 1.0.0\n---\nInstructions."
+            )
+            with pytest.raises(SystemExit):
+                skill_command.publish(is_public=True, org="acme")
+
+    def test_publish_no_org(self, skill_command):
+        with in_temp_dir():
+            Path("SKILL.md").write_text(
+                "---\nname: my-skill\nversion: 1.0.0\ndescription: Test.\n---\nInstructions."
+            )
+            with patch.object(skill_command, "plus_api_client") as mock_client:
+                mock_resp = MagicMock()
+                mock_resp.is_success = True
+                mock_resp.status_code = 200
+                mock_resp.json.return_value = {}
+                mock_client.publish_skill.return_value = mock_resp
+                # No org set → should SystemExit (no org_name in settings)
+                with patch("crewai_cli.skills.main.Settings") as mock_settings_cls:
+                    mock_settings_cls.return_value.org_name = None
+                    mock_settings_cls.return_value.enterprise_base_url = None
+                    with pytest.raises(SystemExit):
+                        skill_command.publish(is_public=True, org=None)
+
+    def test_publish_calls_api(self, skill_command):
+        with in_temp_dir():
+            Path("SKILL.md").write_text(
+                "---\nname: my-skill\nversion: 1.0.0\ndescription: A test skill.\n---\nInstructions."
+            )
+            mock_resp = MagicMock()
+            mock_resp.is_success = True
+            mock_resp.status_code = 200
+            mock_resp.json.return_value = {}
+            skill_command.plus_api_client.publish_skill = MagicMock(return_value=mock_resp)
+            with patch("crewai_cli.skills.main.Settings") as mock_settings_cls:
+                mock_settings_cls.return_value.org_name = "acme"
+                mock_settings_cls.return_value.enterprise_base_url = None
+
+                skill_command.publish(is_public=False, org="acme")
+
+            skill_command.plus_api_client.publish_skill.assert_called_once()
+            call_kwargs = skill_command.plus_api_client.publish_skill.call_args
+            assert call_kwargs.kwargs["name"] == "my-skill"
+            assert call_kwargs.kwargs["version"] == "1.0.0"
+
+
+# ---------------------------------------------------------------------------
+# list_cached
+# ---------------------------------------------------------------------------
+
+class TestSkillListCached:
+    def test_list_cached_empty(self, skill_command, capsys):
+        with in_temp_dir():
+            skill_command.list_cached()
+            # Should not raise
+
+    def test_list_cached_shows_project_skills(self, skill_command, capsys):
+        with in_temp_dir():
+            skill_dir = Path("skills/my-skill")
+            skill_dir.mkdir(parents=True)
+            (skill_dir / "SKILL.md").write_text(
+                "---\nname: my-skill\nversion: 0.5.0\ndescription: A skill.\n---\nBody."
+            )
+            skill_command.list_cached()
+            # Should complete without error

lib/crewai-core/src/crewai_core/__init__.py

@@ -1 +1 @@
-__version__ = "1.14.5a3"
+__version__ = "1.14.5"

lib/crewai-core/src/crewai_core/plus_api.py

@@ -3,36 +3,162 @@
 from __future__ import annotations
 
 import os
-from typing import Any
+from typing import Any, Final, Literal, TypedDict, cast
 from urllib.parse import urljoin
 
 import httpx
+from typing_extensions import NotRequired
 
 from crewai_core.constants import DEFAULT_CREWAI_ENTERPRISE_URL
 from crewai_core.settings import Settings
 from crewai_core.version import get_crewai_version
 
 
+HttpMethod = Literal["GET", "POST", "PATCH", "DELETE"]
+
+
+class AvailableExport(TypedDict):
+    name: str
+
+
+class EnvVarEntry(TypedDict):
+    name: str
+    description: str
+    required: bool
+    default: str | None
+
+
+class ToolMetadata(TypedDict):
+    name: str
+    module: str
+    humanized_name: str
+    description: str
+    run_params_schema: dict[str, Any]
+    init_params_schema: dict[str, Any]
+    env_vars: list[EnvVarEntry]
+
+
+class ToolsMetadataPayload(TypedDict):
+    package: str
+    tools: list[ToolMetadata] | None
+
+
+class PublishToolPayload(TypedDict):
+    handle: str
+    public: bool
+    version: str
+    file: str
+    description: str | None
+    available_exports: list[AvailableExport] | None
+    tools_metadata: ToolsMetadataPayload | None
+
+
+class CrewDeploymentSpec(TypedDict):
+    name: str
+    repo_clone_url: str
+    env: dict[str, str]
+
+
+class CreateCrewPayload(TypedDict):
+    deploy: CrewDeploymentSpec
+
+
+class _WithUserIdentifier(TypedDict):
+    user_identifier: NotRequired[str]
+
+
+class LoginPayload(_WithUserIdentifier):
+    pass
+
+
+class TraceExecutionContext(TypedDict):
+    crew_fingerprint: str | None
+    crew_name: str | None
+    flow_name: str | None
+    crewai_version: str
+    privacy_level: str
+
+
+class TraceExecutionMetadata(TypedDict):
+    expected_duration_estimate: int
+    agent_count: int
+    task_count: int
+    flow_method_count: int
+    execution_started_at: str
+
+
+class TraceBatchInitPayload(_WithUserIdentifier):
+    trace_id: str
+    execution_type: str
+    execution_context: TraceExecutionContext
+    execution_metadata: TraceExecutionMetadata
+    ephemeral_trace_id: NotRequired[str]
+
+
+class TraceBatchMetadata(TypedDict):
+    events_count: int
+    batch_sequence: int
+    is_final_batch: bool
+
+
+class TraceEventsPayload(TypedDict):
+    events: list[dict[str, Any]]
+    batch_metadata: TraceBatchMetadata
+
+
+class TraceFinalizePayload(TypedDict):
+    status: Literal["completed"]
+    duration_ms: float | None
+    final_event_count: int
+
+
+class TraceFailedPayload(TypedDict):
+    status: Literal["failed"]
+    failure_reason: str
+
+
+Headers = TypedDict(
+    "Headers",
+    {
+        "Content-Type": str,
+        "User-Agent": str,
+        "X-Crewai-Version": str,
+        "Authorization": NotRequired[str],
+        "X-Crewai-Organization-Id": NotRequired[str],
+    },
+)
+
+
+class RequestKwargs(TypedDict):
+    headers: dict[str, str]
+    json: NotRequired[Any]
+    params: NotRequired[dict[str, str]]
+    timeout: NotRequired[float]
+
+
 class PlusAPI:
     """Client for working with the CrewAI+ API."""
 
-    TOOLS_RESOURCE = "/crewai_plus/api/v1/tools"
-    ORGANIZATIONS_RESOURCE = "/crewai_plus/api/v1/me/organizations"
-    CREWS_RESOURCE = "/crewai_plus/api/v1/crews"
-    AGENTS_RESOURCE = "/crewai_plus/api/v1/agents"
-    TRACING_RESOURCE = "/crewai_plus/api/v1/tracing"
-    EPHEMERAL_TRACING_RESOURCE = "/crewai_plus/api/v1/tracing/ephemeral"
-    INTEGRATIONS_RESOURCE = "/crewai_plus/api/v1/integrations"
+    TOOLS_RESOURCE: Final = "/crewai_plus/api/v1/tools"
+    SKILLS_RESOURCE: Final = "/crewai_plus/api/v1/skills"
+    ORGANIZATIONS_RESOURCE: Final = "/crewai_plus/api/v1/me/organizations"
+    CREWS_RESOURCE: Final = "/crewai_plus/api/v1/crews"
+    AGENTS_RESOURCE: Final = "/crewai_plus/api/v1/agents"
+    TRACING_RESOURCE: Final = "/crewai_plus/api/v1/tracing"
+    EPHEMERAL_TRACING_RESOURCE: Final = "/crewai_plus/api/v1/tracing/ephemeral"
+    INTEGRATIONS_RESOURCE: Final = "/crewai_plus/api/v1/integrations"
 
     def __init__(self, api_key: str | None = None) -> None:
+        version = get_crewai_version()
         self.api_key = api_key
-        self.headers = {
+        self.headers: Headers = {
             "Content-Type": "application/json",
-            "User-Agent": f"CrewAI-CLI/{get_crewai_version()}",
-            "X-Crewai-Version": get_crewai_version(),
+            "User-Agent": f"CrewAI-CLI/{version}",
+            "X-Crewai-Version": version,
         }
         if api_key:
             self.headers["Authorization"] = f"Bearer {api_key}"
+
         settings = Settings()
         if settings.org_uuid:
             self.headers["X-Crewai-Organization-Id"] = settings.org_uuid
@@ -44,17 +170,30 @@ class PlusAPI:
         )
 
     def _make_request(
-        self, method: str, endpoint: str, **kwargs: Any
+        self,
+        method: HttpMethod,
+        endpoint: str,
+        *,
+        json: Any = None,
+        params: dict[str, str] | None = None,
+        timeout: float | None = None,
+        verify: bool = True,
     ) -> httpx.Response:
         url = urljoin(self.base_url, endpoint)
-        verify = kwargs.pop("verify", True)
+        request_kwargs: RequestKwargs = {"headers": cast(dict[str, str], self.headers)}
+        if json is not None:
+            request_kwargs["json"] = json
+        if params is not None:
+            request_kwargs["params"] = params
+        if timeout is not None:
+            request_kwargs["timeout"] = timeout
         with httpx.Client(trust_env=False, verify=verify) as client:
-            return client.request(method, url, headers=self.headers, **kwargs)
+            return client.request(method, url, **request_kwargs)
 
     def login_to_tool_repository(
         self, user_identifier: str | None = None
     ) -> httpx.Response:
-        payload = {}
+        payload: LoginPayload = {}
         if user_identifier:
             payload["user_identifier"] = user_identifier
         return self._make_request("POST", f"{self.TOOLS_RESOURCE}/login", json=payload)
@@ -65,7 +204,7 @@ class PlusAPI:
     async def get_agent(self, handle: str) -> httpx.Response:
         url = urljoin(self.base_url, f"{self.AGENTS_RESOURCE}/{handle}")
         async with httpx.AsyncClient() as client:
-            return await client.get(url, headers=self.headers)
+            return await client.get(url, headers=cast(dict[str, str], self.headers))
 
     def publish_tool(
         self,
@@ -74,10 +213,10 @@ class PlusAPI:
         version: str,
         description: str | None,
         encoded_file: str,
-        available_exports: list[dict[str, Any]] | None = None,
-        tools_metadata: list[dict[str, Any]] | None = None,
+        available_exports: list[AvailableExport] | None = None,
+        tools_metadata: list[ToolMetadata] | None = None,
     ) -> httpx.Response:
-        params = {
+        params: PublishToolPayload = {
             "handle": handle,
             "public": is_public,
             "version": version,
@@ -90,6 +229,47 @@ class PlusAPI:
         }
         return self._make_request("POST", f"{self.TOOLS_RESOURCE}", json=params)
 
+    def get_skill(
+        self, org: str, name: str, version: str | None = None
+    ) -> httpx.Response:
+        params: dict[str, str] = {}
+        if version is not None:
+            params["version"] = version
+        return self._make_request(
+            "GET",
+            f"{self.SKILLS_RESOURCE}/{org}/{name}",
+            params=params or None,
+        )
+
+    def publish_skill(
+        self,
+        org: str,
+        name: str,
+        version: str,
+        is_public: bool,
+        description: str | None,
+        encoded_file: str,
+    ) -> httpx.Response:
+        payload = {
+            "org": org,
+            "name": name,
+            "version": version,
+            "public": is_public,
+            "description": description,
+            "file": encoded_file,
+        }
+        return self._make_request("POST", self.SKILLS_RESOURCE, json=payload)
+
+    def list_skills(self, org: str | None = None) -> httpx.Response:
+        params: dict[str, str] = {}
+        if org is not None:
+            params["org"] = org
+        return self._make_request(
+            "GET",
+            self.SKILLS_RESOURCE,
+            params=params or None,
+        )
+
     def deploy_by_name(self, project_name: str) -> httpx.Response:
         return self._make_request(
             "POST", f"{self.CREWS_RESOURCE}/by-name/{project_name}/deploy"
@@ -129,13 +309,13 @@ class PlusAPI:
     def list_crews(self) -> httpx.Response:
         return self._make_request("GET", self.CREWS_RESOURCE)
 
-    def create_crew(self, payload: dict[str, Any]) -> httpx.Response:
+    def create_crew(self, payload: CreateCrewPayload) -> httpx.Response:
         return self._make_request("POST", self.CREWS_RESOURCE, json=payload)
 
     def get_organizations(self) -> httpx.Response:
         return self._make_request("GET", self.ORGANIZATIONS_RESOURCE)
 
-    def initialize_trace_batch(self, payload: dict[str, Any]) -> httpx.Response:
+    def initialize_trace_batch(self, payload: TraceBatchInitPayload) -> httpx.Response:
         return self._make_request(
             "POST",
             f"{self.TRACING_RESOURCE}/batches",
@@ -144,7 +324,7 @@ class PlusAPI:
         )
 
     def initialize_ephemeral_trace_batch(
-        self, payload: dict[str, Any]
+        self, payload: TraceBatchInitPayload
     ) -> httpx.Response:
         return self._make_request(
             "POST",
@@ -153,7 +333,7 @@ class PlusAPI:
         )
 
     def send_trace_events(
-        self, trace_batch_id: str, payload: dict[str, Any]
+        self, trace_batch_id: str, payload: TraceEventsPayload
     ) -> httpx.Response:
         return self._make_request(
             "POST",
@@ -163,7 +343,7 @@ class PlusAPI:
         )
 
     def send_ephemeral_trace_events(
-        self, trace_batch_id: str, payload: dict[str, Any]
+        self, trace_batch_id: str, payload: TraceEventsPayload
     ) -> httpx.Response:
         return self._make_request(
             "POST",
@@ -173,7 +353,7 @@ class PlusAPI:
         )
 
     def finalize_trace_batch(
-        self, trace_batch_id: str, payload: dict[str, Any]
+        self, trace_batch_id: str, payload: TraceFinalizePayload
     ) -> httpx.Response:
         return self._make_request(
             "PATCH",
@@ -183,7 +363,7 @@ class PlusAPI:
         )
 
     def finalize_ephemeral_trace_batch(
-        self, trace_batch_id: str, payload: dict[str, Any]
+        self, trace_batch_id: str, payload: TraceFinalizePayload
     ) -> httpx.Response:
         return self._make_request(
             "PATCH",
@@ -195,20 +375,28 @@ class PlusAPI:
     def mark_trace_batch_as_failed(
         self, trace_batch_id: str, error_message: str
     ) -> httpx.Response:
+        payload: TraceFailedPayload = {
+            "status": "failed",
+            "failure_reason": error_message,
+        }
         return self._make_request(
             "PATCH",
             f"{self.TRACING_RESOURCE}/batches/{trace_batch_id}",
-            json={"status": "failed", "failure_reason": error_message},
+            json=payload,
             timeout=30,
         )
 
     def mark_ephemeral_trace_batch_as_failed(
         self, trace_batch_id: str, error_message: str
     ) -> httpx.Response:
+        payload: TraceFailedPayload = {
+            "status": "failed",
+            "failure_reason": error_message,
+        }
         return self._make_request(
             "PATCH",
             f"{self.EPHEMERAL_TRACING_RESOURCE}/batches/{trace_batch_id}",
-            json={"status": "failed", "failure_reason": error_message},
+            json=payload,
             timeout=30,
         )
 

lib/crewai-files/src/crewai_files/__init__.py

@@ -152,4 +152,4 @@ __all__ = [
     "wrap_file_source",
 ]
 
-__version__ = "1.14.5a3"
+__version__ = "1.14.5"

lib/crewai-tools/pyproject.toml

@@ -10,7 +10,7 @@ requires-python = ">=3.10, <3.14"
 dependencies = [
     "pytube~=15.0.0",
     "requests>=2.33.0,<3",
-    "crewai==1.14.5a3",
+    "crewai==1.14.5",
     "tiktoken>=0.8.0,<0.13",
     "beautifulsoup4~=4.13.4",
     "python-docx~=1.2.0",
@@ -107,7 +107,7 @@ stagehand = [
     "stagehand>=0.4.1",
 ]
 github = [
-    "gitpython>=3.1.47,<4",
+    "gitpython>=3.1.50,<4",
     "PyGithub==1.59.1",
 ]
 rag = [

lib/crewai-tools/src/crewai_tools/__init__.py

@@ -330,4 +330,4 @@ __all__ = [
     "ZapierActionTools",
 ]
 
-__version__ = "1.14.5a3"
+__version__ = "1.14.5"

lib/crewai-tools/src/crewai_tools/tools/daytona_sandbox_tool/README.md

@@ -55,10 +55,11 @@ from crewai_tools import DaytonaExecTool, DaytonaFileTool
 exec_tool = DaytonaExecTool(persistent=True)
 file_tool = DaytonaFileTool(persistent=True)
 
-# Agent writes a script, then runs it — both share the same sandbox instance
-# because they each keep their own persistent sandbox. If you need the *same*
-# sandbox across two tools, create one tool, grab the sandbox id via
-# `tool._persistent_sandbox.id`, and pass it to the other via `sandbox_id=...`.
+# Agent writes a script, then runs it — but each tool keeps its OWN persistent
+# sandbox. To share the *same* sandbox across two tools, create and use the
+# first tool, then read its `active_sandbox_id` and pass it to the second:
+#   exec_tool.run(command="pip install httpx")
+#   file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
 ```
 
 ### Attach to an existing sandbox
@@ -99,9 +100,14 @@ tool = DaytonaExecTool(
 - `timeout: int | None` — seconds.
 
 ### `DaytonaFileTool`
-- `action: "read" | "write" | "list" | "delete" | "mkdir" | "info"`
-- `path: str` — absolute path inside the sandbox.
-- `content: str | None` — required for `write`.
+- `action`: one of `read`, `write`, `append`, `list`, `delete`, `mkdir`, `info`, `exists`, `move`, `find`, `search`, `chmod`, `replace`.
+- `path: str | None` — absolute path inside the sandbox. Required for all actions except `replace`.
+- `content: str | None` — required for `append`; optional for `write`.
 - `binary: bool` — if `True`, `content` is base64 on write / returned as base64 on read.
 - `recursive: bool` — for `delete`, removes directories recursively.
-- `mode: str` — for `mkdir`, octal permission string (default `"0755"`).
+- `mode: str | None` — for `mkdir` (defaults to `"0755"`) or for `chmod` (e.g. `"755"`).
+- `destination: str | None` — required for `move`.
+- `pattern: str | None` — required for `find` (content grep), `search` (filename glob), and `replace`.
+- `replacement: str | None` — required for `replace`.
+- `paths: list[str] | None` — required for `replace`; list of files to operate on.
+- `owner: str | None` / `group: str | None` — for `chmod`. Pass at least one of `mode`, `owner`, or `group`.

lib/crewai-tools/src/crewai_tools/tools/daytona_sandbox_tool/daytona_base_tool.py

@@ -196,3 +196,27 @@ class DaytonaBaseTool(BaseTool):
                 "the sandbox may need manual deletion.",
                 exc_info=True,
             )
+
+    @property
+    def active_sandbox_id(self) -> str | None:
+        """The id of the sandbox this tool is currently bound to, if any.
+
+        Returns:
+          - the explicitly attached `sandbox_id`, if set at construction;
+          - the id of the lazily-created persistent sandbox, once a call has
+            triggered creation;
+          - None for ephemeral mode (where no sandbox lives between calls).
+
+        Use this to share one sandbox across multiple tool instances:
+
+            exec_tool = DaytonaExecTool(persistent=True)
+            exec_tool.run(command="pip install httpx")
+            file_tool = DaytonaFileTool(sandbox_id=exec_tool.active_sandbox_id)
+        """
+        if self.sandbox_id:
+            return self.sandbox_id
+        with self._lock:
+            sandbox = self._persistent_sandbox
+        if sandbox is None:
+            return None
+        return getattr(sandbox, "id", None)

lib/crewai-tools/src/crewai_tools/tools/daytona_sandbox_tool/daytona_file_tool.py

@@ -4,7 +4,9 @@ import base64
 from builtins import type as type_
 import logging
 import posixpath
+import shlex
 from typing import Any, Literal
+import uuid
 
 from pydantic import BaseModel, Field, model_validator
 
@@ -14,22 +16,110 @@ from crewai_tools.tools.daytona_sandbox_tool.daytona_base_tool import DaytonaBas
 logger = logging.getLogger(__name__)
 
 
-FileAction = Literal["read", "write", "append", "list", "delete", "mkdir", "info"]
+FileAction = Literal[
+    "read",
+    "write",
+    "append",
+    "list",
+    "delete",
+    "mkdir",
+    "info",
+    "exists",
+    "move",
+    "find",
+    "search",
+    "chmod",
+    "replace",
+]
+
+
+def _daytona_file_schema_extra(schema: dict[str, Any]) -> None:
+    schema["allOf"] = [
+        {
+            "if": {
+                "properties": {
+                    "action": {
+                        "enum": [
+                            "read",
+                            "write",
+                            "append",
+                            "list",
+                            "delete",
+                            "mkdir",
+                            "info",
+                            "exists",
+                            "move",
+                            "find",
+                            "search",
+                            "chmod",
+                        ]
+                    }
+                }
+            },
+            "then": {"required": ["path"]},
+        },
+        {
+            "if": {"properties": {"action": {"const": "append"}}},
+            "then": {"required": ["content"]},
+        },
+        {
+            "if": {"properties": {"action": {"const": "move"}}},
+            "then": {"required": ["destination"]},
+        },
+        {
+            "if": {"properties": {"action": {"enum": ["find", "search"]}}},
+            "then": {"required": ["pattern"]},
+        },
+        {
+            "if": {"properties": {"action": {"const": "replace"}}},
+            "then": {"required": ["paths", "pattern", "replacement"]},
+        },
+        {
+            "if": {"properties": {"action": {"const": "chmod"}}},
+            "then": {
+                "anyOf": [
+                    {"required": ["mode"]},
+                    {"required": ["owner"]},
+                    {"required": ["group"]},
+                ]
+            },
+        },
+    ]
 
 
 class DaytonaFileToolSchema(BaseModel):
+    model_config = {"json_schema_extra": _daytona_file_schema_extra}
+
     action: FileAction = Field(
         ...,
         description=(
-            "The filesystem action to perform: 'read' (returns file contents), "
-            "'write' (create or replace a file with content), 'append' (append "
-            "content to an existing file — use this for writing large files in "
-            "chunks to avoid hitting tool-call size limits), 'list' (lists a "
-            "directory), 'delete' (removes a file/dir), 'mkdir' (creates a "
-            "directory), 'info' (returns file metadata)."
+            "The filesystem action to perform: "
+            "'read' (returns file contents); "
+            "'write' (create or replace a file with content); "
+            "'append' (append content to an existing file — use this for "
+            "writing large files in chunks to avoid hitting tool-call size "
+            "limits); "
+            "'list' (lists a directory); "
+            "'delete' (removes a file/dir); "
+            "'mkdir' (creates a directory); "
+            "'info' (returns file metadata); "
+            "'exists' (returns whether a path exists); "
+            "'move' (rename or relocate a file/dir; requires 'destination'); "
+            "'find' (grep file CONTENTS recursively; requires 'pattern'); "
+            "'search' (find files by NAME pattern; requires 'pattern'); "
+            "'chmod' (change permissions/owner/group; pass at least one of "
+            "'mode', 'owner', 'group'); "
+            "'replace' (find-and-replace text across files; requires "
+            "'paths', 'pattern', and 'replacement')."
+        ),
+    )
+    path: str | None = Field(
+        default=None,
+        description=(
+            "Absolute path inside the sandbox. Required for all actions "
+            "except 'replace' (which uses 'paths' instead)."
         ),
     )
-    path: str = Field(..., description="Absolute path inside the sandbox.")
     content: str | None = Field(
         default=None,
         description=(
@@ -50,18 +140,78 @@ class DaytonaFileToolSchema(BaseModel):
         default=False,
         description="For action='delete': remove directories recursively.",
     )
-    mode: str = Field(
-        default="0755",
-        description="For action='mkdir': octal permission string (default 0755).",
+    mode: str | None = Field(
+        default=None,
+        description=(
+            "Octal permission string. For 'mkdir' it sets the new directory "
+            "permissions (defaults to '0755' if omitted). For 'chmod' it sets "
+            "the target's mode (e.g. '755' to make a script executable). "
+            "Ignored for other actions."
+        ),
+    )
+    destination: str | None = Field(
+        default=None,
+        description="For action='move': absolute destination path.",
+    )
+    pattern: str | None = Field(
+        default=None,
+        description=(
+            "For 'find': substring matched against file CONTENTS. "
+            "For 'search': glob-style pattern matched against file NAMES "
+            "(e.g. '*.py'). "
+            "For 'replace': text to replace inside files."
+        ),
+    )
+    replacement: str | None = Field(
+        default=None,
+        description="For action='replace': replacement text for 'pattern'.",
+    )
+    paths: list[str] | None = Field(
+        default=None,
+        description=(
+            "For action='replace': list of absolute file paths in which to "
+            "replace 'pattern' with 'replacement'."
+        ),
+    )
+    owner: str | None = Field(
+        default=None,
+        description="For action='chmod': new file owner (user name).",
+    )
+    group: str | None = Field(
+        default=None,
+        description="For action='chmod': new file group.",
     )
 
     @model_validator(mode="after")
     def _validate_action_args(self) -> DaytonaFileToolSchema:
+        if self.action != "replace" and not self.path:
+            raise ValueError(f"action={self.action!r} requires 'path'.")
         if self.action == "append" and self.content is None:
             raise ValueError(
                 "action='append' requires 'content'. Pass the chunk to append "
                 "in the 'content' field."
             )
+        if self.action == "move" and not self.destination:
+            raise ValueError("action='move' requires 'destination'.")
+        if self.action == "find" and not self.pattern:
+            raise ValueError(
+                "action='find' requires 'pattern' (text to search for inside files)."
+            )
+        if self.action == "search" and not self.pattern:
+            raise ValueError("action='search' requires 'pattern' (glob, e.g. '*.py').")
+        if self.action == "chmod" and not (self.mode or self.owner or self.group):
+            raise ValueError(
+                "action='chmod' requires at least one of 'mode', 'owner', or 'group'."
+            )
+        if self.action == "replace":
+            if not self.paths:
+                raise ValueError(
+                    "action='replace' requires 'paths' (list of file paths)."
+                )
+            if not self.pattern:
+                raise ValueError("action='replace' requires 'pattern'.")
+            if self.replacement is None:
+                raise ValueError("action='replace' requires 'replacement'.")
         return self
 
 
@@ -75,9 +225,10 @@ class DaytonaFileTool(DaytonaBaseTool):
 
     name: str = "Daytona Sandbox Files"
     description: str = (
-        "Perform filesystem operations inside a Daytona sandbox: read a file, "
-        "write content to a path, append content to an existing file, list a "
-        "directory, delete a path, make a directory, or fetch file metadata. "
+        "Perform filesystem operations inside a Daytona sandbox: read, "
+        "write, append, list, delete, mkdir, info, exists, move, find "
+        "(content grep), search (filename glob), chmod (permissions/owner/"
+        "group), and replace (bulk find-and-replace across files). "
         "For files larger than a few KB, create the file with action='write' "
         "and empty content, then send the body via multiple 'append' calls of "
         "~4KB each to stay within tool-call payload limits."
@@ -87,30 +238,79 @@ class DaytonaFileTool(DaytonaBaseTool):
     def _run(
         self,
         action: FileAction,
-        path: str,
+        path: str | None = None,
         content: str | None = None,
         binary: bool = False,
         recursive: bool = False,
-        mode: str = "0755",
+        mode: str | None = None,
+        destination: str | None = None,
+        pattern: str | None = None,
+        replacement: str | None = None,
+        paths: list[str] | None = None,
+        owner: str | None = None,
+        group: str | None = None,
     ) -> Any:
         sandbox, should_delete = self._acquire_sandbox()
         try:
             if action == "read":
+                if path is None:
+                    raise ValueError("action='read' requires 'path'")
                 return self._read(sandbox, path, binary=binary)
             if action == "write":
+                if path is None:
+                    raise ValueError("action='write' requires 'path'")
                 return self._write(sandbox, path, content or "", binary=binary)
             if action == "append":
+                if path is None:
+                    raise ValueError("action='append' requires 'path'")
                 return self._append(sandbox, path, content or "", binary=binary)
             if action == "list":
+                if path is None:
+                    raise ValueError("action='list' requires 'path'")
                 return self._list(sandbox, path)
             if action == "delete":
+                if path is None:
+                    raise ValueError("action='delete' requires 'path'")
                 sandbox.fs.delete_file(path, recursive=recursive)
                 return {"status": "deleted", "path": path}
             if action == "mkdir":
-                sandbox.fs.create_folder(path, mode)
-                return {"status": "created", "path": path, "mode": mode}
+                if path is None:
+                    raise ValueError("action='mkdir' requires 'path'")
+                mkdir_mode = mode or "0755"
+                sandbox.fs.create_folder(path, mkdir_mode)
+                return {"status": "created", "path": path, "mode": mkdir_mode}
             if action == "info":
+                if path is None:
+                    raise ValueError("action='info' requires 'path'")
                 return self._info(sandbox, path)
+            if action == "exists":
+                if path is None:
+                    raise ValueError("action='exists' requires 'path'")
+                return self._exists(sandbox, path)
+            if action == "move":
+                if path is None or destination is None:
+                    raise ValueError("action='move' requires 'path' and 'destination'")
+                sandbox.fs.move_files(path, destination)
+                return {"status": "moved", "from": path, "to": destination}
+            if action == "find":
+                if path is None or pattern is None:
+                    raise ValueError("action='find' requires 'path' and 'pattern'")
+                return self._find(sandbox, path, pattern)
+            if action == "search":
+                if path is None or pattern is None:
+                    raise ValueError("action='search' requires 'path' and 'pattern'")
+                return self._search(sandbox, path, pattern)
+            if action == "chmod":
+                if path is None:
+                    raise ValueError("action='chmod' requires 'path'")
+                return self._chmod(sandbox, path, mode=mode, owner=owner, group=group)
+            if action == "replace":
+                if paths is None or pattern is None or replacement is None:
+                    raise ValueError(
+                        "action='replace' requires 'paths', 'pattern', and "
+                        "'replacement'"
+                    )
+                return self._replace(sandbox, paths, pattern, replacement)
             raise ValueError(f"Unknown action: {action}")
         finally:
             self._release_sandbox(sandbox, should_delete)
@@ -146,17 +346,46 @@ class DaytonaFileTool(DaytonaBaseTool):
     ) -> dict[str, Any]:
         chunk = base64.b64decode(content) if binary else content.encode("utf-8")
         self._ensure_parent_dir(sandbox, path)
+
+        # Server-side `cat >>` keeps this O(chunk_size) per call. The naive
+        # download-concat-reupload alternative is O(N^2) in total transfer.
+        # /tmp/ is on the sandbox's ephemeral filesystem, not the host.
+        temp_path = f"/tmp/.crewai-append-{uuid.uuid4().hex}"  # noqa: S108
+        sandbox.fs.upload_file(chunk, temp_path)
+
+        quoted_temp = shlex.quote(temp_path)
+        quoted_target = shlex.quote(path)
+        response = sandbox.process.exec(
+            f"cat {quoted_temp} >> {quoted_target}; "
+            f"rc=$?; rm -f {quoted_temp}; exit $rc"
+        )
+
+        exit_code = getattr(response, "exit_code", 0)
+        if exit_code != 0:
+            try:
+                sandbox.fs.delete_file(temp_path)
+            except Exception:
+                logger.debug(
+                    "Best-effort temp-file cleanup failed after append "
+                    "error; the file may need manual deletion.",
+                    exc_info=True,
+                )
+            raise RuntimeError(
+                f"append failed: exit_code={exit_code}, "
+                f"output={getattr(response, 'result', '')!r}"
+            )
+
         try:
-            existing: bytes = sandbox.fs.download_file(path)
+            info = sandbox.fs.get_file_info(path)
+            total_bytes = getattr(info, "size", None)
         except Exception:
-            existing = b""
-        payload = existing + chunk
-        sandbox.fs.upload_file(payload, path)
+            total_bytes = None
+
         return {
             "status": "appended",
             "path": path,
             "appended_bytes": len(chunk),
-            "total_bytes": len(payload),
+            "total_bytes": total_bytes,
         }
 
     @staticmethod
@@ -190,6 +419,77 @@ class DaytonaFileTool(DaytonaBaseTool):
     def _info(self, sandbox: Any, path: str) -> dict[str, Any]:
         return self._file_info_to_dict(sandbox.fs.get_file_info(path))
 
+    def _exists(self, sandbox: Any, path: str) -> dict[str, Any]:
+        try:
+            info = sandbox.fs.get_file_info(path)
+        except Exception:
+            return {"path": path, "exists": False}
+        return {
+            "path": path,
+            "exists": True,
+            "is_dir": getattr(info, "is_dir", False),
+        }
+
+    def _find(self, sandbox: Any, path: str, pattern: str) -> dict[str, Any]:
+        matches = sandbox.fs.find_files(path, pattern)
+        return {
+            "path": path,
+            "pattern": pattern,
+            "matches": [
+                {
+                    "file": getattr(m, "file", None),
+                    "line": getattr(m, "line", None),
+                    "content": getattr(m, "content", None),
+                }
+                for m in matches
+            ],
+        }
+
+    def _search(self, sandbox: Any, path: str, pattern: str) -> dict[str, Any]:
+        response = sandbox.fs.search_files(path, pattern)
+        files = getattr(response, "files", None) or []
+        return {"path": path, "pattern": pattern, "files": list(files)}
+
+    def _chmod(
+        self,
+        sandbox: Any,
+        path: str,
+        *,
+        mode: str | None,
+        owner: str | None,
+        group: str | None,
+    ) -> dict[str, Any]:
+        kwargs: dict[str, str] = {}
+        if mode is not None:
+            kwargs["mode"] = mode
+        if owner is not None:
+            kwargs["owner"] = owner
+        if group is not None:
+            kwargs["group"] = group
+        sandbox.fs.set_file_permissions(path, **kwargs)
+        return {"status": "permissions_set", "path": path, **kwargs}
+
+    def _replace(
+        self,
+        sandbox: Any,
+        paths: list[str],
+        pattern: str,
+        replacement: str,
+    ) -> dict[str, Any]:
+        results = sandbox.fs.replace_in_files(paths, pattern, replacement)
+        return {
+            "pattern": pattern,
+            "replacement": replacement,
+            "results": [
+                {
+                    "file": getattr(r, "file", None),
+                    "success": getattr(r, "success", None),
+                    "error": getattr(r, "error", None),
+                }
+                for r in (results or [])
+            ],
+        }
+
     @staticmethod
     def _file_info_to_dict(info: Any) -> dict[str, Any]:
         fields = (

lib/crewai-tools/tool.specs.json

@@ -7184,7 +7184,7 @@
       }
     },
     {
-      "description": "Perform filesystem operations inside a Daytona sandbox: read a file, write content to a path, append content to an existing file, list a directory, delete a path, make a directory, or fetch file metadata. For files larger than a few KB, create the file with action='write' and empty content, then send the body via multiple 'append' calls of ~4KB each to stay within tool-call payload limits.",
+      "description": "Perform filesystem operations inside a Daytona sandbox: read, write, append, list, delete, mkdir, info, exists, move, find (content grep), search (filename glob), chmod (permissions/owner/group), and replace (bulk find-and-replace across files). For files larger than a few KB, create the file with action='write' and empty content, then send the body via multiple 'append' calls of ~4KB each to stay within tool-call payload limits.",
       "env_vars": [
         {
           "default": null,
@@ -7334,9 +7334,127 @@
         "daytona"
       ],
       "run_params_schema": {
+        "allOf": [
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "enum": [
+                    "read",
+                    "write",
+                    "append",
+                    "list",
+                    "delete",
+                    "mkdir",
+                    "info",
+                    "exists",
+                    "move",
+                    "find",
+                    "search",
+                    "chmod"
+                  ]
+                }
+              }
+            },
+            "then": {
+              "required": [
+                "path"
+              ]
+            }
+          },
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "const": "append"
+                }
+              }
+            },
+            "then": {
+              "required": [
+                "content"
+              ]
+            }
+          },
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "const": "move"
+                }
+              }
+            },
+            "then": {
+              "required": [
+                "destination"
+              ]
+            }
+          },
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "enum": [
+                    "find",
+                    "search"
+                  ]
+                }
+              }
+            },
+            "then": {
+              "required": [
+                "pattern"
+              ]
+            }
+          },
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "const": "replace"
+                }
+              }
+            },
+            "then": {
+              "required": [
+                "paths",
+                "pattern",
+                "replacement"
+              ]
+            }
+          },
+          {
+            "if": {
+              "properties": {
+                "action": {
+                  "const": "chmod"
+                }
+              }
+            },
+            "then": {
+              "anyOf": [
+                {
+                  "required": [
+                    "mode"
+                  ]
+                },
+                {
+                  "required": [
+                    "owner"
+                  ]
+                },
+                {
+                  "required": [
+                    "group"
+                  ]
+                }
+              ]
+            }
+          }
+        ],
         "properties": {
           "action": {
-            "description": "The filesystem action to perform: 'read' (returns file contents), 'write' (create or replace a file with content), 'append' (append content to an existing file \u2014 use this for writing large files in chunks to avoid hitting tool-call size limits), 'list' (lists a directory), 'delete' (removes a file/dir), 'mkdir' (creates a directory), 'info' (returns file metadata).",
+            "description": "The filesystem action to perform: 'read' (returns file contents); 'write' (create or replace a file with content); 'append' (append content to an existing file \u2014 use this for writing large files in chunks to avoid hitting tool-call size limits); 'list' (lists a directory); 'delete' (removes a file/dir); 'mkdir' (creates a directory); 'info' (returns file metadata); 'exists' (returns whether a path exists); 'move' (rename or relocate a file/dir; requires 'destination'); 'find' (grep file CONTENTS recursively; requires 'pattern'); 'search' (find files by NAME pattern; requires 'pattern'); 'chmod' (change permissions/owner/group; pass at least one of 'mode', 'owner', 'group'); 'replace' (find-and-replace text across files; requires 'paths', 'pattern', and 'replacement').",
             "enum": [
               "read",
               "write",
@@ -7344,7 +7462,13 @@
               "list",
               "delete",
               "mkdir",
-              "info"
+              "info",
+              "exists",
+              "move",
+              "find",
+              "search",
+              "chmod",
+              "replace"
             ],
             "title": "Action",
             "type": "string"
@@ -7368,27 +7492,122 @@
             "description": "Content to write or append. If omitted for 'write', an empty file is created. For files larger than a few KB, prefer one 'write' with empty content followed by multiple 'append' calls of ~4KB each to stay within tool-call payload limits.",
             "title": "Content"
           },
+          "destination": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For action='move': absolute destination path.",
+            "title": "Destination"
+          },
+          "group": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For action='chmod': new file group.",
+            "title": "Group"
+          },
           "mode": {
-            "default": "0755",
-            "description": "For action='mkdir': octal permission string (default 0755).",
-            "title": "Mode",
-            "type": "string"
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Octal permission string. For 'mkdir' it sets the new directory permissions (defaults to '0755' if omitted). For 'chmod' it sets the target's mode (e.g. '755' to make a script executable). Ignored for other actions.",
+            "title": "Mode"
+          },
+          "owner": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For action='chmod': new file owner (user name).",
+            "title": "Owner"
           },
           "path": {
-            "description": "Absolute path inside the sandbox.",
-            "title": "Path",
-            "type": "string"
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Absolute path inside the sandbox. Required for all actions except 'replace' (which uses 'paths' instead).",
+            "title": "Path"
+          },
+          "paths": {
+            "anyOf": [
+              {
+                "items": {
+                  "type": "string"
+                },
+                "type": "array"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For action='replace': list of absolute file paths in which to replace 'pattern' with 'replacement'.",
+            "title": "Paths"
+          },
+          "pattern": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For 'find': substring matched against file CONTENTS. For 'search': glob-style pattern matched against file NAMES (e.g. '*.py'). For 'replace': text to replace inside files.",
+            "title": "Pattern"
           },
           "recursive": {
             "default": false,
             "description": "For action='delete': remove directories recursively.",
             "title": "Recursive",
             "type": "boolean"
+          },
+          "replacement": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "For action='replace': replacement text for 'pattern'.",
+            "title": "Replacement"
           }
         },
         "required": [
-          "action",
-          "path"
+          "action"
         ],
         "title": "DaytonaFileToolSchema",
         "type": "object"

lib/crewai/pyproject.toml

@@ -8,8 +8,8 @@ authors = [
 ]
 requires-python = ">=3.10, <3.14"
 dependencies = [
-    "crewai-core==1.14.5a3",
-    "crewai-cli==1.14.5a3",
+    "crewai-core==1.14.5",
+    "crewai-cli==1.14.5",
     # Core Dependencies
     "pydantic>=2.11.9,<2.13",
     "openai>=2.30.0,<3",
@@ -28,8 +28,6 @@ dependencies = [
     # Authentication and Security
     "python-dotenv>=1.2.2,<2",
     "pyjwt>=2.9.0,<3",
-    # TUI
-    "textual>=7.5.0",
     # Configuration and Utils
     "click~=8.1.7",
     "appdirs~=1.4.4",
@@ -56,7 +54,7 @@ Repository = "https://github.com/crewAIInc/crewAI"
 
 [project.optional-dependencies]
 tools = [
-    "crewai-tools==1.14.5a3",
+    "crewai-tools==1.14.5",
 ]
 embeddings = [
     "tiktoken>=0.8.0,<0.13"
@@ -67,7 +65,7 @@ pandas = [
 openpyxl = [
     "openpyxl~=3.1.5",
 ]
-mem0 = ["mem0ai~=0.1.94"]
+mem0 = ["mem0ai>=2.0.0,<3"]
 docling = [
     "docling~=2.84.0",
 ]

lib/crewai/src/crewai/__init__.py

@@ -48,7 +48,7 @@ def _suppress_pydantic_deprecation_warnings() -> None:
 
 _suppress_pydantic_deprecation_warnings()
 
-__version__ = "1.14.5a3"
+__version__ = "1.14.5"
 
 _LAZY_IMPORTS: dict[str, tuple[str, str]] = {
     "Memory": ("crewai.memory.unified_memory", "Memory"),

lib/crewai/src/crewai/agent/core.py

@@ -7,6 +7,7 @@ from collections.abc import Callable, Coroutine, Sequence
 import concurrent.futures
 import contextvars
 from datetime import datetime
+import inspect
 import json
 import os
 from pathlib import Path
@@ -35,13 +36,11 @@ from typing_extensions import Self, TypeIs
 from crewai.agent.planning_config import PlanningConfig
 from crewai.agent.utils import (
     ahandle_knowledge_retrieval,
-    append_skill_context,
     apply_training_data,
     build_task_prompt_with_schema,
     format_task_with_context,
     get_knowledge_config,
     handle_knowledge_retrieval,
-    handle_reasoning,
     prepare_tools,
     process_tool_results,
     save_last_messages,
@@ -150,7 +149,17 @@ def _validate_executor_class(value: Any) -> Any:
         cls = _EXECUTOR_CLASS_MAP.get(value)
         if cls is None:
             raise ValueError(f"Unknown executor class: {value}")
-        return cls
+        value = cls
+    import warnings
+
+    if value is CrewAgentExecutor:
+        warnings.warn(
+            "CrewAgentExecutor is deprecated and will be removed in a future release. "
+            "Agents inside Crews now use AgentExecutor by default. "
+            "Switch to crewai.experimental.AgentExecutor.",
+            DeprecationWarning,
+            stacklevel=3,
+        )
     return value
 
 
@@ -211,7 +220,11 @@ class Agent(BaseAgent):
         str | BaseLLM | None,
         BeforeValidator(_validate_llm_ref),
         PlainSerializer(_serialize_llm_ref, return_type=dict | None, when_used="json"),
-    ] = Field(description="Language model that will run the agent.", default=None)
+    ] = Field(
+        description="Language model that will run the agent.",
+        default=None,
+        deprecated="function_calling_llm is deprecated and will be removed in a future release.",
+    )
     system_template: str | None = Field(
         default=None, description="System format for the agent."
     )
@@ -325,8 +338,8 @@ class Agent(BaseAgent):
         BeforeValidator(_validate_executor_class),
         PlainSerializer(_serialize_executor_class, return_type=str, when_used="json"),
     ] = Field(
-        default=CrewAgentExecutor,
-        description="Class to use for the agent executor. Defaults to CrewAgentExecutor, can optionally use AgentExecutor.",
+        default=AgentExecutor,
+        description="Class to use for the agent executor. Defaults to AgentExecutor, can optionally use CrewAgentExecutor.",
     )
 
     @model_validator(mode="before")
@@ -421,7 +434,7 @@ class Agent(BaseAgent):
         from crewai.crew import Crew
 
         if resolved_crew_skills is None:
-            crew_skills: list[Path | SkillModel] | None = (
+            crew_skills: list[Path | SkillModel | str] | None = (
                 self.crew.skills
                 if isinstance(self.crew, Crew) and isinstance(self.crew.skills, list)
                 else None
@@ -433,7 +446,7 @@ class Agent(BaseAgent):
             return
 
         needs_work = self.skills and any(
-            isinstance(s, Path)
+            isinstance(s, (Path, str))
             or (isinstance(s, SkillModel) and s.disclosure_level < INSTRUCTIONS)
             for s in self.skills
         )
@@ -441,14 +454,28 @@ class Agent(BaseAgent):
             return
 
         seen: set[str] = set()
-        resolved: list[Path | SkillModel] = []
-        items: list[Path | SkillModel] = list(self.skills) if self.skills else []
+        resolved: list[Path | SkillModel | str] = []
+        items: list[Path | SkillModel | str] = list(self.skills) if self.skills else []
 
         if crew_skills:
             items.extend(crew_skills)
 
         for item in items:
-            if isinstance(item, Path):
+            if isinstance(item, str):
+                from crewai.skills.registry import (
+                    is_registry_ref,
+                    parse_registry_ref,
+                    resolve_registry_ref,
+                )
+
+                if is_registry_ref(item):
+                    skill = resolve_registry_ref(item, source=self)
+                    org, _ = parse_registry_ref(item)
+                    dedup_key = f"{org}/{skill.name}"
+                    if dedup_key not in seen:
+                        seen.add(dedup_key)
+                        resolved.append(skill)
+            elif isinstance(item, Path):
                 discovered = discover_skills(item, source=self)
                 for skill in discovered:
                     if skill.name not in seen:
@@ -512,8 +539,6 @@ class Agent(BaseAgent):
             The task prompt after memory retrieval, ready for knowledge lookup.
         """
         get_env_context()
-        if self.executor_class is not AgentExecutor:
-            handle_reasoning(self, task)
 
         self._inject_date_to_task(task)
 
@@ -541,7 +566,6 @@ class Agent(BaseAgent):
         Returns:
             The fully prepared task prompt.
         """
-        task_prompt = append_skill_context(self, task_prompt)
         prepare_tools(self, tools, task)
 
         return apply_training_data(self, task_prompt)
@@ -843,18 +867,22 @@ class Agent(BaseAgent):
         if not self.agent_executor:
             raise RuntimeError("Agent executor is not initialized.")
 
-        result = cast(
-            dict[str, Any],
-            self.agent_executor.invoke(
-                {
-                    "input": task_prompt,
-                    "tool_names": self.agent_executor.tools_names,
-                    "tools": self.agent_executor.tools_description,
-                    "ask_for_human_input": task.human_input,
-                }
-            ),
+        invoke_result = self.agent_executor.invoke(
+            {
+                "input": task_prompt,
+                "tool_names": self.agent_executor.tools_names,
+                "tools": self.agent_executor.tools_description,
+                "ask_for_human_input": task.human_input,
+            }
         )
-        return result["output"]
+        if inspect.isawaitable(invoke_result):
+            invoke_result.close()
+            raise RuntimeError(
+                "Agent execution was invoked synchronously from within a running "
+                "event loop. Use `agent.kickoff_async()` / `crew.kickoff_async()` "
+                "(or `await agent.aexecute_task(...)`) when calling from async code."
+            )
+        return invoke_result["output"]
 
     async def aexecute_task(
         self,
@@ -1474,8 +1502,6 @@ class Agent(BaseAgent):
                     ),
                 )
 
-        formatted_messages = append_skill_context(self, formatted_messages)
-
         inputs: dict[str, Any] = {
             "input": formatted_messages,
             "tool_names": get_tool_names(parsed_tools),

lib/crewai/src/crewai/agent/utils.py

@@ -213,30 +213,6 @@ def _combine_knowledge_context(agent: Agent) -> str:
     return agent_ctx + separator + crew_ctx
 
 
-def append_skill_context(agent: Agent, task_prompt: str) -> str:
-    """Append activated skill context sections to the task prompt.
-
-    Args:
-        agent: The agent with optional skills.
-        task_prompt: The current task prompt.
-
-    Returns:
-        The task prompt with skill context appended.
-    """
-    if not agent.skills:
-        return task_prompt
-
-    from crewai.skills.loader import format_skill_context
-    from crewai.skills.models import Skill
-
-    skill_sections = [
-        format_skill_context(s) for s in agent.skills if isinstance(s, Skill)
-    ]
-    if skill_sections:
-        task_prompt += "\n\n" + "\n\n".join(skill_sections)
-    return task_prompt
-
-
 def apply_training_data(agent: Agent, task_prompt: str) -> str:
     """Apply training data to the task prompt.
 

lib/crewai/src/crewai/agents/__init__.py

@@ -1,13 +1,28 @@
+from typing import TYPE_CHECKING, Any
+
 from crewai.agents.cache.cache_handler import CacheHandler
 from crewai.agents.parser import AgentAction, AgentFinish, OutputParserError, parse
 from crewai.agents.tools_handler import ToolsHandler
 
 
+if TYPE_CHECKING:
+    from crewai.agents.crew_agent_executor import CrewAgentExecutor
+
+
 __all__ = [
     "AgentAction",
     "AgentFinish",
     "CacheHandler",
+    "CrewAgentExecutor",
     "OutputParserError",
     "ToolsHandler",
     "parse",
 ]
+
+
+def __getattr__(name: str) -> Any:
+    if name == "CrewAgentExecutor":
+        from crewai.agents.crew_agent_executor import CrewAgentExecutor
+
+        return CrewAgentExecutor
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

lib/crewai/src/crewai/agents/agent_adapters/langgraph/langgraph_adapter.py

@@ -51,7 +51,10 @@ class LangGraphAgentAdapter(BaseAgentAdapter):
     _graph: Any = PrivateAttr(default=None)
     _memory: Any = PrivateAttr(default=None)
     _max_iterations: int = PrivateAttr(default=10)
-    function_calling_llm: Any = Field(default=None)
+    function_calling_llm: Any = Field(
+        default=None,
+        deprecated="function_calling_llm is deprecated and will be removed in a future release.",
+    )
     step_callback: SerializableCallable | None = Field(default=None)
 
     model: str = Field(default="gpt-4o")

lib/crewai/src/crewai/agents/agent_adapters/openai_agents/openai_adapter.py

@@ -60,7 +60,10 @@ class OpenAIAgentAdapter(BaseAgentAdapter):
     _openai_agent: OpenAIAgentProtocol = PrivateAttr()
     _logger: Logger = PrivateAttr(default_factory=Logger)
     _active_thread: str | None = PrivateAttr(default=None)
-    function_calling_llm: Any = Field(default=None)
+    function_calling_llm: Any = Field(
+        default=None,
+        deprecated="function_calling_llm is deprecated and will be removed in a future release.",
+    )
     step_callback: Any = Field(default=None)
     _tool_adapter: OpenAIAgentToolAdapter = PrivateAttr()
     _converter_adapter: OpenAIConverterAdapter = PrivateAttr()

lib/crewai/src/crewai/agents/agent_builder/base_agent.py

@@ -31,13 +31,13 @@ from crewai.agents.tools_handler import ToolsHandler
 from crewai.events.base_events import set_emission_counter
 from crewai.events.event_bus import crewai_event_bus
 from crewai.events.event_context import restore_event_scope, set_last_event_id
-from crewai.knowledge.knowledge import Knowledge
+from crewai.knowledge.knowledge import Knowledge, _resolve_knowledge_sources
 from crewai.knowledge.knowledge_config import KnowledgeConfig
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.knowledge.storage.base_knowledge_storage import BaseKnowledgeStorage
 from crewai.llms.base_llm import BaseLLM
 from crewai.mcp.config import MCPServerConfig
-from crewai.memory.memory_scope import MemoryScope, MemorySlice
+from crewai.memory.memory_scope import MemoryScope, MemorySlice, _ensure_memory_kind
 from crewai.memory.unified_memory import Memory
 from crewai.rag.embeddings.types import EmbedderConfig
 from crewai.security.security_config import SecurityConfig
@@ -127,6 +127,13 @@ def _validate_executor_ref(value: Any) -> Any:
     return value
 
 
+def _serialize_executor_ref(value: Any) -> dict[str, Any] | None:
+    if value is None:
+        return None
+    result: dict[str, Any] = value.model_dump(mode="json")
+    return result
+
+
 def _serialize_llm_ref(value: Any) -> dict[str, Any] | None:
     if value is None:
         return None
@@ -251,14 +258,13 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
     max_iter: int = Field(
         default=25, description="Maximum iterations for an agent to execute a task"
     )
-    agent_executor: SerializeAsAny[BaseAgentExecutor] | None = Field(
-        default=None, description="An instance of the CrewAgentExecutor class."
-    )
-
-    @field_validator("agent_executor", mode="before")
-    @classmethod
-    def _validate_agent_executor(cls, v: Any) -> Any:
-        return _validate_executor_ref(v)
+    agent_executor: Annotated[
+        SerializeAsAny[BaseAgentExecutor] | None,
+        BeforeValidator(_validate_executor_ref),
+        PlainSerializer(
+            _serialize_executor_ref, return_type=dict | None, when_used="json"
+        ),
+    ] = Field(default=None, description="An instance of the CrewAgentExecutor class.")
 
     llm: Annotated[
         str | BaseLLM | None,
@@ -288,7 +294,10 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
     knowledge: Knowledge | None = Field(
         default=None, description="Knowledge for the agent."
     )
-    knowledge_sources: list[BaseKnowledgeSource] | None = Field(
+    knowledge_sources: Annotated[
+        list[BaseKnowledgeSource] | None,
+        BeforeValidator(_resolve_knowledge_sources),
+    ] = Field(
         default=None,
         description="Knowledge sources for the agent.",
     )
@@ -326,7 +335,14 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
         default=None,
         description="List of MCP server references. Supports 'https://server.com/path' for external servers and bare slugs like 'notion' for connected MCP integrations. Use '#tool_name' suffix for specific tools.",
     )
-    memory: bool | Memory | MemoryScope | MemorySlice | None = Field(
+    memory: Annotated[
+        bool
+        | Annotated[
+            Memory | MemoryScope | MemorySlice, Field(discriminator="memory_kind")
+        ]
+        | None,
+        BeforeValidator(_ensure_memory_kind),
+    ] = Field(
         default=None,
         description=(
             "Enable agent memory. Pass True for default Memory(), "
@@ -334,9 +350,9 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
             "If not set, falls back to crew memory."
         ),
     )
-    skills: list[Path | Skill] | None = Field(
+    skills: list[Path | Skill | str] | None = Field(
         default=None,
-        description="Agent Skills. Accepts paths for discovery or pre-loaded Skill objects.",
+        description="Agent Skills. Accepts paths for discovery, pre-loaded Skill objects, or '@org/name' registry refs.",
         min_length=1,
     )
     execution_context: ExecutionContext | None = Field(default=None)
@@ -397,8 +413,21 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
             self.agent_executor._resuming = True
         if self.checkpoint_kickoff_event_id is not None:
             self._kickoff_event_id = self.checkpoint_kickoff_event_id
+        self._rebind_memory_view()
         self._restore_event_scope(state)
 
+    def _rebind_memory_view(self) -> None:
+        """Reattach a fresh ``Memory`` to a restored ``MemoryScope``/``MemorySlice``.
+
+        Checkpoint JSON omits the live ``Memory`` dependency, so scoped
+        memory views raise ``RuntimeError`` on first use after restore.
+        """
+        if (
+            isinstance(self.memory, MemoryScope | MemorySlice)
+            and self.memory._memory is None
+        ):
+            self.memory.bind(Memory())
+
     def _restore_event_scope(self, state: RuntimeState) -> None:
         """Rebuild the event scope stack from the checkpoint's event record.
 
@@ -429,6 +458,20 @@ class BaseAgent(BaseModel, ABC, metaclass=AgentMeta):
     def process_model_config(cls, values: Any) -> dict[str, Any]:
         return process_config(values, cls)
 
+    @field_validator("skills", mode="before")
+    @classmethod
+    def coerce_skill_strings(cls, skills: Any) -> Any:
+        """Coerce plain path strings to Path objects; keep @-prefixed refs as str."""
+        if not isinstance(skills, list):
+            return skills
+        result = []
+        for item in skills:
+            if isinstance(item, str) and not item.startswith("@"):
+                result.append(Path(item))
+            else:
+                result.append(item)
+        return result
+
     @field_validator("tools")
     @classmethod
     def validate_tools(cls, tools: list[Any]) -> list[BaseTool]:

lib/crewai/src/crewai/agents/crew_agent_executor.py

@@ -14,6 +14,7 @@ import contextvars
 import inspect
 import logging
 from typing import TYPE_CHECKING, Annotated, Any, Literal, cast
+import warnings
 
 from crewai_core.printer import PRINTER
 from pydantic import (
@@ -138,6 +139,13 @@ class CrewAgentExecutor(BaseAgentExecutor):
 
     def __init__(self, **kwargs: Any) -> None:
         super().__init__(**kwargs)
+        warnings.warn(
+            "CrewAgentExecutor is deprecated and will be removed in a future release.\n"
+            "Agents inside Crews now use AgentExecutor (crewai.experimental.AgentExecutor) by default.\n"
+            "To suppress this warning, migrate to AgentExecutor.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
         if not self.before_llm_call_hooks:
             self.before_llm_call_hooks.extend(get_before_llm_call_hooks())
         if not self.after_llm_call_hooks:
@@ -166,6 +174,8 @@ class CrewAgentExecutor(BaseAgentExecutor):
         if provider.setup_messages(cast(ExecutorContext, cast(object, self))):
             return
 
+        from crewai.llms.cache import mark_cache_breakpoint
+
         if self.prompt is not None and "system" in self.prompt:
             system_prompt = self._format_prompt(
                 cast(str, self.prompt.get("system", "")), inputs
@@ -173,11 +183,22 @@ class CrewAgentExecutor(BaseAgentExecutor):
             user_prompt = self._format_prompt(
                 cast(str, self.prompt.get("user", "")), inputs
             )
-            self.messages.append(format_message_for_llm(system_prompt, role="system"))
-            self.messages.append(format_message_for_llm(user_prompt))
+            # Cache breakpoints: end-of-system caches the per-agent stable
+            # prefix; end-of-user caches the per-task stable prefix across
+            # ReAct-loop iterations.
+            self.messages.append(
+                mark_cache_breakpoint(
+                    format_message_for_llm(system_prompt, role="system")
+                )
+            )
+            self.messages.append(
+                mark_cache_breakpoint(format_message_for_llm(user_prompt))
+            )
         elif self.prompt is not None:
             user_prompt = self._format_prompt(self.prompt.get("prompt", ""), inputs)
-            self.messages.append(format_message_for_llm(user_prompt))
+            self.messages.append(
+                mark_cache_breakpoint(format_message_for_llm(user_prompt))
+            )
 
         provider.post_setup_messages(cast(ExecutorContext, cast(object, self)))
 

lib/crewai/src/crewai/crew.py

@@ -93,11 +93,11 @@ from crewai.events.types.crew_events import (
     CrewTrainStartedEvent,
 )
 from crewai.flow.flow_trackable import FlowTrackable
-from crewai.knowledge.knowledge import Knowledge
+from crewai.knowledge.knowledge import Knowledge, _resolve_knowledge_sources
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.llm import LLM
 from crewai.llms.base_llm import BaseLLM
-from crewai.memory.memory_scope import MemoryScope, MemorySlice
+from crewai.memory.memory_scope import MemoryScope, MemorySlice, _ensure_memory_kind
 from crewai.memory.unified_memory import Memory
 from crewai.process import Process
 from crewai.rag.embeddings.types import EmbedderConfig
@@ -223,7 +223,14 @@ class Crew(FlowTrackable, BaseModel):
     ] = Field(default_factory=list)
     process: Process = Field(default=Process.sequential)
     verbose: bool = Field(default=False)
-    memory: bool | Memory | MemoryScope | MemorySlice | None = Field(
+    memory: Annotated[
+        bool
+        | Annotated[
+            Memory | MemoryScope | MemorySlice, Field(discriminator="memory_kind")
+        ]
+        | None,
+        BeforeValidator(_ensure_memory_kind),
+    ] = Field(
         default=False,
         description=(
             "Enable crew memory. Pass True for default Memory(), "
@@ -251,7 +258,11 @@ class Crew(FlowTrackable, BaseModel):
         str | LLM | None,
         BeforeValidator(_validate_llm_ref),
         PlainSerializer(_serialize_llm_ref, return_type=dict | None, when_used="json"),
-    ] = Field(description="Language model that will run the agent.", default=None)
+    ] = Field(
+        description="Language model that will run the agent.",
+        default=None,
+        deprecated="function_calling_llm is deprecated and will be removed in a future release.",
+    )
     config: Json[dict[str, Any]] | dict[str, Any] | None = Field(default=None)
     id: UUID4 = Field(default_factory=uuid.uuid4, frozen=True)
     share_crew: bool | None = Field(default=False)
@@ -318,7 +329,10 @@ class Crew(FlowTrackable, BaseModel):
         default_factory=list,
         description="list of execution logs for tasks",
     )
-    knowledge_sources: list[BaseKnowledgeSource] | None = Field(
+    knowledge_sources: Annotated[
+        list[BaseKnowledgeSource] | None,
+        BeforeValidator(_resolve_knowledge_sources),
+    ] = Field(
         default=None,
         description=(
             "Knowledge sources for the crew. Add knowledge sources to the "
@@ -337,9 +351,9 @@ class Crew(FlowTrackable, BaseModel):
         default=None,
         description="Knowledge for the crew.",
     )
-    skills: list[Path | Skill] | None = Field(
+    skills: list[Path | Skill | str] | None = Field(
         default=None,
-        description="Skill search paths or pre-loaded Skill objects applied to all agents in the crew.",
+        description="Skill search paths, pre-loaded Skill objects, or '@org/name' registry refs applied to all agents in the crew.",
     )
 
     security_config: SecurityConfig = Field(
@@ -473,8 +487,42 @@ class Crew(FlowTrackable, BaseModel):
         if self.checkpoint_train is not None:
             self._train = self.checkpoint_train
 
+        self._rebind_memory_views()
         self._restore_event_scope()
 
+    def _rebind_memory_views(self) -> None:
+        """Reattach a live ``Memory`` to restored ``MemoryScope``/``MemorySlice`` views.
+
+        Checkpoint JSON omits the live ``Memory`` dependency on scope/slice
+        views, so after restore they raise ``RuntimeError`` on first use.
+        Prefer the crew's restored ``Memory`` (from ``create_crew_memory``
+        or a ``Crew.memory=Memory(...)`` instance) so all views share one
+        backing store; fall back to a fresh ``Memory()`` only if nothing is
+        available.
+        """
+        from crewai.memory.memory_scope import MemoryScope, MemorySlice
+        from crewai.memory.unified_memory import Memory
+
+        backing: Memory | None = None
+        if isinstance(self._memory, Memory):
+            backing = self._memory
+        elif isinstance(self.memory, Memory):
+            backing = self.memory
+
+        def _ensure(view: Any) -> None:
+            nonlocal backing
+            if not isinstance(view, MemoryScope | MemorySlice):
+                return
+            if view._memory is not None:
+                return
+            if backing is None:
+                backing = Memory()
+            view.bind(backing)
+
+        _ensure(self.memory)
+        for agent in self.agents:
+            _ensure(agent.memory)
+
     def _restore_event_scope(self) -> None:
         """Rebuild the event scope stack from the checkpoint's event record."""
         from crewai.events.base_events import set_emission_counter
@@ -522,6 +570,20 @@ class Crew(FlowTrackable, BaseModel):
         if max_seq > 0:
             set_emission_counter(max_seq)
 
+    @field_validator("skills", mode="before")
+    @classmethod
+    def coerce_skill_strings(cls, skills: Any) -> Any:
+        """Coerce plain path strings to Path objects; keep @-prefixed refs as str."""
+        if not isinstance(skills, list):
+            return skills
+        result = []
+        for item in skills:
+            if isinstance(item, str) and not item.startswith("@"):
+                result.append(Path(item))
+            else:
+                result.append(item)
+        return result
+
     @field_validator("id", mode="before")
     @classmethod
     def _deny_user_set_id(cls, v: UUID4 | None, info: Any) -> UUID4 | None:

lib/crewai/src/crewai/events/listeners/tracing/trace_batch_manager.py

@@ -6,6 +6,14 @@ import time
 from typing import Any
 import uuid
 
+from crewai_core.plus_api import (
+    TraceBatchInitPayload,
+    TraceBatchMetadata,
+    TraceEventsPayload,
+    TraceExecutionContext,
+    TraceExecutionMetadata,
+    TraceFinalizePayload,
+)
 from crewai_core.settings import Settings
 from rich.console import Console
 from rich.panel import Panel
@@ -123,25 +131,27 @@ class TraceBatchManager:
             return None
 
         try:
-            payload = {
+            execution_context: TraceExecutionContext = {
+                "crew_fingerprint": execution_metadata.get("crew_fingerprint"),
+                "crew_name": execution_metadata.get("crew_name", None),
+                "flow_name": execution_metadata.get("flow_name", None),
+                "crewai_version": self.current_batch.version,
+                "privacy_level": user_context.get("privacy_level", "standard"),
+            }
+            execution_metadata_payload: TraceExecutionMetadata = {
+                "expected_duration_estimate": execution_metadata.get(
+                    "expected_duration_estimate", 300
+                ),
+                "agent_count": execution_metadata.get("agent_count", 0),
+                "task_count": execution_metadata.get("task_count", 0),
+                "flow_method_count": execution_metadata.get("flow_method_count", 0),
+                "execution_started_at": datetime.now(timezone.utc).isoformat(),
+            }
+            payload: TraceBatchInitPayload = {
                 "trace_id": self.current_batch.batch_id,
                 "execution_type": execution_metadata.get("execution_type", "crew"),
-                "execution_context": {
-                    "crew_fingerprint": execution_metadata.get("crew_fingerprint"),
-                    "crew_name": execution_metadata.get("crew_name", None),
-                    "flow_name": execution_metadata.get("flow_name", None),
-                    "crewai_version": self.current_batch.version,
-                    "privacy_level": user_context.get("privacy_level", "standard"),
-                },
-                "execution_metadata": {
-                    "expected_duration_estimate": execution_metadata.get(
-                        "expected_duration_estimate", 300
-                    ),
-                    "agent_count": execution_metadata.get("agent_count", 0),
-                    "task_count": execution_metadata.get("task_count", 0),
-                    "flow_method_count": execution_metadata.get("flow_method_count", 0),
-                    "execution_started_at": datetime.now(timezone.utc).isoformat(),
-                },
+                "execution_context": execution_context,
+                "execution_metadata": execution_metadata_payload,
             }
             if use_ephemeral:
                 payload["ephemeral_trace_id"] = self.current_batch.batch_id
@@ -264,13 +274,14 @@ class TraceBatchManager:
         if not self.plus_api or not self.trace_batch_id or not self.event_buffer:
             return 500
         try:
-            payload = {
+            batch_metadata: TraceBatchMetadata = {
+                "events_count": len(self.event_buffer),
+                "batch_sequence": 1,
+                "is_final_batch": False,
+            }
+            payload: TraceEventsPayload = {
                 "events": [event.to_dict() for event in self.event_buffer],
-                "batch_metadata": {
-                    "events_count": len(self.event_buffer),
-                    "batch_sequence": 1,
-                    "is_final_batch": False,
-                },
+                "batch_metadata": batch_metadata,
             }
 
             response = (
@@ -364,7 +375,7 @@ class TraceBatchManager:
             return
 
         try:
-            payload = {
+            payload: TraceFinalizePayload = {
                 "status": "completed",
                 "duration_ms": self.calculate_duration("execution"),
                 "final_event_count": events_count,

lib/crewai/src/crewai/events/types/skill_events.py

@@ -60,3 +60,20 @@ class SkillLoadFailedEvent(SkillEvent):
 
     type: Literal["skill_load_failed"] = "skill_load_failed"
     error: str
+
+
+class SkillDownloadStartedEvent(SkillEvent):
+    """Event emitted when a registry skill download begins."""
+
+    type: Literal["skill_download_started"] = "skill_download_started"
+    registry_ref: str
+    version: str | None = None
+
+
+class SkillDownloadCompletedEvent(SkillEvent):
+    """Event emitted when a registry skill download completes."""
+
+    type: Literal["skill_download_completed"] = "skill_download_completed"
+    registry_ref: str
+    version: str | None = None
+    cache_path: Path | None = None

lib/crewai/src/crewai/experimental/agent_executor.py

@@ -1191,6 +1191,13 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
     @router("force_final_answer")
     def ensure_force_final_answer(self) -> Literal["agent_finished"]:
         """Force agent to provide final answer when max iterations exceeded."""
+        # The flow framework can route here more than once per execution when the
+        # "initialized" label is emitted by both initialize_reasoning and
+        # increment_and_continue in the same listener pass. Skip the extra LLM
+        # round-trip once we've already produced a forced final answer.
+        if self.state.is_finished:
+            return "agent_finished"
+
         formatted_answer = handle_max_iterations_exceeded(
             formatted_answer=None,
             printer=PRINTER,
@@ -2579,16 +2586,26 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
             self._kickoff_input = inputs.get("input", "")
 
             if "system" in self.prompt:
+                from crewai.llms.cache import mark_cache_breakpoint
+
                 prompt = cast("SystemPromptResult", self.prompt)
                 system_prompt = self._format_prompt(prompt["system"], inputs)
                 user_prompt = self._format_prompt(prompt["user"], inputs)
                 self.state.messages.append(
-                    format_message_for_llm(system_prompt, role="system")
+                    mark_cache_breakpoint(
+                        format_message_for_llm(system_prompt, role="system")
+                    )
+                )
+                self.state.messages.append(
+                    mark_cache_breakpoint(format_message_for_llm(user_prompt))
                 )
-                self.state.messages.append(format_message_for_llm(user_prompt))
             else:
+                from crewai.llms.cache import mark_cache_breakpoint
+
                 user_prompt = self._format_prompt(self.prompt["prompt"], inputs)
-                self.state.messages.append(format_message_for_llm(user_prompt))
+                self.state.messages.append(
+                    mark_cache_breakpoint(format_message_for_llm(user_prompt))
+                )
 
             self._inject_files_from_inputs(inputs)
 
@@ -2670,16 +2687,26 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
             self._kickoff_input = inputs.get("input", "")
 
             if "system" in self.prompt:
+                from crewai.llms.cache import mark_cache_breakpoint
+
                 prompt = cast("SystemPromptResult", self.prompt)
                 system_prompt = self._format_prompt(prompt["system"], inputs)
                 user_prompt = self._format_prompt(prompt["user"], inputs)
                 self.state.messages.append(
-                    format_message_for_llm(system_prompt, role="system")
+                    mark_cache_breakpoint(
+                        format_message_for_llm(system_prompt, role="system")
+                    )
+                )
+                self.state.messages.append(
+                    mark_cache_breakpoint(format_message_for_llm(user_prompt))
                 )
-                self.state.messages.append(format_message_for_llm(user_prompt))
             else:
+                from crewai.llms.cache import mark_cache_breakpoint
+
                 user_prompt = self._format_prompt(self.prompt["prompt"], inputs)
-                self.state.messages.append(format_message_for_llm(user_prompt))
+                self.state.messages.append(
+                    mark_cache_breakpoint(format_message_for_llm(user_prompt))
+                )
 
             self._inject_files_from_inputs(inputs)
 

lib/crewai/src/crewai/flow/flow.py

@@ -113,7 +113,7 @@ from crewai.flow.utils import (
     is_flow_method_name,
     is_simple_flow_condition,
 )
-from crewai.memory.memory_scope import MemoryScope, MemorySlice
+from crewai.memory.memory_scope import MemoryScope, MemorySlice, _ensure_memory_kind
 from crewai.memory.unified_memory import Memory
 from crewai.state.checkpoint_config import (
     CheckpointConfig,
@@ -159,6 +159,39 @@ def _resolve_persistence(value: Any) -> Any:
     return value
 
 
+def _serialize_persistence(value: Any) -> dict[str, Any] | None:
+    if value is None:
+        return None
+    if isinstance(value, FlowPersistence):
+        return value.model_dump(mode="json")
+    raise TypeError(
+        f"Cannot serialize Flow.persistence of type {type(value).__name__}: "
+        "expected FlowPersistence or None."
+    )
+
+
+def _validate_input_provider(value: Any) -> Any:
+    if value is None or isinstance(value, InputProvider):
+        return value
+    from crewai.types.callback import _dotted_path_to_instance
+
+    resolved = _dotted_path_to_instance(value)
+    if resolved is None or isinstance(resolved, InputProvider):
+        return resolved
+    raise ValueError(
+        f"Resolved input_provider {resolved!r} does not implement the "
+        "InputProvider protocol (missing request_input)."
+    )
+
+
+def _serialize_input_provider(value: Any) -> str | None:
+    if value is None:
+        return None
+    from crewai.types.callback import _instance_to_dotted_path
+
+    return _instance_to_dotted_path(value)
+
+
 _INITIAL_STATE_CLASS_MARKER = "__crewai_pydantic_class_schema__"
 
 
@@ -949,15 +982,30 @@ class Flow(BaseModel, Generic[T], metaclass=FlowMeta):
     name: str | None = Field(default=None)
     tracing: bool | None = Field(default=None)
     stream: bool = Field(default=False)
-    memory: Memory | MemoryScope | MemorySlice | None = Field(default=None)
-    input_provider: InputProvider | None = Field(default=None)
+    memory: Annotated[
+        Annotated[
+            Memory | MemoryScope | MemorySlice, Field(discriminator="memory_kind")
+        ]
+        | None,
+        BeforeValidator(_ensure_memory_kind),
+    ] = Field(default=None)
+    input_provider: Annotated[
+        InputProvider | None,
+        BeforeValidator(_validate_input_provider),
+        PlainSerializer(
+            _serialize_input_provider, return_type=str | None, when_used="json"
+        ),
+    ] = Field(default=None)
     suppress_flow_events: bool = Field(default=False)
     human_feedback_history: list[HumanFeedbackResult] = Field(default_factory=list)
     last_human_feedback: HumanFeedbackResult | None = Field(default=None)
 
     persistence: Annotated[
-        SerializeAsAny[FlowPersistence] | Any,
+        SerializeAsAny[FlowPersistence] | None,
         BeforeValidator(lambda v, _: _resolve_persistence(v)),
+        PlainSerializer(
+            _serialize_persistence, return_type=dict | None, when_used="json"
+        ),
     ] = Field(default=None)
     max_method_calls: int = Field(default=100)
 
@@ -1050,6 +1098,11 @@ class Flow(BaseModel, Generic[T], metaclass=FlowMeta):
             }
         if self.checkpoint_state is not None:
             self._restore_state(self.checkpoint_state)
+        if (
+            isinstance(self.memory, MemoryScope | MemorySlice)
+            and self.memory._memory is None
+        ):
+            self.memory.bind(Memory())
         restore_event_scope(())
         reset_last_event_id()
 

lib/crewai/src/crewai/flow/human_feedback.py

@@ -60,6 +60,7 @@ from collections.abc import Callable, Sequence
 from dataclasses import dataclass, field
 from datetime import datetime
 from functools import wraps
+import logging
 from typing import TYPE_CHECKING, Any, TypeVar
 
 from pydantic import BaseModel, Field
@@ -73,6 +74,8 @@ if TYPE_CHECKING:
     from crewai.llms.base_llm import BaseLLM
 
 
+logger = logging.getLogger(__name__)
+
 F = TypeVar("F", bound=Callable[..., Any])
 
 
@@ -188,6 +191,7 @@ class HumanFeedbackConfig:
     provider: HumanFeedbackProvider | None = None
     learn: bool = False
     learn_source: str = "hitl"
+    learn_strict: bool = False
 
 
 class HumanFeedbackMethod(FlowMethod[Any, Any]):
@@ -237,6 +241,7 @@ def human_feedback(
     provider: HumanFeedbackProvider | None = None,
     learn: bool = False,
     learn_source: str = "hitl",
+    learn_strict: bool = False,
 ) -> Callable[[F], F]:
     """Decorator for Flow methods that require human feedback.
 
@@ -275,6 +280,14 @@ def human_feedback(
             external systems like Slack, Teams, or webhooks. When the
             provider raises HumanFeedbackPending, the flow pauses and
             can be resumed later with Flow.resume().
+        learn: Enable HITL learning. Recall past lessons to pre-review
+            output before the human sees it, and distill new lessons
+            from feedback after.
+        learn_source: Memory source tag for stored/recalled lessons.
+        learn_strict: When True, re-raise exceptions from the pre-review
+            and distillation steps instead of falling back to raw output.
+            Default False preserves graceful degradation; failures are
+            always logged via ``logger.warning`` regardless of this flag.
 
     Returns:
         A decorator function that wraps the method with human feedback
@@ -404,7 +417,19 @@ def human_feedback(
                 reviewed = llm_inst.call(messages)
                 return reviewed if isinstance(reviewed, str) else str(reviewed)
             except Exception:
-                return method_output  # fallback to raw output on any failure
+                if learn_strict:
+                    logger.warning(
+                        "HITL pre-review failed for %s; re-raising (learn_strict=True)",
+                        func.__name__,
+                        exc_info=True,
+                    )
+                    raise
+                logger.warning(
+                    "HITL pre-review failed for %s; falling back to raw output",
+                    func.__name__,
+                    exc_info=True,
+                )
+                return method_output
 
         def _distill_and_store_lessons(
             flow_instance: Flow[Any], method_output: Any, raw_feedback: str
@@ -446,8 +471,19 @@ def human_feedback(
 
                 if lessons:
                     mem.remember_many(lessons, source=learn_source)  # type: ignore[union-attr]
-            except Exception:  # noqa: S110
-                pass  # non-critical: don't fail the flow because lesson storage failed
+            except Exception:
+                if learn_strict:
+                    logger.warning(
+                        "HITL lesson distillation failed for %s; re-raising (learn_strict=True)",
+                        func.__name__,
+                        exc_info=True,
+                    )
+                    raise
+                logger.warning(
+                    "HITL lesson distillation failed for %s; no lessons stored",
+                    func.__name__,
+                    exc_info=True,
+                )
 
         # -- Core feedback helpers ------------------------------------
 
@@ -654,6 +690,7 @@ def human_feedback(
             provider=provider,
             learn=learn,
             learn_source=learn_source,
+            learn_strict=learn_strict,
         )
         wrapper.__is_flow_method__ = True
 

lib/crewai/src/crewai/knowledge/knowledge.py

@@ -1,16 +1,89 @@
 import os
+from typing import Annotated, Any
 
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, BeforeValidator, ConfigDict, Field, PlainSerializer
 
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
+from crewai.knowledge.source.crew_docling_source import CrewDoclingSource
+from crewai.knowledge.source.csv_knowledge_source import CSVKnowledgeSource
+from crewai.knowledge.source.excel_knowledge_source import ExcelKnowledgeSource
+from crewai.knowledge.source.json_knowledge_source import JSONKnowledgeSource
+from crewai.knowledge.source.pdf_knowledge_source import PDFKnowledgeSource
+from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+from crewai.knowledge.source.text_file_knowledge_source import (
+    TextFileKnowledgeSource,
+)
 from crewai.knowledge.storage.knowledge_storage import KnowledgeStorage
+from crewai.rag.core.base_embeddings_provider import BaseEmbeddingsProvider
 from crewai.rag.embeddings.types import EmbedderConfig
 from crewai.rag.types import SearchResult
 
 
+_KNOWN_SOURCES: dict[str, type[BaseKnowledgeSource]] = {
+    "string": StringKnowledgeSource,
+    "docling": CrewDoclingSource,
+    "csv": CSVKnowledgeSource,
+    "excel": ExcelKnowledgeSource,
+    "json": JSONKnowledgeSource,
+    "pdf": PDFKnowledgeSource,
+    "text_file": TextFileKnowledgeSource,
+}
+
+
+def _resolve_knowledge_sources(value: Any) -> Any:
+    """Coerce list of dicts into typed BaseKnowledgeSource subclasses via source_type.
+
+    Pass-through for anything else (existing instances, mocks).
+    """
+    if not isinstance(value, list):
+        return value
+    resolved: list[Any] = []
+    for idx, item in enumerate(value):
+        if isinstance(item, dict):
+            tag = item.get("source_type")
+            if not isinstance(tag, str):
+                resolved.append(item)
+                continue
+            cls = _KNOWN_SOURCES.get(tag)
+            if cls is None:
+                raise ValueError(
+                    f"Unknown source_type={tag!r} at index {idx}: "
+                    f"expected one of {sorted(_KNOWN_SOURCES)}"
+                )
+            try:
+                resolved.append(cls.model_validate(item))
+            except Exception as exc:
+                raise ValueError(
+                    f"Failed to validate knowledge source at index {idx} "
+                    f"with source_type={tag!r}: {exc}"
+                ) from exc
+        else:
+            resolved.append(item)
+    return resolved
+
+
 os.environ["TOKENIZERS_PARALLELISM"] = "false"  # removes logging from fastembed
 
 
+def _serialize_embedder_spec(value: Any) -> dict[str, Any] | None:
+    if value is None:
+        return None
+    if isinstance(value, BaseEmbeddingsProvider):
+        return value.model_dump(mode="json")
+    if isinstance(value, dict):
+        return value
+    if isinstance(value, type) and issubclass(value, BaseEmbeddingsProvider):
+        raise TypeError(
+            f"Cannot checkpoint embedder class {value.__module__}.{value.__qualname__}: "
+            "build_embedder requires an instance or ProviderSpec dict, not a class. "
+            "Instantiate the provider before assigning it to Knowledge.embedder."
+        )
+    raise TypeError(
+        f"Cannot serialize embedder of type {type(value).__name__}: "
+        "expected ProviderSpec dict or BaseEmbeddingsProvider instance."
+    )
+
+
 class Knowledge(BaseModel):
     """
     Knowledge is a collection of sources and setup for the vector store to save and query relevant context.
@@ -20,10 +93,18 @@ class Knowledge(BaseModel):
         embedder: EmbedderConfig | None = None
     """
 
-    sources: list[BaseKnowledgeSource] = Field(default_factory=list)
+    sources: Annotated[
+        list[BaseKnowledgeSource],
+        BeforeValidator(_resolve_knowledge_sources),
+    ] = Field(default_factory=list)
     model_config = ConfigDict(arbitrary_types_allowed=True)
     storage: KnowledgeStorage | None = Field(default=None)
-    embedder: EmbedderConfig | None = None
+    embedder: Annotated[
+        EmbedderConfig | None,
+        PlainSerializer(
+            _serialize_embedder_spec, return_type=dict | None, when_used="json"
+        ),
+    ] = None
     collection_name: str | None = None
 
     def __init__(

lib/crewai/src/crewai/knowledge/source/base_knowledge_source.py

@@ -13,7 +13,9 @@ class BaseKnowledgeSource(BaseModel, ABC):
     chunk_size: int = 4000
     chunk_overlap: int = 200
     chunks: list[str] = Field(default_factory=list)
-    chunk_embeddings: list[np.ndarray[Any, np.dtype[Any]]] = Field(default_factory=list)
+    chunk_embeddings: list[np.ndarray[Any, np.dtype[Any]]] = Field(
+        default_factory=list, exclude=True
+    )
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
     storage: KnowledgeStorage | None = Field(default=None)

lib/crewai/src/crewai/knowledge/source/crew_docling_source.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from collections.abc import Iterator
 from pathlib import Path
-from typing import TYPE_CHECKING, Any
+from typing import TYPE_CHECKING, Any, Literal
 from urllib.parse import urlparse
 
 
@@ -45,6 +45,7 @@ class CrewDoclingSource(BaseKnowledgeSource):
 
     _logger: Logger = Logger(verbose=True)
 
+    source_type: Literal["docling"] = "docling"
     file_path: list[Path | str] | None = Field(default=None)
     file_paths: list[Path | str] = Field(default_factory=list)
     chunks: list[str] = Field(default_factory=list)

lib/crewai/src/crewai/knowledge/source/csv_knowledge_source.py

@@ -1,5 +1,6 @@
 import csv
 from pathlib import Path
+from typing import Literal
 
 from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledgeSource
 
@@ -7,6 +8,8 @@ from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledge
 class CSVKnowledgeSource(BaseFileKnowledgeSource):
     """A knowledge source that stores and queries CSV file content using embeddings."""
 
+    source_type: Literal["csv"] = "csv"
+
     def load_content(self) -> dict[Path, str]:
         """Load and preprocess CSV file content."""
         content_dict = {}

lib/crewai/src/crewai/knowledge/source/excel_knowledge_source.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 from types import ModuleType
-from typing import Any
+from typing import Any, Literal
 
 from pydantic import Field, field_validator
 
@@ -16,6 +16,7 @@ class ExcelKnowledgeSource(BaseKnowledgeSource):
 
     _logger: Logger = Logger(verbose=True)
 
+    source_type: Literal["excel"] = "excel"
     file_path: Path | list[Path] | str | list[str] | None = Field(
         default=None,
         description="[Deprecated] The path to the file. Use file_paths instead.",

lib/crewai/src/crewai/knowledge/source/json_knowledge_source.py

@@ -1,6 +1,6 @@
 import json
 from pathlib import Path
-from typing import Any
+from typing import Any, Literal
 
 from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledgeSource
 
@@ -8,6 +8,8 @@ from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledge
 class JSONKnowledgeSource(BaseFileKnowledgeSource):
     """A knowledge source that stores and queries JSON file content using embeddings."""
 
+    source_type: Literal["json"] = "json"
+
     def load_content(self) -> dict[Path, str]:
         """Load and preprocess JSON file content."""
         content: dict[Path, str] = {}

lib/crewai/src/crewai/knowledge/source/pdf_knowledge_source.py

@@ -1,5 +1,6 @@
 from pathlib import Path
 from types import ModuleType
+from typing import Literal
 
 from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledgeSource
 
@@ -7,6 +8,8 @@ from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledge
 class PDFKnowledgeSource(BaseFileKnowledgeSource):
     """A knowledge source that stores and queries PDF file content using embeddings."""
 
+    source_type: Literal["pdf"] = "pdf"
+
     def load_content(self) -> dict[Path, str]:
         """Load and preprocess PDF file content."""
         pdfplumber = self._import_pdfplumber()

lib/crewai/src/crewai/knowledge/source/string_knowledge_source.py

@@ -1,4 +1,4 @@
-from typing import Any
+from typing import Any, Literal
 
 from pydantic import Field
 
@@ -8,6 +8,7 @@ from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 class StringKnowledgeSource(BaseKnowledgeSource):
     """A knowledge source that stores and queries plain text content using embeddings."""
 
+    source_type: Literal["string"] = "string"
     content: str = Field(...)
     collection_name: str | None = Field(default=None)
 

lib/crewai/src/crewai/knowledge/source/text_file_knowledge_source.py

@@ -1,4 +1,5 @@
 from pathlib import Path
+from typing import Literal
 
 from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledgeSource
 
@@ -6,6 +7,8 @@ from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledge
 class TextFileKnowledgeSource(BaseFileKnowledgeSource):
     """A knowledge source that stores and queries text file content using embeddings."""
 
+    source_type: Literal["text_file"] = "text_file"
+
     def load_content(self) -> dict[Path, str]:
         """Load and preprocess text file content."""
         content = {}

lib/crewai/src/crewai/llm.py

@@ -940,6 +940,21 @@ class LLM(BaseLLM):
                 self._track_token_usage_internal(usage_info)
             self._handle_streaming_callbacks(callbacks, usage_info, last_chunk)
 
+            if accumulated_tool_args and not available_functions:
+                tool_calls_list: list[ChatCompletionDeltaToolCall] = [
+                    ChatCompletionDeltaToolCall(
+                        index=idx,
+                        function=Function(
+                            name=tool_arg.function.name,
+                            arguments=tool_arg.function.arguments,
+                        ),
+                    )
+                    for idx, tool_arg in sorted(accumulated_tool_args.items())
+                    if tool_arg.function.name
+                ]
+                if tool_calls_list:
+                    return tool_calls_list
+
             if not tool_calls or not available_functions:
                 if response_model and self.is_litellm:
                     instructor_instance = InternalInstructor(
@@ -1535,8 +1550,7 @@ class LLM(BaseLLM):
             if usage_info:
                 self._track_token_usage_internal(usage_info)
 
-            if accumulated_tool_args and available_functions:
-                # Convert accumulated tool args to ChatCompletionDeltaToolCall objects
+            if accumulated_tool_args:
                 tool_calls_list: list[ChatCompletionDeltaToolCall] = [
                     ChatCompletionDeltaToolCall(
                         index=idx,
@@ -1545,21 +1559,24 @@ class LLM(BaseLLM):
                             arguments=tool_arg.function.arguments,
                         ),
                     )
-                    for idx, tool_arg in accumulated_tool_args.items()
+                    for idx, tool_arg in sorted(accumulated_tool_args.items())
                     if tool_arg.function.name
                 ]
 
                 if tool_calls_list:
-                    result = self._handle_streaming_tool_calls(
-                        tool_calls=tool_calls_list,
-                        accumulated_tool_args=accumulated_tool_args,
-                        available_functions=available_functions,
-                        from_task=from_task,
-                        from_agent=from_agent,
-                        response_id=response_id,
-                    )
-                    if result is not None:
-                        return result
+                    if available_functions:
+                        result = self._handle_streaming_tool_calls(
+                            tool_calls=tool_calls_list,
+                            accumulated_tool_args=accumulated_tool_args,
+                            available_functions=available_functions,
+                            from_task=from_task,
+                            from_agent=from_agent,
+                            response_id=response_id,
+                        )
+                        if result is not None:
+                            return result
+                    else:
+                        return tool_calls_list
 
             usage_dict = self._usage_to_dict(usage_info)
             self._handle_emit_call_events(

lib/crewai/src/crewai/llms/base_llm.py

@@ -14,7 +14,7 @@ from datetime import datetime
 import json
 import logging
 import re
-from typing import TYPE_CHECKING, Any, Final, Literal
+from typing import TYPE_CHECKING, Any, Final, Literal, cast
 import uuid
 
 from pydantic import (
@@ -703,10 +703,19 @@ class BaseLLM(BaseModel, ABC):
         Raises:
             ValueError: If message format is invalid
         """
+        from crewai.llms.cache import CACHE_BREAKPOINT_KEY
+        from crewai.utilities.types import LLMMessage as _LLMMessage
+
         if isinstance(messages, str):
             return [{"role": "user", "content": messages}]
 
-        # Validate message format
+        # Validate then copy each message, dropping the cache-breakpoint
+        # flag in the copy only. The caller (e.g. CrewAgentExecutor,
+        # experimental.AgentExecutor) reuses its messages buffer across
+        # many LLM calls in the tool-use loop; mutating their dicts
+        # in place would erase the markers after the first call and
+        # break prompt caching for every subsequent iteration.
+        cleaned: list[LLMMessage] = []
         for i, msg in enumerate(messages):
             if not isinstance(msg, dict):
                 raise ValueError(f"Message at index {i} must be a dictionary")
@@ -714,8 +723,12 @@ class BaseLLM(BaseModel, ABC):
                 raise ValueError(
                     f"Message at index {i} must have 'role' and 'content' keys"
                 )
+            copy: dict[str, Any] = {
+                k: v for k, v in msg.items() if k != CACHE_BREAKPOINT_KEY
+            }
+            cleaned.append(cast(_LLMMessage, copy))
 
-        return self._process_message_files(messages)
+        return self._process_message_files(cleaned)
 
     def _process_message_files(self, messages: list[LLMMessage]) -> list[LLMMessage]:
         """Process files attached to messages and format for the provider.

lib/crewai/src/crewai/llms/cache.py

@@ -0,0 +1,37 @@
+"""Provider-agnostic prompt-cache breakpoint marker.
+
+Application code (prompt builders, agent executors) marks messages where a
+stable prefix ends. Provider adapters then translate the marker into the
+cache directive their API expects, or strip it for providers that cache
+implicitly (OpenAI, Gemini) or do not cache at all.
+
+Usage:
+
+    from crewai.llms.cache import mark_cache_breakpoint
+
+    messages = [
+        mark_cache_breakpoint({"role": "system", "content": stable_system}),
+        mark_cache_breakpoint({"role": "user", "content": stable_user_prefix}),
+        {"role": "user", "content": volatile_query},
+    ]
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+
+CACHE_BREAKPOINT_KEY = "cache_breakpoint"
+
+
+def mark_cache_breakpoint(message: dict[str, Any]) -> dict[str, Any]:
+    """Return ``message`` with the cache-breakpoint flag set.
+
+    Returns a new dict so callers can safely pass literal dicts.
+    """
+    return {**message, CACHE_BREAKPOINT_KEY: True}
+
+
+def strip_cache_breakpoint(message: dict[str, Any]) -> None:
+    """Remove the breakpoint flag from a message in place."""
+    message.pop(CACHE_BREAKPOINT_KEY, None)

lib/crewai/src/crewai/llms/providers/anthropic/completion.py

@@ -425,7 +425,7 @@ class AnthropicCompletion(BaseLLM):
     def _prepare_completion_params(
         self,
         messages: list[LLMMessage],
-        system_message: str | None = None,
+        system_message: str | list[dict[str, Any]] | None = None,
         tools: list[dict[str, Any]] | None = None,
         available_functions: dict[str, Any] | None = None,
     ) -> dict[str, Any]:
@@ -665,7 +665,7 @@ class AnthropicCompletion(BaseLLM):
 
     def _format_messages_for_anthropic(
         self, messages: str | list[LLMMessage]
-    ) -> tuple[list[LLMMessage], str | None]:
+    ) -> tuple[list[LLMMessage], str | list[dict[str, Any]] | None]:
         """Format messages for Anthropic API.
 
         Anthropic has specific requirements:
@@ -679,8 +679,51 @@ class AnthropicCompletion(BaseLLM):
             messages: Input messages
 
         Returns:
-            Tuple of (formatted_messages, system_message)
+            Tuple of (formatted_messages, system_message). `system_message` is
+            a list of content blocks (with cache_control stamped) when any
+            system message in the input carried a cache_breakpoint flag;
+            otherwise a plain string for backwards compatibility.
         """
+        from crewai.llms.cache import CACHE_BREAKPOINT_KEY
+
+        # Read cache_breakpoint flags from raw input BEFORE super strips them.
+        # We track the CONTENT of marked user/assistant messages so we can
+        # locate the corresponding block in formatted_messages — Anthropic
+        # rewrites tool results into user messages, so positional indices
+        # do not survive the conversion. We must stamp the original stable
+        # message (typically the initial task prompt), not whatever happens
+        # to be the trailing user-role block after tool_result expansion.
+        cache_system = False
+        cache_match_contents: list[str] = []
+        if not isinstance(messages, str):
+            for m in messages:
+                if not (isinstance(m, dict) and m.get(CACHE_BREAKPOINT_KEY)):
+                    continue
+                role = m.get("role")
+                if role == "system":
+                    cache_system = True
+                    continue
+                if role != "user":
+                    # Only user messages survive Anthropic's role-coalescing
+                    # in a stable, addressable position. Markers on assistant
+                    # or tool messages have no reliable stamp target after
+                    # tool_result expansion, so we ignore them.
+                    continue
+                raw_content = m.get("content")
+                if isinstance(raw_content, str) and raw_content:
+                    cache_match_contents.append(raw_content)
+                    continue
+                if isinstance(raw_content, list):
+                    # Pull text from a single-text-block list so callers that
+                    # pre-format content blocks still match cleanly.
+                    text_blocks = [
+                        b.get("text")
+                        for b in raw_content
+                        if isinstance(b, dict) and b.get("type") == "text"
+                    ]
+                    if len(text_blocks) == 1 and isinstance(text_blocks[0], str):
+                        cache_match_contents.append(text_blocks[0])
+
         # Use base class formatting first
         base_formatted = super()._format_messages(messages)
 
@@ -788,7 +831,62 @@ class AnthropicCompletion(BaseLLM):
             # If first message is not from user, insert a user message at the beginning
             formatted_messages.insert(0, {"role": "user", "content": "Hello"})
 
-        return formatted_messages, system_message
+        # Stamp cache_control on the message(s) whose original content was
+        # marked. We scan formatted_messages in order and stamp the first
+        # match per marked content — Anthropic permits up to 4 cache
+        # breakpoints per request, which is more than enough for our usage.
+        # Matching by content (rather than position) handles the ReAct
+        # case where tool_result blocks get expanded into trailing user
+        # messages: the stable initial-task prompt still maps cleanly.
+        for needle in cache_match_contents:
+            for fm in formatted_messages:
+                if fm.get("role") != "user":
+                    continue
+                content = fm.get("content")
+                if isinstance(content, str) and content == needle:
+                    self._stamp_cache_control_on_message(fm)
+                    break
+                if isinstance(content, list):
+                    fm_texts: list[str] = [
+                        b.get("text", "")
+                        for b in content
+                        if isinstance(b, dict) and b.get("type") == "text"
+                    ]
+                    if len(fm_texts) == 1 and fm_texts[0] == needle:
+                        self._stamp_cache_control_on_message(fm)
+                        break
+
+        # Convert system to content-block form when caching is requested.
+        system_payload: str | list[dict[str, Any]] | None = system_message
+        if system_message and cache_system:
+            system_payload = [
+                {
+                    "type": "text",
+                    "text": system_message,
+                    "cache_control": {"type": "ephemeral"},
+                }
+            ]
+
+        return formatted_messages, system_payload
+
+    @staticmethod
+    def _stamp_cache_control_on_message(message: LLMMessage) -> None:
+        """Stamp cache_control on the last content block of an Anthropic message."""
+        msg = cast(dict[str, Any], message)
+        content = msg.get("content")
+        if isinstance(content, str):
+            msg["content"] = [
+                {
+                    "type": "text",
+                    "text": content,
+                    "cache_control": {"type": "ephemeral"},
+                }
+            ]
+            return
+        if isinstance(content, list) and content:
+            last = content[-1]
+            if isinstance(last, dict):
+                last["cache_control"] = {"type": "ephemeral"}
 
     def _handle_completion(
         self,

lib/crewai/src/crewai/memory/memory_scope.py

@@ -6,6 +6,7 @@ from datetime import datetime
 from typing import Any, Literal
 
 from pydantic import BaseModel, ConfigDict, Field, PrivateAttr, model_validator
+from typing_extensions import Self
 
 from crewai.memory.types import (
     _RECALL_OVERSAMPLE_FACTOR,
@@ -16,15 +17,35 @@ from crewai.memory.types import (
 from crewai.memory.unified_memory import Memory
 
 
+def _ensure_memory_kind(value: Any) -> Any:
+    """Backfill ``memory_kind`` on legacy dicts that predate the discriminator.
+
+    Lets pre-1.14.6 configs/checkpoints flow into the discriminated
+    ``Memory | MemoryScope | MemorySlice`` union without crashing. Inference:
+    ``scopes`` key → ``slice``; ``root_path`` → ``scope``; else ``memory``.
+    Pass-through for non-dict values (instances, ``bool``, ``None``).
+    """
+    if isinstance(value, dict) and "memory_kind" not in value:
+        if "scopes" in value:
+            value["memory_kind"] = "slice"
+        elif "root_path" in value:
+            value["memory_kind"] = "scope"
+        else:
+            value["memory_kind"] = "memory"
+    return value
+
+
 class MemoryScope(BaseModel):
     """View of Memory restricted to a root path. All operations are scoped under that path."""
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
 
+    memory_kind: Literal["scope"] = "scope"
+
     root_path: str = Field(default="/")
 
-    _memory: Memory = PrivateAttr()
-    _root: str = PrivateAttr()
+    _memory: Memory | None = PrivateAttr(default=None)
+    _root: str = PrivateAttr(default="")
 
     @model_validator(mode="wrap")
     @classmethod
@@ -34,21 +55,38 @@ class MemoryScope(BaseModel):
             return data
         if not isinstance(data, dict):
             raise ValueError(f"Expected dict or MemoryScope, got {type(data).__name__}")
-        if "memory" not in data:
-            raise ValueError("MemoryScope requires a 'memory' key")
-        memory = data.pop("memory")
+        memory = data.pop("memory", None)
         instance: MemoryScope = handler(data)
-        instance._memory = memory
+        if memory is not None:
+            instance._memory = memory
         root = instance.root_path.rstrip("/") or ""
         if root and not root.startswith("/"):
             root = "/" + root
         instance._root = root
         return instance
 
+    def bind(self, memory: Memory) -> Self:
+        """Rebind the runtime ``Memory`` dependency after restore.
+
+        Required after deserializing from a checkpoint, since the live
+        ``Memory`` cannot be serialized.
+        """
+        self._memory = memory
+        return self
+
+    def _require_memory(self) -> Memory:
+        """Return the bound ``Memory`` or raise a clear error if missing."""
+        if self._memory is None:
+            raise RuntimeError(
+                "MemoryScope is not bound to a Memory; call .bind(memory) "
+                "after restore."
+            )
+        return self._memory
+
     @property
     def read_only(self) -> bool:
         """Whether the underlying memory is read-only."""
-        return self._memory.read_only
+        return self._require_memory().read_only
 
     def _scope_path(self, scope: str | None) -> str:
         if not scope or scope == "/":
@@ -73,7 +111,7 @@ class MemoryScope(BaseModel):
     ) -> MemoryRecord | None:
         """Remember content; scope is relative to this scope's root."""
         path = self._scope_path(scope)
-        return self._memory.remember(
+        return self._require_memory().remember(
             content,
             scope=path,
             categories=categories,
@@ -96,7 +134,7 @@ class MemoryScope(BaseModel):
     ) -> list[MemoryRecord]:
         """Remember multiple items; scope is relative to this scope's root."""
         path = self._scope_path(scope)
-        return self._memory.remember_many(
+        return self._require_memory().remember_many(
             contents,
             scope=path,
             categories=categories,
@@ -119,7 +157,7 @@ class MemoryScope(BaseModel):
     ) -> list[MemoryMatch]:
         """Recall within this scope (root path and below)."""
         search_scope = self._scope_path(scope) if scope else (self._root or "/")
-        return self._memory.recall(
+        return self._require_memory().recall(
             query,
             scope=search_scope,
             categories=categories,
@@ -131,7 +169,7 @@ class MemoryScope(BaseModel):
 
     def extract_memories(self, content: str) -> list[str]:
         """Extract discrete memories from content; delegates to underlying Memory."""
-        return self._memory.extract_memories(content)
+        return self._require_memory().extract_memories(content)
 
     def forget(
         self,
@@ -143,7 +181,7 @@ class MemoryScope(BaseModel):
     ) -> int:
         """Forget within this scope."""
         prefix = self._scope_path(scope) if scope else (self._root or "/")
-        return self._memory.forget(
+        return self._require_memory().forget(
             scope=prefix,
             categories=categories,
             older_than=older_than,
@@ -154,27 +192,27 @@ class MemoryScope(BaseModel):
     def list_scopes(self, path: str = "/") -> list[str]:
         """List child scopes under path (relative to this scope's root)."""
         full = self._scope_path(path)
-        return self._memory.list_scopes(full)
+        return self._require_memory().list_scopes(full)
 
     def info(self, path: str = "/") -> ScopeInfo:
         """Info for path under this scope."""
         full = self._scope_path(path)
-        return self._memory.info(full)
+        return self._require_memory().info(full)
 
     def tree(self, path: str = "/", max_depth: int = 3) -> str:
         """Tree under path within this scope."""
         full = self._scope_path(path)
-        return self._memory.tree(full, max_depth=max_depth)
+        return self._require_memory().tree(full, max_depth=max_depth)
 
     def list_categories(self, path: str | None = None) -> dict[str, int]:
         """Categories in this scope; path None means this scope root."""
         full = self._scope_path(path) if path else (self._root or "/")
-        return self._memory.list_categories(full)
+        return self._require_memory().list_categories(full)
 
     def reset(self, scope: str | None = None) -> None:
         """Reset within this scope."""
         prefix = self._scope_path(scope) if scope else (self._root or "/")
-        self._memory.reset(scope=prefix)
+        self._require_memory().reset(scope=prefix)
 
     def subscope(self, path: str) -> MemoryScope:
         """Return a narrower scope under this scope."""
@@ -191,11 +229,13 @@ class MemorySlice(BaseModel):
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
 
+    memory_kind: Literal["slice"] = "slice"
+
     scopes: list[str] = Field(default_factory=list)
     categories: list[str] | None = Field(default=None)
     read_only: bool = Field(default=True)
 
-    _memory: Memory = PrivateAttr()
+    _memory: Memory | None = PrivateAttr(default=None)
 
     @model_validator(mode="wrap")
     @classmethod
@@ -205,14 +245,27 @@ class MemorySlice(BaseModel):
             return data
         if not isinstance(data, dict):
             raise ValueError(f"Expected dict or MemorySlice, got {type(data).__name__}")
-        if "memory" not in data:
-            raise ValueError("MemorySlice requires a 'memory' key")
-        memory = data.pop("memory")
+        memory = data.pop("memory", None)
         data["scopes"] = [s.rstrip("/") or "/" for s in data.get("scopes", [])]
         instance: MemorySlice = handler(data)
-        instance._memory = memory
+        if memory is not None:
+            instance._memory = memory
         return instance
 
+    def bind(self, memory: Memory) -> Self:
+        """Rebind the runtime ``Memory`` dependency after restore."""
+        self._memory = memory
+        return self
+
+    def _require_memory(self) -> Memory:
+        """Return the bound ``Memory`` or raise a clear error if missing."""
+        if self._memory is None:
+            raise RuntimeError(
+                "MemorySlice is not bound to a Memory; call .bind(memory) "
+                "after restore."
+            )
+        return self._memory
+
     def remember(
         self,
         content: str,
@@ -226,7 +279,7 @@ class MemorySlice(BaseModel):
         """Remember into an explicit scope. No-op when read_only=True."""
         if self.read_only:
             return None
-        return self._memory.remember(
+        return self._require_memory().remember(
             content,
             scope=scope,
             categories=categories,
@@ -250,7 +303,7 @@ class MemorySlice(BaseModel):
         cats = categories or self.categories
         all_matches: list[MemoryMatch] = []
         for sc in self.scopes:
-            matches = self._memory.recall(
+            matches = self._require_memory().recall(
                 query,
                 scope=sc,
                 categories=cats,
@@ -272,14 +325,14 @@ class MemorySlice(BaseModel):
 
     def extract_memories(self, content: str) -> list[str]:
         """Extract discrete memories from content; delegates to underlying Memory."""
-        return self._memory.extract_memories(content)
+        return self._require_memory().extract_memories(content)
 
     def list_scopes(self, path: str = "/") -> list[str]:
         """List scopes across all slice roots."""
         out: list[str] = []
         for sc in self.scopes:
             full = f"{sc.rstrip('/')}{path}" if sc != "/" else path
-            out.extend(self._memory.list_scopes(full))
+            out.extend(self._require_memory().list_scopes(full))
         return sorted(set(out))
 
     def info(self, path: str = "/") -> ScopeInfo:
@@ -291,7 +344,7 @@ class MemorySlice(BaseModel):
         children: list[str] = []
         for sc in self.scopes:
             full = f"{sc.rstrip('/')}{path}" if sc != "/" else path
-            inf = self._memory.info(full)
+            inf = self._require_memory().info(full)
             total_records += inf.record_count
             all_categories.update(inf.categories)
             if inf.oldest_record:
@@ -321,6 +374,6 @@ class MemorySlice(BaseModel):
         counts: dict[str, int] = {}
         for sc in self.scopes:
             full = (f"{sc.rstrip('/')}{path}" if sc != "/" else path) if path else sc
-            for k, v in self._memory.list_categories(full).items():
+            for k, v in self._require_memory().list_categories(full).items():
                 counts[k] = counts.get(k, 0) + v
         return counts

lib/crewai/src/crewai/memory/unified_memory.py

@@ -63,6 +63,8 @@ class Memory(BaseModel):
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
 
+    memory_kind: Literal["memory"] = "memory"
+
     llm: Annotated[BaseLLM | str, PlainValidator(_passthrough)] = Field(
         default="gpt-4o-mini",
         description="LLM for analysis (model name or BaseLLM instance).",

lib/crewai/src/crewai/skills/__init__.py

@@ -3,15 +3,20 @@
 Provides filesystem-based skill packaging with progressive disclosure.
 """
 
+from crewai.skills.cache import SkillCacheManager
 from crewai.skills.loader import activate_skill, discover_skills
 from crewai.skills.models import Skill, SkillFrontmatter
 from crewai.skills.parser import SkillParseError
+from crewai.skills.registry import is_registry_ref, resolve_registry_ref
 
 
 __all__ = [
     "Skill",
+    "SkillCacheManager",
     "SkillFrontmatter",
     "SkillParseError",
     "activate_skill",
     "discover_skills",
+    "is_registry_ref",
+    "resolve_registry_ref",
 ]

lib/crewai/src/crewai/skills/cache.py

@@ -0,0 +1,148 @@
+"""Cache manager for registry-downloaded skills.
+
+Manages ~/.crewai/skills/{org}/{name}/ as the global skill cache.
+One version is stored per skill (last install wins).
+"""
+
+from __future__ import annotations
+
+from datetime import datetime, timezone
+import json
+import logging
+from pathlib import Path
+import tarfile
+from typing import TypedDict
+import zipfile
+
+
+_logger = logging.getLogger(__name__)
+
+
+_CACHE_ROOT = Path.home() / ".crewai" / "skills"
+_META_FILENAME = ".crewai_meta.json"
+
+
+class SkillMetadata(TypedDict):
+    org: str
+    name: str
+    version: str | None
+    installed_at: str
+
+
+class SkillCacheManager:
+    """Manages the global skill cache at ~/.crewai/skills/."""
+
+    def __init__(self, cache_root: Path | None = None) -> None:
+        self._root = cache_root or _CACHE_ROOT
+
+    def _skill_dir(self, org: str, name: str) -> Path:
+        return self._root / org / name
+
+    def get_cached_path(self, org: str, name: str) -> Path | None:
+        """Return the cached skill directory path if it exists, else None."""
+        skill_dir = self._skill_dir(org, name)
+        meta_file = skill_dir / _META_FILENAME
+        if skill_dir.is_dir() and meta_file.exists():
+            return skill_dir
+        return None
+
+    def store(
+        self, org: str, name: str, version: str | None, archive_bytes: bytes
+    ) -> Path:
+        """Unpack an archive into the cache and write metadata.
+
+        Uses tarfile with filter='data' for path-traversal protection.
+
+        Args:
+            org: Organisation slug.
+            name: Skill name.
+            version: Semantic version string, or None if unknown.
+            archive_bytes: Raw bytes of a .tar.gz archive.
+
+        Returns:
+            Path to the stored skill directory.
+        """
+        skill_dir = self._skill_dir(org, name)
+        # Wipe any previous version
+        if skill_dir.exists():
+            import shutil
+
+            shutil.rmtree(skill_dir)
+        skill_dir.mkdir(parents=True, exist_ok=True)
+
+        import io
+
+        # Try tar.gz first, fall back to zip
+        try:
+            with tarfile.open(fileobj=io.BytesIO(archive_bytes), mode="r:gz") as tf:
+                try:
+                    tf.extractall(skill_dir, filter="data")
+                except TypeError:
+                    _safe_extractall(tf, skill_dir)
+        except tarfile.TarError:
+            with zipfile.ZipFile(io.BytesIO(archive_bytes)) as zf:
+                _safe_extract_zip(zf, skill_dir)
+
+        meta: SkillMetadata = {
+            "org": org,
+            "name": name,
+            "version": version,
+            "installed_at": datetime.now(tz=timezone.utc).isoformat(),
+        }
+        (skill_dir / _META_FILENAME).write_text(json.dumps(meta, indent=2))
+        return skill_dir
+
+    def list_cached(self) -> list[SkillMetadata]:
+        """Return metadata for every cached skill."""
+        results: list[SkillMetadata] = []
+        if not self._root.exists():
+            return results
+        for org_dir in sorted(self._root.iterdir()):
+            if not org_dir.is_dir():
+                continue
+            for skill_dir in sorted(org_dir.iterdir()):
+                meta_file = skill_dir / _META_FILENAME
+                if meta_file.exists():
+                    try:
+                        results.append(json.loads(meta_file.read_text()))
+                    except (json.JSONDecodeError, KeyError):
+                        _logger.debug(
+                            "Skipping malformed cache entry: %s",
+                            meta_file,
+                            exc_info=True,
+                        )
+        return results
+
+    def invalidate(self, org: str, name: str) -> bool:
+        """Remove a cached skill.
+
+        Returns:
+            True if the cache entry existed and was removed, False otherwise.
+        """
+        skill_dir = self._skill_dir(org, name)
+        if skill_dir.exists():
+            import shutil
+
+            shutil.rmtree(skill_dir)
+            return True
+        return False
+
+
+def _safe_extractall(tf: tarfile.TarFile, dest: Path) -> None:
+    """Path-traversal-safe extraction for Python < 3.12."""
+    dest_resolved = dest.resolve()
+    for member in tf.getmembers():
+        member_path = (dest / member.name).resolve()
+        if not member_path.is_relative_to(dest_resolved):
+            raise ValueError(f"Blocked path traversal attempt: {member.name!r}")
+    tf.extractall(dest)  # noqa: S202
+
+
+def _safe_extract_zip(zf: zipfile.ZipFile, dest: Path) -> None:
+    """Path-traversal-safe ZIP extraction."""
+    dest_resolved = dest.resolve()
+    for member in zf.namelist():
+        member_path = (dest / member).resolve()
+        if not member_path.is_relative_to(dest_resolved):
+            raise ValueError(f"Blocked path traversal attempt: {member!r}")
+    zf.extractall(dest)  # noqa: S202

lib/crewai/src/crewai/skills/loader.py

@@ -161,6 +161,9 @@ def format_skill_context(skill: Skill) -> str:
     At METADATA level: returns name and description only.
     At INSTRUCTIONS level or above: returns full SKILL.md body.
 
+    Output is wrapped in <skill name="..."> XML tags so the block can serve
+    as a stable cache anchor when injected into the system prompt.
+
     Args:
         skill: The skill to format.
 
@@ -169,7 +172,7 @@ def format_skill_context(skill: Skill) -> str:
     """
     if skill.disclosure_level >= INSTRUCTIONS and skill.instructions:
         parts = [
-            f"## Skill: {skill.name}",
+            f'<skill name="{skill.name}">',
             skill.description,
             "",
             skill.instructions,
@@ -180,5 +183,6 @@ def format_skill_context(skill: Skill) -> str:
             for dir_name, files in sorted(skill.resource_files.items()):
                 if files:
                     parts.append(f"- **{dir_name}/**: {', '.join(files)}")
+        parts.append("</skill>")
         return "\n".join(parts)
-    return f"## Skill: {skill.name}\n{skill.description}"
+    return f'<skill name="{skill.name}">\n{skill.description}\n</skill>'