fix: HuggingFace embedder configuration validation error (#3995)

- Update HuggingFaceProvider to use HuggingFaceEmbeddingFunction instead of
  HuggingFaceEmbeddingServer for HuggingFace Inference API support
- Add api_key, model_name, and api_key_env_var fields to match documented config
- Accept api_url for compatibility but exclude from model_dump (not used by
  HuggingFace Inference API)
- Add validation aliases for model (maps to model_name) and environment variables
- Update HuggingFaceProviderConfig TypedDict with new fields
- Add comprehensive tests for HuggingFace provider configuration
- Regenerate uv.lock (was corrupted)

Fixes #3995

Co-Authored-By: João <joao@crewai.com>

lib/crewai-tools/pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
     "pytube>=15.0.0",
     "requests>=2.32.5",
     "docker>=7.1.0",
-    "crewai==1.6.0",
+    "crewai==1.6.1",
     "lancedb>=0.5.4",
     "tiktoken>=0.8.0",
     "beautifulsoup4>=4.13.4",

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

@@ -291,4 +291,4 @@ __all__ = [
     "ZapierActionTools",
 ]
 
-__version__ = "1.6.0"
+__version__ = "1.6.1"

lib/crewai/pyproject.toml

@@ -48,7 +48,7 @@ Repository = "https://github.com/crewAIInc/crewAI"
 
 [project.optional-dependencies]
 tools = [
-    "crewai-tools==1.6.0",
+    "crewai-tools==1.6.1",
 ]
 embeddings = [
     "tiktoken~=0.8.0"

lib/crewai/src/crewai/__init__.py

@@ -40,7 +40,7 @@ def _suppress_pydantic_deprecation_warnings() -> None:
 
 _suppress_pydantic_deprecation_warnings()
 
-__version__ = "1.6.0"
+__version__ = "1.6.1"
 _telemetry_submitted = False
 
 

lib/crewai/src/crewai/cli/templates/crew/pyproject.toml

@@ -5,7 +5,7 @@ description = "{{name}} using crewAI"
 authors = [{ name = "Your Name", email = "you@example.com" }]
 requires-python = ">=3.10,<3.14"
 dependencies = [
-    "crewai[tools]==1.6.0"
+    "crewai[tools]==1.6.1"
 ]
 
 [project.scripts]

lib/crewai/src/crewai/cli/templates/flow/pyproject.toml

@@ -5,7 +5,7 @@ description = "{{name}} using crewAI"
 authors = [{ name = "Your Name", email = "you@example.com" }]
 requires-python = ">=3.10,<3.14"
 dependencies = [
-    "crewai[tools]==1.6.0"
+    "crewai[tools]==1.6.1"
 ]
 
 [project.scripts]

lib/crewai/src/crewai/llm.py

@@ -406,46 +406,100 @@ class LLM(BaseLLM):
         instance.is_litellm = True
         return instance
 
+    @classmethod
+    def _matches_provider_pattern(cls, model: str, provider: str) -> bool:
+        """Check if a model name matches provider-specific patterns.
+
+        This allows supporting models that aren't in the hardcoded constants list,
+        including "latest" versions and new models that follow provider naming conventions.
+
+        Args:
+            model: The model name to check
+            provider: The provider to check against (canonical name)
+
+        Returns:
+            True if the model matches the provider's naming pattern, False otherwise
+        """
+        model_lower = model.lower()
+
+        if provider == "openai":
+            return any(
+                model_lower.startswith(prefix)
+                for prefix in ["gpt-", "o1", "o3", "o4", "whisper-"]
+            )
+
+        if provider == "anthropic" or provider == "claude":
+            return any(
+                model_lower.startswith(prefix) for prefix in ["claude-", "anthropic."]
+            )
+
+        if provider == "gemini" or provider == "google":
+            return any(
+                model_lower.startswith(prefix)
+                for prefix in ["gemini-", "gemma-", "learnlm-"]
+            )
+
+        if provider == "bedrock":
+            return "." in model_lower
+
+        if provider == "azure":
+            return any(
+                model_lower.startswith(prefix)
+                for prefix in ["gpt-", "gpt-35-", "o1", "o3", "o4", "azure-"]
+            )
+
+        return False
+
     @classmethod
     def _validate_model_in_constants(cls, model: str, provider: str) -> bool:
-        """Validate if a model name exists in the provider's constants.
+        """Validate if a model name exists in the provider's constants or matches provider patterns.
+
+        This method first checks the hardcoded constants list for known models.
+        If not found, it falls back to pattern matching to support new models,
+        "latest" versions, and models that follow provider naming conventions.
 
         Args:
             model: The model name to validate
             provider: The provider to check against (canonical name)
 
         Returns:
-            True if the model exists in the provider's constants, False otherwise
+            True if the model exists in constants or matches provider patterns, False otherwise
         """
-        if provider == "openai":
-            return model in OPENAI_MODELS
+        if provider == "openai" and model in OPENAI_MODELS:
+            return True
 
-        if provider == "anthropic" or provider == "claude":
-            return model in ANTHROPIC_MODELS
+        if (
+            provider == "anthropic" or provider == "claude"
+        ) and model in ANTHROPIC_MODELS:
+            return True
 
-        if provider == "gemini":
-            return model in GEMINI_MODELS
+        if (provider == "gemini" or provider == "google") and model in GEMINI_MODELS:
+            return True
 
-        if provider == "bedrock":
-            return model in BEDROCK_MODELS
+        if provider == "bedrock" and model in BEDROCK_MODELS:
+            return True
 
         if provider == "azure":
             # azure does not provide a list of available models, determine a better way to handle this
             return True
 
-        return False
+        # Fallback to pattern matching for models not in constants
+        return cls._matches_provider_pattern(model, provider)
 
     @classmethod
     def _infer_provider_from_model(cls, model: str) -> str:
         """Infer the provider from the model name.
 
+        This method first checks the hardcoded constants list for known models.
+        If not found, it uses pattern matching to infer the provider from model name patterns.
+        This allows supporting new models and "latest" versions without hardcoding.
+
         Args:
             model: The model name without provider prefix
 
         Returns:
             The inferred provider name, defaults to "openai"
         """
-
         if model in OPENAI_MODELS:
             return "openai"
 
@@ -1699,12 +1753,14 @@ class LLM(BaseLLM):
             max_tokens=self.max_tokens,
             presence_penalty=self.presence_penalty,
             frequency_penalty=self.frequency_penalty,
-            logit_bias=copy.deepcopy(self.logit_bias, memo)
-            if self.logit_bias
-            else None,
-            response_format=copy.deepcopy(self.response_format, memo)
-            if self.response_format
-            else None,
+            logit_bias=(
+                copy.deepcopy(self.logit_bias, memo) if self.logit_bias else None
+            ),
+            response_format=(
+                copy.deepcopy(self.response_format, memo)
+                if self.response_format
+                else None
+            ),
             seed=self.seed,
             logprobs=self.logprobs,
             top_logprobs=self.top_logprobs,

lib/crewai/src/crewai/llms/constants.py

@@ -182,6 +182,8 @@ OPENAI_MODELS: list[OpenAIModels] = [
 
 
 AnthropicModels: TypeAlias = Literal[
+    "claude-opus-4-5-20251101",
+    "claude-opus-4-5",
     "claude-3-7-sonnet-latest",
     "claude-3-7-sonnet-20250219",
     "claude-3-5-haiku-latest",
@@ -208,6 +210,8 @@ AnthropicModels: TypeAlias = Literal[
     "claude-3-haiku-20240307",
 ]
 ANTHROPIC_MODELS: list[AnthropicModels] = [
+    "claude-opus-4-5-20251101",
+    "claude-opus-4-5",
     "claude-3-7-sonnet-latest",
     "claude-3-7-sonnet-20250219",
     "claude-3-5-haiku-latest",
@@ -452,6 +456,7 @@ BedrockModels: TypeAlias = Literal[
     "anthropic.claude-3-sonnet-20240229-v1:0:28k",
     "anthropic.claude-haiku-4-5-20251001-v1:0",
     "anthropic.claude-instant-v1:2:100k",
+    "anthropic.claude-opus-4-5-20251101-v1:0",
     "anthropic.claude-opus-4-1-20250805-v1:0",
     "anthropic.claude-opus-4-20250514-v1:0",
     "anthropic.claude-sonnet-4-20250514-v1:0",
@@ -524,6 +529,7 @@ BEDROCK_MODELS: list[BedrockModels] = [
     "anthropic.claude-3-sonnet-20240229-v1:0:28k",
     "anthropic.claude-haiku-4-5-20251001-v1:0",
     "anthropic.claude-instant-v1:2:100k",
+    "anthropic.claude-opus-4-5-20251101-v1:0",
     "anthropic.claude-opus-4-1-20250805-v1:0",
     "anthropic.claude-opus-4-20250514-v1:0",
     "anthropic.claude-sonnet-4-20250514-v1:0",

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

@@ -8,6 +8,7 @@ from typing import TYPE_CHECKING, Any
 from pydantic import BaseModel
 
 from crewai.utilities.agent_utils import is_context_length_exceeded
+from crewai.utilities.converter import generate_model_description
 from crewai.utilities.exceptions.context_window_exceeding_exception import (
     LLMContextLengthExceededError,
 )
@@ -26,6 +27,7 @@ try:
     from azure.ai.inference.models import (
         ChatCompletions,
         ChatCompletionsToolCall,
+        JsonSchemaFormat,
         StreamingChatCompletionsUpdate,
     )
     from azure.core.credentials import (
@@ -278,13 +280,16 @@ class AzureCompletion(BaseLLM):
         }
 
         if response_model and self.is_openai_model:
-            params["response_format"] = {
-                "type": "json_schema",
-                "json_schema": {
-                    "name": response_model.__name__,
-                    "schema": response_model.model_json_schema(),
-                },
-            }
+            model_description = generate_model_description(response_model)
+            json_schema_info = model_description["json_schema"]
+            json_schema_name = json_schema_info["name"]
+
+            params["response_format"] = JsonSchemaFormat(
+                name=json_schema_name,
+                schema=json_schema_info["schema"],
+                description=f"Schema for {json_schema_name}",
+                strict=json_schema_info["strict"],
+            )
 
         # Only include model parameter for non-Azure OpenAI endpoints
         # Azure OpenAI endpoints have the deployment name in the URL
@@ -311,8 +316,8 @@ class AzureCompletion(BaseLLM):
             params["tool_choice"] = "auto"
 
         additional_params = self.additional_params
-        additional_drop_params = additional_params.get('additional_drop_params')
-        drop_params = additional_params.get('drop_params')
+        additional_drop_params = additional_params.get("additional_drop_params")
+        drop_params = additional_params.get("drop_params")
 
         if drop_params and isinstance(additional_drop_params, list):
             for drop_param in additional_drop_params:

lib/crewai/src/crewai/project/annotations.py

@@ -2,8 +2,10 @@
 
 from __future__ import annotations
 
+import asyncio
 from collections.abc import Callable
 from functools import wraps
+import inspect
 from typing import TYPE_CHECKING, Any, Concatenate, ParamSpec, TypeVar, overload
 
 from crewai.project.utils import memoize
@@ -156,6 +158,23 @@ def cache_handler(meth: Callable[P, R]) -> CacheHandlerMethod[P, R]:
     return CacheHandlerMethod(memoize(meth))
 
 
+def _call_method(method: Callable[..., Any], *args: Any, **kwargs: Any) -> Any:
+    """Call a method, awaiting it if async and running in an event loop."""
+    result = method(*args, **kwargs)
+    if inspect.iscoroutine(result):
+        try:
+            loop = asyncio.get_running_loop()
+        except RuntimeError:
+            loop = None
+        if loop and loop.is_running():
+            import concurrent.futures
+
+            with concurrent.futures.ThreadPoolExecutor() as pool:
+                return pool.submit(asyncio.run, result).result()
+        return asyncio.run(result)
+    return result
+
+
 @overload
 def crew(
     meth: Callable[Concatenate[SelfT, P], Crew],
@@ -198,7 +217,7 @@ def crew(
 
         # Instantiate tasks in order
         for _, task_method in tasks:
-            task_instance = task_method(self)
+            task_instance = _call_method(task_method, self)
             instantiated_tasks.append(task_instance)
             agent_instance = getattr(task_instance, "agent", None)
             if agent_instance and agent_instance.role not in agent_roles:
@@ -207,7 +226,7 @@ def crew(
 
         # Instantiate agents not included by tasks
         for _, agent_method in agents:
-            agent_instance = agent_method(self)
+            agent_instance = _call_method(agent_method, self)
             if agent_instance.role not in agent_roles:
                 instantiated_agents.append(agent_instance)
                 agent_roles.add(agent_instance.role)
@@ -215,7 +234,7 @@ def crew(
         self.agents = instantiated_agents
         self.tasks = instantiated_tasks
 
-        crew_instance = meth(self, *args, **kwargs)
+        crew_instance: Crew = _call_method(meth, self, *args, **kwargs)
 
         def callback_wrapper(
             hook: Callable[Concatenate[CrewInstance, P2], R2], instance: CrewInstance

lib/crewai/src/crewai/project/utils.py

@@ -1,7 +1,8 @@
 """Utility functions for the crewai project module."""
 
-from collections.abc import Callable
+from collections.abc import Callable, Coroutine
 from functools import wraps
+import inspect
 from typing import Any, ParamSpec, TypeVar, cast
 
 from pydantic import BaseModel
@@ -37,8 +38,8 @@ def _make_hashable(arg: Any) -> Any:
 def memoize(meth: Callable[P, R]) -> Callable[P, R]:
     """Memoize a method by caching its results based on arguments.
 
-    Handles Pydantic BaseModel instances by converting them to JSON strings
-    before hashing for cache lookup.
+    Handles both sync and async methods. Pydantic BaseModel instances are
+    converted to JSON strings before hashing for cache lookup.
 
     Args:
         meth: The method to memoize.
@@ -46,18 +47,16 @@ def memoize(meth: Callable[P, R]) -> Callable[P, R]:
     Returns:
         A memoized version of the method that caches results.
     """
+    if inspect.iscoroutinefunction(meth):
+        return cast(Callable[P, R], _memoize_async(meth))
+    return _memoize_sync(meth)
+
+
+def _memoize_sync(meth: Callable[P, R]) -> Callable[P, R]:
+    """Memoize a synchronous method."""
 
     @wraps(meth)
     def wrapper(*args: P.args, **kwargs: P.kwargs) -> R:
-        """Wrapper that converts arguments to hashable form before caching.
-
-        Args:
-            *args: Positional arguments to the memoized method.
-            **kwargs: Keyword arguments to the memoized method.
-
-        Returns:
-            The result of the memoized method call.
-        """
         hashable_args = tuple(_make_hashable(arg) for arg in args)
         hashable_kwargs = tuple(
             sorted((k, _make_hashable(v)) for k, v in kwargs.items())
@@ -73,3 +72,27 @@ def memoize(meth: Callable[P, R]) -> Callable[P, R]:
         return result
 
     return cast(Callable[P, R], wrapper)
+
+
+def _memoize_async(
+    meth: Callable[P, Coroutine[Any, Any, R]],
+) -> Callable[P, Coroutine[Any, Any, R]]:
+    """Memoize an async method."""
+
+    @wraps(meth)
+    async def wrapper(*args: P.args, **kwargs: P.kwargs) -> R:
+        hashable_args = tuple(_make_hashable(arg) for arg in args)
+        hashable_kwargs = tuple(
+            sorted((k, _make_hashable(v)) for k, v in kwargs.items())
+        )
+        cache_key = str((hashable_args, hashable_kwargs))
+
+        cached_result: R | None = cache.read(tool=meth.__name__, input=cache_key)
+        if cached_result is not None:
+            return cached_result
+
+        result = await meth(*args, **kwargs)
+        cache.add(tool=meth.__name__, input=cache_key, output=result)
+        return result
+
+    return wrapper

lib/crewai/src/crewai/project/wrappers.py

@@ -2,8 +2,10 @@
 
 from __future__ import annotations
 
+import asyncio
 from collections.abc import Callable
 from functools import partial
+import inspect
 from pathlib import Path
 from typing import (
     TYPE_CHECKING,
@@ -132,6 +134,22 @@ class CrewClass(Protocol):
     crew: Callable[..., Crew]
 
 
+def _resolve_result(result: Any) -> Any:
+    """Resolve a potentially async result to its value."""
+    if inspect.iscoroutine(result):
+        try:
+            loop = asyncio.get_running_loop()
+        except RuntimeError:
+            loop = None
+        if loop and loop.is_running():
+            import concurrent.futures
+
+            with concurrent.futures.ThreadPoolExecutor() as pool:
+                return pool.submit(asyncio.run, result).result()
+        return asyncio.run(result)
+    return result
+
+
 class DecoratedMethod(Generic[P, R]):
     """Base wrapper for methods with decorator metadata.
 
@@ -162,7 +180,12 @@ class DecoratedMethod(Generic[P, R]):
         """
         if obj is None:
             return self
-        bound = partial(self._meth, obj)
+        inner = partial(self._meth, obj)
+
+        def _bound(*args: Any, **kwargs: Any) -> R:
+            result: R = _resolve_result(inner(*args, **kwargs))  # type: ignore[call-arg]
+            return result
+
         for attr in (
             "is_agent",
             "is_llm",
@@ -174,8 +197,8 @@ class DecoratedMethod(Generic[P, R]):
             "is_crew",
         ):
             if hasattr(self, attr):
-                setattr(bound, attr, getattr(self, attr))
-        return bound
+                setattr(_bound, attr, getattr(self, attr))
+        return _bound
 
     def __call__(self, *args: P.args, **kwargs: P.kwargs) -> R:
         """Call the wrapped method.
@@ -236,6 +259,7 @@ class BoundTaskMethod(Generic[TaskResultT]):
             The task result with name ensured.
         """
         result = self._task_method.unwrap()(self._obj, *args, **kwargs)
+        result = _resolve_result(result)
         return self._task_method.ensure_task_name(result)
 
 
@@ -292,7 +316,9 @@ class TaskMethod(Generic[P, TaskResultT]):
         Returns:
             The task instance with name set if not already provided.
         """
-        return self.ensure_task_name(self._meth(*args, **kwargs))
+        result = self._meth(*args, **kwargs)
+        result = _resolve_result(result)
+        return self.ensure_task_name(result)
 
     def unwrap(self) -> Callable[P, TaskResultT]:
         """Get the original unwrapped method.

lib/crewai/src/crewai/rag/embeddings/providers/huggingface/huggingface_provider.py

@@ -1,21 +1,66 @@
 """HuggingFace embeddings provider."""
 
 from chromadb.utils.embedding_functions.huggingface_embedding_function import (
-    HuggingFaceEmbeddingServer,
+    HuggingFaceEmbeddingFunction,
 )
 from pydantic import AliasChoices, Field
 
 from crewai.rag.core.base_embeddings_provider import BaseEmbeddingsProvider
 
 
-class HuggingFaceProvider(BaseEmbeddingsProvider[HuggingFaceEmbeddingServer]):
-    """HuggingFace embeddings provider."""
+class HuggingFaceProvider(BaseEmbeddingsProvider[HuggingFaceEmbeddingFunction]):
+    """HuggingFace embeddings provider using the Inference API.
 
-    embedding_callable: type[HuggingFaceEmbeddingServer] = Field(
-        default=HuggingFaceEmbeddingServer,
+    This provider uses the HuggingFace Inference API for text embeddings.
+    It supports configuration via direct parameters or environment variables.
+
+    Example:
+        embedder={
+            "provider": "huggingface",
+            "config": {
+                "api_key": "your-hf-token",
+                "model": "sentence-transformers/all-MiniLM-L6-v2"
+            }
+        }
+    """
+
+    embedding_callable: type[HuggingFaceEmbeddingFunction] = Field(
+        default=HuggingFaceEmbeddingFunction,
         description="HuggingFace embedding function class",
     )
-    url: str = Field(
-        description="HuggingFace API URL",
-        validation_alias=AliasChoices("EMBEDDINGS_HUGGINGFACE_URL", "HUGGINGFACE_URL"),
+    api_key: str | None = Field(
+        default=None,
+        description="HuggingFace API key for authentication",
+        validation_alias=AliasChoices(
+            "EMBEDDINGS_HUGGINGFACE_API_KEY",
+            "HUGGINGFACE_API_KEY",
+            "HF_TOKEN",
+        ),
+    )
+    model_name: str = Field(
+        default="sentence-transformers/all-MiniLM-L6-v2",
+        description="Model name to use for embeddings",
+        validation_alias=AliasChoices(
+            "EMBEDDINGS_HUGGINGFACE_MODEL",
+            "HUGGINGFACE_MODEL",
+            "model",
+        ),
+    )
+    api_key_env_var: str = Field(
+        default="CHROMA_HUGGINGFACE_API_KEY",
+        description="Environment variable name containing the API key",
+        validation_alias=AliasChoices(
+            "EMBEDDINGS_HUGGINGFACE_API_KEY_ENV_VAR",
+            "HUGGINGFACE_API_KEY_ENV_VAR",
+        ),
+    )
+    api_url: str | None = Field(
+        default=None,
+        description="API URL (accepted for compatibility but not used by HuggingFace Inference API)",
+        validation_alias=AliasChoices(
+            "EMBEDDINGS_HUGGINGFACE_URL",
+            "HUGGINGFACE_URL",
+            "url",
+        ),
+        exclude=True,
     )

lib/crewai/src/crewai/rag/embeddings/providers/huggingface/types.py

@@ -6,8 +6,24 @@ from typing_extensions import Required, TypedDict
 
 
 class HuggingFaceProviderConfig(TypedDict, total=False):
-    """Configuration for HuggingFace provider."""
+    """Configuration for HuggingFace provider.
 
+    Supports HuggingFace Inference API for text embeddings.
+
+    Attributes:
+        api_key: HuggingFace API key for authentication.
+        model: Model name to use for embeddings (e.g., "sentence-transformers/all-MiniLM-L6-v2").
+        model_name: Alias for model.
+        api_key_env_var: Environment variable name containing the API key.
+        api_url: Optional API URL (accepted but not used, for compatibility).
+        url: Alias for api_url (accepted but not used, for compatibility).
+    """
+
+    api_key: str
+    model: str
+    model_name: str
+    api_key_env_var: str
+    api_url: str
     url: str
 
 

lib/crewai/tests/rag/embeddings/test_embedding_factory.py

@@ -176,6 +176,98 @@ class TestEmbeddingFactory:
             "crewai.rag.embeddings.providers.ibm.watsonx.WatsonXProvider"
         )
 
+    @patch("crewai.rag.embeddings.factory.import_and_validate_definition")
+    def test_build_embedder_huggingface(self, mock_import):
+        """Test building HuggingFace embedder with api_key and model."""
+        mock_provider_class = MagicMock()
+        mock_provider_instance = MagicMock()
+        mock_embedding_function = MagicMock()
+
+        mock_import.return_value = mock_provider_class
+        mock_provider_class.return_value = mock_provider_instance
+        mock_provider_instance.embedding_callable.return_value = mock_embedding_function
+
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "hf-test-key",
+                "model": "sentence-transformers/all-MiniLM-L6-v2",
+            },
+        }
+
+        build_embedder(config)
+
+        mock_import.assert_called_once_with(
+            "crewai.rag.embeddings.providers.huggingface.huggingface_provider.HuggingFaceProvider"
+        )
+        mock_provider_class.assert_called_once()
+
+        call_kwargs = mock_provider_class.call_args.kwargs
+        assert call_kwargs["api_key"] == "hf-test-key"
+        assert call_kwargs["model"] == "sentence-transformers/all-MiniLM-L6-v2"
+
+    @patch("crewai.rag.embeddings.factory.import_and_validate_definition")
+    def test_build_embedder_huggingface_with_api_url(self, mock_import):
+        """Test building HuggingFace embedder with api_url (for compatibility)."""
+        mock_provider_class = MagicMock()
+        mock_provider_instance = MagicMock()
+        mock_embedding_function = MagicMock()
+
+        mock_import.return_value = mock_provider_class
+        mock_provider_class.return_value = mock_provider_instance
+        mock_provider_instance.embedding_callable.return_value = mock_embedding_function
+
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "hf-test-key",
+                "model": "Qwen/Qwen3-Embedding-0.6B",
+                "api_url": "https://api-inference.huggingface.co",
+            },
+        }
+
+        build_embedder(config)
+
+        mock_import.assert_called_once_with(
+            "crewai.rag.embeddings.providers.huggingface.huggingface_provider.HuggingFaceProvider"
+        )
+        mock_provider_class.assert_called_once()
+
+        call_kwargs = mock_provider_class.call_args.kwargs
+        assert call_kwargs["api_key"] == "hf-test-key"
+        assert call_kwargs["model"] == "Qwen/Qwen3-Embedding-0.6B"
+        assert call_kwargs["api_url"] == "https://api-inference.huggingface.co"
+
+    @patch("crewai.rag.embeddings.factory.import_and_validate_definition")
+    def test_build_embedder_huggingface_with_model_name(self, mock_import):
+        """Test building HuggingFace embedder with model_name alias."""
+        mock_provider_class = MagicMock()
+        mock_provider_instance = MagicMock()
+        mock_embedding_function = MagicMock()
+
+        mock_import.return_value = mock_provider_class
+        mock_provider_class.return_value = mock_provider_instance
+        mock_provider_instance.embedding_callable.return_value = mock_embedding_function
+
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "hf-test-key",
+                "model_name": "sentence-transformers/all-MiniLM-L6-v2",
+            },
+        }
+
+        build_embedder(config)
+
+        mock_import.assert_called_once_with(
+            "crewai.rag.embeddings.providers.huggingface.huggingface_provider.HuggingFaceProvider"
+        )
+        mock_provider_class.assert_called_once()
+
+        call_kwargs = mock_provider_class.call_args.kwargs
+        assert call_kwargs["api_key"] == "hf-test-key"
+        assert call_kwargs["model_name"] == "sentence-transformers/all-MiniLM-L6-v2"
+
     def test_build_embedder_unknown_provider(self):
         """Test error handling for unknown provider."""
         config = {"provider": "unknown-provider", "config": {}}

lib/crewai/tests/rag/embeddings/test_huggingface_provider.py

@@ -0,0 +1,143 @@
+"""Tests for HuggingFace embedding provider."""
+
+import pytest
+from chromadb.utils.embedding_functions.huggingface_embedding_function import (
+    HuggingFaceEmbeddingFunction,
+)
+
+from crewai.rag.embeddings.factory import build_embedder
+from crewai.rag.embeddings.providers.huggingface.huggingface_provider import (
+    HuggingFaceProvider,
+)
+
+
+class TestHuggingFaceProvider:
+    """Test HuggingFace embedding provider."""
+
+    def test_provider_with_api_key_and_model(self):
+        """Test provider initialization with api_key and model.
+
+        This tests the fix for GitHub issue #3995 where users couldn't
+        configure HuggingFace embedder with api_key and model.
+        """
+        provider = HuggingFaceProvider(
+            api_key="test-hf-token",
+            model_name="sentence-transformers/all-MiniLM-L6-v2",
+        )
+
+        assert provider.api_key == "test-hf-token"
+        assert provider.model_name == "sentence-transformers/all-MiniLM-L6-v2"
+        assert provider.embedding_callable == HuggingFaceEmbeddingFunction
+
+    def test_provider_with_model_alias(self):
+        """Test provider initialization with 'model' alias for model_name."""
+        provider = HuggingFaceProvider(
+            api_key="test-hf-token",
+            model="Qwen/Qwen3-Embedding-0.6B",
+        )
+
+        assert provider.api_key == "test-hf-token"
+        assert provider.model_name == "Qwen/Qwen3-Embedding-0.6B"
+
+    def test_provider_with_api_url_compatibility(self):
+        """Test provider accepts api_url for compatibility but excludes it from model_dump.
+
+        The api_url parameter is accepted for compatibility with the documented
+        configuration format but is not passed to HuggingFaceEmbeddingFunction
+        since it uses a fixed API endpoint.
+        """
+        provider = HuggingFaceProvider(
+            api_key="test-hf-token",
+            model="sentence-transformers/all-MiniLM-L6-v2",
+            api_url="https://api-inference.huggingface.co",
+        )
+
+        assert provider.api_key == "test-hf-token"
+        assert provider.model_name == "sentence-transformers/all-MiniLM-L6-v2"
+        assert provider.api_url == "https://api-inference.huggingface.co"
+
+        # api_url should be excluded from model_dump
+        dumped = provider.model_dump(exclude={"embedding_callable"})
+        assert "api_url" not in dumped
+
+    def test_provider_default_model(self):
+        """Test provider uses default model when not specified."""
+        provider = HuggingFaceProvider(api_key="test-hf-token")
+
+        assert provider.model_name == "sentence-transformers/all-MiniLM-L6-v2"
+
+    def test_provider_default_api_key_env_var(self):
+        """Test provider uses default api_key_env_var."""
+        provider = HuggingFaceProvider(api_key="test-hf-token")
+
+        assert provider.api_key_env_var == "CHROMA_HUGGINGFACE_API_KEY"
+
+
+class TestHuggingFaceProviderIntegration:
+    """Integration tests for HuggingFace provider with build_embedder."""
+
+    def test_build_embedder_with_documented_config(self):
+        """Test build_embedder with the documented configuration format.
+
+        This tests the exact configuration format shown in the documentation
+        that was failing before the fix for GitHub issue #3995.
+        """
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "test-hf-token",
+                "model": "sentence-transformers/all-MiniLM-L6-v2",
+                "api_url": "https://api-inference.huggingface.co",
+            },
+        }
+
+        # This should not raise a validation error
+        embedder = build_embedder(config)
+
+        assert isinstance(embedder, HuggingFaceEmbeddingFunction)
+        assert embedder.model_name == "sentence-transformers/all-MiniLM-L6-v2"
+
+    def test_build_embedder_with_minimal_config(self):
+        """Test build_embedder with minimal configuration."""
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "test-hf-token",
+            },
+        }
+
+        embedder = build_embedder(config)
+
+        assert isinstance(embedder, HuggingFaceEmbeddingFunction)
+        # Default model should be used
+        assert embedder.model_name == "sentence-transformers/all-MiniLM-L6-v2"
+
+    def test_build_embedder_with_model_name_config(self):
+        """Test build_embedder with model_name instead of model."""
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "test-hf-token",
+                "model_name": "sentence-transformers/paraphrase-MiniLM-L6-v2",
+            },
+        }
+
+        embedder = build_embedder(config)
+
+        assert isinstance(embedder, HuggingFaceEmbeddingFunction)
+        assert embedder.model_name == "sentence-transformers/paraphrase-MiniLM-L6-v2"
+
+    def test_build_embedder_with_custom_model(self):
+        """Test build_embedder with a custom model name."""
+        config = {
+            "provider": "huggingface",
+            "config": {
+                "api_key": "test-hf-token",
+                "model": "Qwen/Qwen3-Embedding-0.6B",
+            },
+        }
+
+        embedder = build_embedder(config)
+
+        assert isinstance(embedder, HuggingFaceEmbeddingFunction)
+        assert embedder.model_name == "Qwen/Qwen3-Embedding-0.6B"

lib/crewai/tests/test_llm.py

@@ -243,7 +243,11 @@ def test_validate_call_params_not_supported():
 
     # Patch supports_response_schema to simulate an unsupported model.
     with patch("crewai.llm.supports_response_schema", return_value=False):
-        llm = LLM(model="gemini/gemini-1.5-pro", response_format=DummyResponse, is_litellm=True)
+        llm = LLM(
+            model="gemini/gemini-1.5-pro",
+            response_format=DummyResponse,
+            is_litellm=True,
+        )
         with pytest.raises(ValueError) as excinfo:
             llm._validate_call_params()
         assert "does not support response_format" in str(excinfo.value)
@@ -702,13 +706,16 @@ def test_ollama_does_not_modify_when_last_is_user(ollama_llm):
 
     assert formatted == original_messages
 
+
 def test_native_provider_raises_error_when_supported_but_fails():
     """Test that when a native provider is in SUPPORTED_NATIVE_PROVIDERS but fails to instantiate, we raise the error."""
     with patch("crewai.llm.SUPPORTED_NATIVE_PROVIDERS", ["openai"]):
         with patch("crewai.llm.LLM._get_native_provider") as mock_get_native:
             # Mock that provider exists but throws an error when instantiated
             mock_provider = MagicMock()
-            mock_provider.side_effect = ValueError("Native provider initialization failed")
+            mock_provider.side_effect = ValueError(
+                "Native provider initialization failed"
+            )
             mock_get_native.return_value = mock_provider
 
             with pytest.raises(ImportError) as excinfo:
@@ -751,16 +758,16 @@ def test_prefixed_models_with_valid_constants_use_native_sdk():
 
 
 def test_prefixed_models_with_invalid_constants_use_litellm():
-    """Test that models with native provider prefixes use LiteLLM when model is NOT in constants."""
+    """Test that models with native provider prefixes use LiteLLM when model is NOT in constants and does NOT match patterns."""
     # Test openai/ prefix with non-OpenAI model (not in OPENAI_MODELS) → LiteLLM
     llm = LLM(model="openai/gemini-2.5-flash", is_litellm=False)
     assert llm.is_litellm is True
     assert llm.model == "openai/gemini-2.5-flash"
 
-    # Test openai/ prefix with unknown future model → LiteLLM
-    llm2 = LLM(model="openai/gpt-future-6", is_litellm=False)
+    # Test openai/ prefix with model that doesn't match patterns (e.g. no gpt- prefix) → LiteLLM
+    llm2 = LLM(model="openai/custom-finetune-model", is_litellm=False)
     assert llm2.is_litellm is True
-    assert llm2.model == "openai/gpt-future-6"
+    assert llm2.model == "openai/custom-finetune-model"
 
     # Test anthropic/ prefix with non-Anthropic model → LiteLLM
     llm3 = LLM(model="anthropic/gpt-4o", is_litellm=False)
@@ -768,6 +775,21 @@ def test_prefixed_models_with_invalid_constants_use_litellm():
     assert llm3.model == "anthropic/gpt-4o"
 
 
+def test_prefixed_models_with_valid_patterns_use_native_sdk():
+    """Test that models matching provider patterns use native SDK even if not in constants."""
+    with patch.dict(os.environ, {"OPENAI_API_KEY": "test-key"}):
+        llm = LLM(model="openai/gpt-future-6", is_litellm=False)
+        assert llm.is_litellm is False
+        assert llm.provider == "openai"
+        assert llm.model == "gpt-future-6"
+
+    with patch.dict(os.environ, {"ANTHROPIC_API_KEY": "test-key"}):
+        llm2 = LLM(model="anthropic/claude-future-5", is_litellm=False)
+        assert llm2.is_litellm is False
+        assert llm2.provider == "anthropic"
+        assert llm2.model == "claude-future-5"
+
+
 def test_prefixed_models_with_non_native_providers_use_litellm():
     """Test that models with non-native provider prefixes always use LiteLLM."""
     # Test groq/ prefix (not a native provider) → LiteLLM
@@ -821,19 +843,36 @@ def test_validate_model_in_constants():
     """Test the _validate_model_in_constants method."""
     # OpenAI models
     assert LLM._validate_model_in_constants("gpt-4o", "openai") is True
-    assert LLM._validate_model_in_constants("gpt-future-6", "openai") is False
+    assert LLM._validate_model_in_constants("gpt-future-6", "openai") is True
+    assert LLM._validate_model_in_constants("o1-latest", "openai") is True
+    assert LLM._validate_model_in_constants("unknown-model", "openai") is False
 
     # Anthropic models
     assert LLM._validate_model_in_constants("claude-opus-4-0", "claude") is True
-    assert LLM._validate_model_in_constants("claude-future-5", "claude") is False
+    assert LLM._validate_model_in_constants("claude-future-5", "claude") is True
+    assert (
+        LLM._validate_model_in_constants("claude-3-5-sonnet-latest", "claude") is True
+    )
+    assert LLM._validate_model_in_constants("unknown-model", "claude") is False
 
     # Gemini models
     assert LLM._validate_model_in_constants("gemini-2.5-pro", "gemini") is True
-    assert LLM._validate_model_in_constants("gemini-future", "gemini") is False
+    assert LLM._validate_model_in_constants("gemini-future", "gemini") is True
+    assert LLM._validate_model_in_constants("gemma-3-latest", "gemini") is True
+    assert LLM._validate_model_in_constants("unknown-model", "gemini") is False
 
     # Azure models
     assert LLM._validate_model_in_constants("gpt-4o", "azure") is True
     assert LLM._validate_model_in_constants("gpt-35-turbo", "azure") is True
 
     # Bedrock models
-    assert LLM._validate_model_in_constants("anthropic.claude-opus-4-1-20250805-v1:0", "bedrock") is True
+    assert (
+        LLM._validate_model_in_constants(
+            "anthropic.claude-opus-4-1-20250805-v1:0", "bedrock"
+        )
+        is True
+    )
+    assert (
+        LLM._validate_model_in_constants("anthropic.claude-future-v1:0", "bedrock")
+        is True
+    )

lib/crewai/tests/test_project.py

@@ -272,6 +272,99 @@ def another_simple_tool():
     return "Hi!"
 
 
+class TestAsyncDecoratorSupport:
+    """Tests for async method support in @agent, @task decorators."""
+
+    def test_async_agent_memoization(self):
+        """Async agent methods should be properly memoized."""
+
+        class AsyncAgentCrew:
+            call_count = 0
+
+            @agent
+            async def async_agent(self):
+                AsyncAgentCrew.call_count += 1
+                return Agent(
+                    role="Async Agent", goal="Async Goal", backstory="Async Backstory"
+                )
+
+        crew = AsyncAgentCrew()
+        first_call = crew.async_agent()
+        second_call = crew.async_agent()
+
+        assert first_call is second_call, "Async agent memoization failed"
+        assert AsyncAgentCrew.call_count == 1, "Async agent called more than once"
+
+    def test_async_task_memoization(self):
+        """Async task methods should be properly memoized."""
+
+        class AsyncTaskCrew:
+            call_count = 0
+
+            @task
+            async def async_task(self):
+                AsyncTaskCrew.call_count += 1
+                return Task(
+                    description="Async Description", expected_output="Async Output"
+                )
+
+        crew = AsyncTaskCrew()
+        first_call = crew.async_task()
+        second_call = crew.async_task()
+
+        assert first_call is second_call, "Async task memoization failed"
+        assert AsyncTaskCrew.call_count == 1, "Async task called more than once"
+
+    def test_async_task_name_inference(self):
+        """Async task should have name inferred from method name."""
+
+        class AsyncTaskNameCrew:
+            @task
+            async def my_async_task(self):
+                return Task(
+                    description="Async Description", expected_output="Async Output"
+                )
+
+        crew = AsyncTaskNameCrew()
+        task_instance = crew.my_async_task()
+
+        assert task_instance.name == "my_async_task", (
+            "Async task name not inferred correctly"
+        )
+
+    def test_async_agent_returns_agent_not_coroutine(self):
+        """Async agent decorator should return Agent, not coroutine."""
+
+        class AsyncAgentTypeCrew:
+            @agent
+            async def typed_async_agent(self):
+                return Agent(
+                    role="Typed Agent", goal="Typed Goal", backstory="Typed Backstory"
+                )
+
+        crew = AsyncAgentTypeCrew()
+        result = crew.typed_async_agent()
+
+        assert isinstance(result, Agent), (
+            f"Expected Agent, got {type(result).__name__}"
+        )
+
+    def test_async_task_returns_task_not_coroutine(self):
+        """Async task decorator should return Task, not coroutine."""
+
+        class AsyncTaskTypeCrew:
+            @task
+            async def typed_async_task(self):
+                return Task(
+                    description="Typed Description", expected_output="Typed Output"
+                )
+
+        crew = AsyncTaskTypeCrew()
+        result = crew.typed_async_task()
+
+        assert isinstance(result, Task), f"Expected Task, got {type(result).__name__}"
+
+
 def test_internal_crew_with_mcp():
     from crewai_tools.adapters.tool_collection import ToolCollection
 

lib/devtools/src/crewai_devtools/__init__.py

@@ -1,3 +1,3 @@
 """CrewAI development tools."""
 
-__version__ = "1.6.0"
+__version__ = "1.6.1"