PgVector Vector Database

Use PgVector as a vector database for your Knowledge Base.

Setup

uv pip install -U sqlalchemy psycopg pgvector pypdf openai agno

Run PgVector with Docker:

docker run -d \
  -e POSTGRES_DB=ai \
  -e POSTGRES_USER=ai \
  -e POSTGRES_PASSWORD=ai \
  -e PGDATA=/var/lib/postgresql/data/pgdata \
  -v pgvolume:/var/lib/postgresql/data \
  -p 5532:5432 \
  --name pgvector \
  agnohq/pgvector:18

Example

agent_with_knowledge.py

from agno.agent import Agent
from agno.models.openai import OpenAIResponses
from agno.knowledge.knowledge import Knowledge
from agno.vectordb.pgvector import PgVector, SearchType

db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"
knowledge_base = Knowledge(
    vector_db=PgVector(table_name="recipes", db_url=db_url, search_type=SearchType.hybrid),
)

if __name__ == "__main__":
    knowledge_base.insert(
        url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
    )

    agent = Agent(
        model=OpenAIResponses(id="gpt-5.2"),
        knowledge=knowledge_base,
        # Add a tool to read chat history.
        read_chat_history=True,
        markdown=True,
        # debug_mode=True,
    )
    agent.print_response("How do I make chicken and galangal in coconut milk soup", stream=True)
    agent.print_response("What was my last question?", stream=True)

Async Support ⚡

PgVector also supports asynchronous operations, enabling concurrency and leading to better performance.

async_pgvector.py

import asyncio

from agno.agent import Agent
from agno.knowledge.knowledge import Knowledge
from agno.vectordb.pgvector import PgVector

db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"

vector_db = PgVector(table_name="recipes", db_url=db_url)

knowledge_base = Knowledge(
    vector_db=vector_db,
)

agent = Agent(knowledge=knowledge_base)

if __name__ == "__main__":
    # Load knowledge base asynchronously
    asyncio.run(knowledge_base.ainsert(
            url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
        )
    )

    # Create and use the agent asynchronously
    asyncio.run(agent.aprint_response("How to make Tom Kha Gai", markdown=True))

Use ainsert() and aprint_response() methods with asyncio.run() for non-blocking operations in high-throughput applications.

PgVector Params

Parameter	Type	Default	Description
table_name	str	-	Name of the table to store vector data.
schema	str	"ai"	Database schema name.
name	Optional[str]	None	Name of the vector database.
description	Optional[str]	None	Description of the vector database.
id	Optional[str]	None	ID of the vector database. Generated from db_url and table_name if not provided.
db_url	Optional[str]	None	Database connection URL.
db_engine	Optional[Engine]	None	SQLAlchemy database engine.
embedder	Optional[Embedder]	None	Embedder for creating embeddings. Defaults to OpenAIEmbedder if not provided.
search_type	SearchType	vector	Type of search to perform.
vector_index	Union[Ivfflat, HNSW]	HNSW()	Vector index configuration.
distance	Distance	cosine	Distance metric for vector comparisons.
prefix_match	bool	False	Enable prefix matching for full-text search.
vector_score_weight	float	0.5	Weight for vector similarity in hybrid search. Must be between 0 and 1.
content_language	str	"english"	Language for full-text search.
schema_version	int	1	Version of the database schema.
auto_upgrade_schema	bool	False	Automatically upgrade the schema if True.
reranker	Optional[Reranker]	None	Reranker for reranking search results.
create_schema	bool	True	Create the database schema if it does not exist. Set to False if schema is managed externally.
similarity_threshold	Optional[float]	None	Minimum similarity score (0.0-1.0) to filter results.