Knowledge Storage

Vector store backends for knowledge and document retrieval

Knowledge Storage

Knowledge uses vector stores to index and retrieve documents. Choose the right backend for your use case.

Quick Setup

from praisonaiagents import Agent

# Default: Chroma (zero-config, local)
agent = Agent(knowledge=["docs/"])

# With specific provider (chroma ships out of the box)
agent = Agent(
    knowledge={
        "sources": ["docs/"],
        "vector_store": {
            "provider": "chroma",
            "config": {"collection_name": "my_docs"}
        }
    }
)

chroma is the only top-level Knowledge provider that ships out of the box. qdrant, pinecone, weaviate, milvus, and pgvector are plugin adapters — call register_knowledge_adapter(name, adapter_cls) before use. See Custom Knowledge Adapters.

Where is my data stored? By default, the Chroma store persists at <project>/.praisonai/knowledge/chroma (absolute, per project). Override it with vector_store.config.path. See Knowledge Storage.

Supported Vector Stores

Qdrant

High-performance vector search

Chroma

Zero-config local storage

Pinecone

Managed cloud service

Weaviate

Open-source, scalable

LanceDB

Serverless vector DB

Milvus

Enterprise-grade

PGVector

PostgreSQL extension

Cassandra

Distributed storage

ClickHouse

Analytics-optimized

Choosing a Vector Store

Use Case	Recommended	Why
Local development	Chroma	Zero config, embedded
Production (hosted)	Pinecone	Managed, scalable
Production (self-hosted)	Qdrant/Weaviate	Performant, open-source
PostgreSQL stack	PGVector	Same infrastructure
Enterprise	Milvus	High availability

Configuration Examples

Chroma

agent = Agent(
    knowledge={
        "sources": ["docs/"],
        "vector_store": {
            "provider": "chroma",
            "config": {
                "collection_name": "my_docs",
                "path": ".praison/chroma"
            }
        }
    }
)

Qdrant

agent = Agent(
    knowledge={
        "sources": ["docs/"],
        "vector_store": {
            "provider": "qdrant",
            "config": {
                "url": "http://localhost:6333",
                "collection_name": "my_docs"
            }
        }
    }
)

Pinecone

agent = Agent(
    knowledge={
        "sources": ["docs/"],
        "vector_store": {
            "provider": "pinecone",
            "config": {
                "api_key": "your-api-key",
                "index_name": "my-index"
            }
        }
    }
)

Related

Database Overview

All database backends

Knowledge Quick Start

Get started with knowledge