ezrag 0.2.0b1

Async-first RAG SDK. The fastest way to build RAG applications with Python.

ezrag

A modular, hexagonal-architecture RAG SDK for Python.

Install

pip install ezrag
# optional extras
pip install "ezrag[openai]"   # OpenAI embeddings + generation
pip install "ezrag[ollama]"   # Ollama (local LLM) support
pip install "ezrag[firestore]" # Firestore vector store

Quickstart

import asyncio
from ezrag import rag, Document
from ezrag.embeddings import OpenAIEmbedder
from ezrag.stores import InMemoryVectorStore
from ezrag.generators import OpenAIGenerator

async def main() -> None:
    sdk = rag(
        embedder=OpenAIEmbedder(api_key="..."),
        store=InMemoryVectorStore(),
        generator=OpenAIGenerator(api_key="..."),
    )
    await sdk.ingest([Document(content="RAG combines retrieval with generation.")])
    result = await sdk.query("What is RAG?")
    answer = await sdk.generate("Explain RAG simply")
    print(result.results)
    print(answer.answer)

asyncio.run(main())

Documentation

• docs/quickstart.md — install and first local RAG flow
• docs/extras.md — optional dependencies matrix
• docs/adapters.md — public adapter namespaces
• docs/stores.md — in-memory vs Firestore
• docs/chunkers.md — chunker selection guide
• docs/operations.md — ingest / query / generate / generate_stream
• docs/environments.md — local, OpenAI, Ollama, Firestore setup
• docs/limitations.md — beta constraints

Examples

Runnable scripts in examples/:

• examples/local_memory.py — zero-service, local adapters
• examples/openai_memory.py — OpenAI embedder + in-memory store
• examples/ollama_memory.py — Ollama embedder + in-memory store
• examples/openai_firestore.py — OpenAI embedder + Firestore store
• examples/openai_pgvector.py — OpenAI embedder + PgVectorStore