Ollama Embedder - Phidata

The OllamaEmbedder can be used to embed text data into vectors locally using Ollama.

The model used for generating embeddings needs to run locally.

Usage

cookbook/embedders/ollama_embedder.py

from phi.agent import AgentKnowledge
from phi.vectordb.pgvector import PgVector
from phi.embedder.ollama import OllamaEmbedder

embeddings = OllamaEmbedder().get_embedding("The quick brown fox jumps over the lazy dog.")

# Print the embeddings and their dimensions
print(f"Embeddings: {embeddings[:5]}")
print(f"Dimensions: {len(embeddings)}")

# Example usage:
knowledge_base = AgentKnowledge(
    vector_db=PgVector(
        db_url="postgresql+psycopg://ai:ai@localhost:5532/ai",
        table_name="ollama_embeddings",
        embedder=OllamaEmbedder(),
    ),
    num_documents=2,
)

Params

Parameter	Type	Default	Description
`model`	`str`	`"openhermes"`	The name of the model used for generating embeddings.
`dimensions`	`int`	`4096`	The dimensionality of the embeddings generated by the model.
`host`	`str`	-	The host address for the API endpoint.
`timeout`	`Any`	-	The timeout duration for API requests.
`options`	`Any`	-	Additional options for configuring the API request.
`client_kwargs`	`Optional[Dict[str, Any]]`	-	Additional keyword arguments for configuring the API client. Optional.
`ollama_client`	`Optional[OllamaClient]`	-	An instance of the OllamaClient to use for making API requests. Optional.