Implementation:CrewAIInc CrewAI Embedder Config

Metadata

Field	Value
Implementation Name	Embedder Config
Workflow	Knowledge_RAG_Pipeline
Category	Vector Representation
Repository	crewAIInc/crewAI
Implements	Principle:CrewAIInc_CrewAI_Embedding_Configuration

Overview

Concrete type alias and provider specifications for configuring embedding models across 18+ providers provided by the CrewAI RAG subsystem. The EmbedderConfig type alias unifies provider-specific TypedDicts, base provider instances, and provider classes into a single configurable type.

Source Reference

File	Lines
src/crewai/rag/embeddings/types.py	L32-76

Signature

EmbedderConfig: TypeAlias = (
    ProviderSpec
    | BaseEmbeddingsProvider[Any]
    | type[BaseEmbeddingsProvider[Any]]
)

Where ProviderSpec is a union of 18 provider-specific TypedDicts, each following the pattern:

class OpenAIProviderSpec(TypedDict):
    provider: Literal["openai"]
    config: dict[str, Any]

class OllamaProviderSpec(TypedDict):
    provider: Literal["ollama"]
    config: dict[str, Any]

# ... and so on for each provider

Allowed embedding providers:

AllowedEmbeddingProviders = Literal[
    "azure",
    "amazon-bedrock",
    "cohere",
    "custom",
    "google-generativeai",
    "google-vertex",
    "huggingface",
    "instructor",
    "jina",
    "ollama",
    "onnx",
    "openai",
    "openclip",
    "roboflow",
    "sentence-transformer",
    "text2vec",
    "voyageai",
    "watsonx",
]

Import

from crewai.rag.embeddings.types import EmbedderConfig

I/O Contract

Direction	Type	Description
Input	dict with provider and config keys	Provider specification dictionary matching one of the 18 provider TypedDicts
Input (alt)	BaseEmbeddingsProvider instance	A pre-configured embedding provider instance
Input (alt)	type[BaseEmbeddingsProvider]	An embedding provider class (instantiated by the framework)
Output	EmbedderConfig	Configuration value accepted by Knowledge, Crew, Agent, or KnowledgeStorage

Provider Configuration Details

Each provider accepts a config dictionary with provider-specific keys:

Provider	Common Config Keys
openai	model, api_key, api_base
ollama	model, url
cohere	model, api_key
huggingface	model, api_key
google-generativeai	model, api_key
azure	model, api_key, api_base, api_version
amazon-bedrock	model, region_name
sentence-transformer	model
jina	model, api_key
voyageai	model, api_key

Code Examples

OpenAI Embedder Configuration

from crewai.rag.embeddings.types import EmbedderConfig

openai_embedder: EmbedderConfig = {
    "provider": "openai",
    "config": {
        "model": "text-embedding-3-small",
        "api_key": "sk-...",
    },
}

Ollama Local Embedder Configuration

ollama_embedder: EmbedderConfig = {
    "provider": "ollama",
    "config": {
        "model": "nomic-embed-text",
        "url": "http://localhost:11434",
    },
}

Using Embedder Config with a Crew

from crewai import Crew

crew = Crew(
    agents=[...],
    tasks=[...],
    knowledge_sources=[...],
    embedder={
        "provider": "openai",
        "config": {
            "model": "text-embedding-3-small",
        },
    },
)

Using a Provider Instance

from crewai.rag.embeddings.types import EmbedderConfig

# You can also pass a pre-configured provider instance
# instead of a dictionary specification
custom_provider = MyCustomEmbeddingsProvider(model="my-model")
embedder_config: EmbedderConfig = custom_provider

Related Pages

• Principle:CrewAIInc_CrewAI_Embedding_Configuration -- The principle this implements
• Implementation:CrewAIInc_CrewAI_Knowledge_Source_Classes -- Source classes that produce chunks for embedding
• Implementation:CrewAIInc_CrewAI_Knowledge_Constructor -- Knowledge class that accepts embedder config
• Implementation:CrewAIInc_CrewAI_Knowledge_Storage_Search -- Storage that uses embedder for indexing and search
• Environment:CrewAIInc_CrewAI_Optional_Provider_Dependencies