mirror of
https://github.com/crewAIInc/crewAI.git
synced 2026-07-05 06:59:23 +00:00
a237ebabba
* feat: adopt directory-based docs versioning with Edge channel Switch docs.crewai.com from navigation-only versioning (every version selector entry rendered the same docs/<lang>/* source files) to Mintlify's directory-based versioning so each version selector entry renders its own snapshot. Add an "Edge" channel under docs/edge/<lang>/* that always reflects main HEAD for unreleased work, eliminating pre-release leakage onto frozen release labels. External links to canonical /<lang>/* URLs are preserved via wildcard redirects that always land on the current default version. Layout: - docs/edge/<lang>/* rolling source (you edit here) - docs/edge/enterprise-api.*.yaml - docs/v<X.Y.Z>/<lang>/* frozen, immutable snapshots - docs/v<X.Y.Z>/enterprise-api.*.yaml - docs/images/ shared, append-only - docs/docs.json nav + redirects URLs follow the Mintlify-idiomatic shape: /edge/<lang>/<page> for Edge, /v<X.Y.Z>/<lang>/<page> for every frozen snapshot. The wildcard redirects /<lang>/:slug* -> /<default>/<lang>/:slug* keep stale links working, and every freeze rewrites them (plus all per-section/per-page redirects) so destinations always resolve to the current default without depending on a second redirect hop. Release flow integration (devtools release): - New module crewai_devtools.docs_versioning.freeze() materialises docs/v<X.Y.Z>/ from docs/edge/, rewrites openapi: refs inside the snapshot, inserts the version into every language block in docs.json, and refreshes all redirect destinations. - _update_docs_and_create_pr() in cli.py now calls that freeze during Phase 2 of devtools release. Edge changelogs are updated first (so the snapshot freeze picks them up), then the snapshot is staged alongside docs.json, branched as docs/freeze-v<X.Y.Z>, and the PR is titled [docs-freeze] docs: snapshot and changelog for v<X.Y.Z> — the title prefix the new CI guard reads. - The PR still gates tag, GitHub release, PyPI publish, and the enterprise release as before; no new PRs are added. - Pre-releases (1.X.YaN, 1.X.YbN, ...) skip the snapshot — they ride Edge — and the docs PR title omits the [docs-freeze] prefix. - docs_check (AI-generated docs scaffolding) writes to docs/edge/<lang>/* so newly-generated unreleased docs land in Edge and never accidentally touch a frozen snapshot. Migration scripts (one-shot): - scripts/docs/freeze_historical_versions.py reconstructs all 16 historical snapshots (v1.10.0 .. v1.14.7) from git tags via git archive | tar, rewriting openapi: MDX refs so each snapshot reads its own enterprise-api YAML rather than the live one. - scripts/docs/prefix_version_paths.py one-shot-migrates docs.json: rewrites every page path in 16 versioned blocks to point under docs/v<X.Y.Z>/, inserts a new Edge entry per language, tags v1.14.7 as Latest (default), prunes pages whose target file doesn't exist in the snapshot (e.g. docs/ar/ didn't exist before v1.12.0), and writes the wildcard + per-section redirects. - scripts/docs/freeze_current_edge.py is now a thin CLI wrapper around docs_versioning.freeze for manual one-off freezes (e.g. retroactively snapshotting a forgotten release). CI guards (.github/workflows/docs-snapshots.yml): - Frozen snapshots under docs/v[0-9]*/ are immutable; only PRs whose title contains [docs-freeze] (i.e. release-cut PRs generated by devtools release or the manual wrapper) may modify them. - Images under docs/images/ are append-only since snapshots share a single image directory. Deleting or renaming an image breaks every historical snapshot that still references it. Restored docs/images/crewai-otel-export.png from PR #3673; it was deleted in PR #4908 but v1.10.0 / v1.10.1 snapshots still reference it. Restoring instead of editing the snapshots preserves historical rendering fidelity and validates the new append-only rule retroactively. Tests: - lib/devtools/tests/test_docs_versioning.py covers the freeze: file copy, openapi rewrite, version insertion, default demotion, redirect upserts, per-section redirect rewriting, idempotency, and invalid inputs. Verified locally with mintlify broken-links: 0 broken links across the full site (Edge + 16 frozen versions, 4 locales). AGENTS.md (repo root) is the contributor guide for the new model; RELEASING.md is the release-cut runbook; README's Contribution section links to both. Co-authored-by: Cursor <cursoragent@cursor.com> * style: resolve linter issues --------- Co-authored-by: Cursor <cursoragent@cursor.com>
133 lines
5.4 KiB
Plaintext
133 lines
5.4 KiB
Plaintext
---
|
|
title: "Training Crews"
|
|
description: "Train your deployed crews directly from the CrewAI AMP platform to improve agent performance over time"
|
|
icon: "dumbbell"
|
|
mode: "wide"
|
|
---
|
|
|
|
Training lets you improve crew performance by running iterative training sessions directly from the **Training** tab in CrewAI AMP. The platform uses **auto-train mode** — it handles the iterative process automatically, unlike CLI training which requires interactive human feedback per iteration.
|
|
|
|
After training completes, CrewAI evaluates agent outputs and consolidates feedback into actionable suggestions for each agent. These suggestions are then applied to future crew runs to improve output quality.
|
|
|
|
<Tip>
|
|
For details on how CrewAI training works under the hood, see the [Training Concepts](/en/concepts/training) page.
|
|
</Tip>
|
|
|
|
## Prerequisites
|
|
|
|
<CardGroup cols={2}>
|
|
<Card title="Active deployment" icon="rocket">
|
|
You need a CrewAI AMP account with an active deployment in **Ready** status (Crew type).
|
|
</Card>
|
|
<Card title="Run permission" icon="key">
|
|
Your account must have run permission for the deployment you want to train.
|
|
</Card>
|
|
</CardGroup>
|
|
|
|
## How to train a crew
|
|
|
|
<Steps>
|
|
<Step title="Open the Training tab">
|
|
Navigate to **Deployments**, click your deployment, then select the **Training** tab.
|
|
</Step>
|
|
|
|
<Step title="Enter a training name">
|
|
Provide a **Training Name** — this becomes the `.pkl` filename used to store training results. For example, "Expert Mode Training" produces `expert_mode_training.pkl`.
|
|
</Step>
|
|
|
|
<Step title="Fill in the crew inputs">
|
|
Enter the crew's input fields. These are the same inputs you'd provide for a normal kickoff — they're dynamically loaded based on your crew's configuration.
|
|
</Step>
|
|
|
|
<Step title="Start training">
|
|
Click **Train Crew**. The button changes to "Training..." with a spinner while the process runs.
|
|
|
|
Behind the scenes:
|
|
- A training record is created for your deployment
|
|
- The platform calls the deployment's auto-train endpoint
|
|
- The crew runs its iterations automatically — no manual feedback required
|
|
</Step>
|
|
|
|
<Step title="Monitor progress">
|
|
The **Current Training Status** panel displays:
|
|
- **Status** — Current state of the training run
|
|
- **Nº Iterations** — Number of training iterations configured
|
|
- **Filename** — The `.pkl` file being generated
|
|
- **Started At** — When training began
|
|
- **Training Inputs** — The inputs you provided
|
|
</Step>
|
|
</Steps>
|
|
|
|
## Understanding training results
|
|
|
|
Once training completes, you'll see per-agent result cards with the following information:
|
|
|
|
- **Agent Role** — The name/role of the agent in your crew
|
|
- **Final Quality** — A score from 0 to 10 evaluating the agent's output quality
|
|
- **Final Summary** — A summary of the agent's performance during training
|
|
- **Suggestions** — Actionable recommendations for improving the agent's behavior
|
|
|
|
### Editing suggestions
|
|
|
|
You can refine the suggestions for any agent:
|
|
|
|
<Steps>
|
|
<Step title="Click Edit">
|
|
On any agent's result card, click the **Edit** button next to the suggestions.
|
|
</Step>
|
|
|
|
<Step title="Modify suggestions">
|
|
Update the suggestions text to better reflect the improvements you want.
|
|
</Step>
|
|
|
|
<Step title="Save changes">
|
|
Click **Save**. The edited suggestions sync back to the deployment and are used in all future runs.
|
|
</Step>
|
|
</Steps>
|
|
|
|
## Using trained data
|
|
|
|
To apply training results to your crew:
|
|
|
|
1. Note the **Training Filename** (the `.pkl` file) from your completed training session.
|
|
2. Specify this filename in your deployment's kickoff or run configuration.
|
|
3. The crew automatically loads the training file and applies the stored suggestions to each agent.
|
|
|
|
This means agents benefit from the feedback generated during training on every subsequent run.
|
|
|
|
## Previous trainings
|
|
|
|
The bottom of the Training tab displays a **history of all past training sessions** for the deployment. Use this to review previous training runs, compare results, or select a different training file to use.
|
|
|
|
## Error handling
|
|
|
|
If a training run fails, the status panel shows an error state along with a message describing what went wrong.
|
|
|
|
Common causes of training failures:
|
|
- **Deployment runtime not updated** — Ensure your deployment is running the latest version
|
|
- **Crew execution errors** — Issues within the crew's task logic or agent configuration
|
|
- **Network issues** — Connectivity problems between the platform and the deployment
|
|
|
|
## Limitations
|
|
|
|
<Info>
|
|
Keep these constraints in mind when planning your training workflow:
|
|
- **One active training at a time** per deployment — wait for the current run to finish before starting another
|
|
- **Auto-train mode only** — the platform does not support interactive per-iteration feedback like the CLI does
|
|
- **Training data is deployment-specific** — training results are tied to the specific deployment instance and version
|
|
</Info>
|
|
|
|
## Related resources
|
|
|
|
<CardGroup cols={3}>
|
|
<Card title="Training Concepts" icon="book" href="/en/concepts/training">
|
|
Learn how CrewAI training works under the hood.
|
|
</Card>
|
|
<Card title="Kickoff Crew" icon="play" href="/en/enterprise/guides/kickoff-crew">
|
|
Run your deployed crew from the AMP platform.
|
|
</Card>
|
|
<Card title="Deploy to AMP" icon="cloud-arrow-up" href="/en/enterprise/guides/deploy-to-amp">
|
|
Get your crew deployed and ready for training.
|
|
</Card>
|
|
</CardGroup>
|