rag-python 0.3.0

Production-grade RAG for Python: multi-LLM, query rewriting, reranking, guardrails, and evaluation.

rag-python

rag-python is a production-oriented Python library for Retrieval-Augmented Generation (RAG).

Ingest your documents, ask questions, get grounded answers — with query rewriting, multi-query retrieval, reranking, guardrails, and multi-LLM support.

Author: Raghav Singla
Repository: github.com/RaghavOG/rag-python

---

Features

• Document pipeline: loaders → cleaning → chunking → embeddings → ChromaDB
• Query pipeline: rewriting → multi-query / hybrid retrieval → reranking
• Generation with guardrails (prompt injection + hallucination checks)
• Evaluation scores + self-correction retry loop
• LLM providers: OpenAI, Azure OpenAI, Anthropic, Gemini, Ollama
• Loaders: TXT, MD, PDF, DOCX, CSV, JSON, HTML

---

Install

pip install rag-python
# or from source
pip install -e .
# with reranking + extra providers
pip install -e ".[rerank,local,hybrid,anthropic,gemini,all]"

---

Quickstart

from rag_python import RAG

rag = RAG(
    llm_provider="openai",
    llm_model="gpt-4o-mini",
    embedding_provider="openai",
    embedding_model="text-embedding-3-small",
)

rag.ingest(["./data"], reindex=True)
answer = rag.query("How many days of annual leave?")
print(answer.text)

Hybrid search + metadata filter

from rag_python import RAG, SearchConfig

rag = RAG(
    retriever="hybrid",  # pip install rag-python[hybrid]
    metadata_filter={"filename": "leave-policy.pdf"},
)
rag.ingest(["./policies/leave-policy.pdf", "./policies/handbook.pdf"])
answer = rag.query("How many days of annual leave?")

CLI

export OPENAI_API_KEY=sk-...
rag-python ingest ./data --reindex
rag-python query "How many days of annual leave?" -v
rag-python query "leave policy" --retriever hybrid --metadata-filter '{"filename": "leave-policy.pdf"}'

---

Environment variables

Variable	Required	Description
OPENAI_API_KEY	For OpenAI	Default LLM + embeddings
ANTHROPIC_API_KEY	For Claude	LLM only
GEMINI_API_KEY	For Gemini	LLM only
AZURE_OPENAI_ENDPOINT	For Azure	Azure OpenAI
AZURE_OPENAI_API_KEY	For Azure	Azure OpenAI
OPENAI_API_VERSION	Azure	Default 2023-09-01-preview
OLLAMA_BASE_URL	Ollama	Default http://localhost:11434
RAG_PYTHON_DATA_DIR	Optional	Default ./data
RAG_PYTHON_CHROMA_DIR	Optional	Default ./chroma_db

See .env.example for all tuning options.

---

Project structure

.
├── src/rag_python/      # Installable package (PyPI: rag-python)
│   ├── client.py        # High-level RAG API
│   ├── rag_pipeline.py  # Full pipeline
│   └── providers/       # OpenAI, Azure, Anthropic, Gemini, Ollama
├── tests/
├── examples/
├── docs/
├── data/                # Sample documents
├── pyproject.toml
└── main.py              # Local dev CLI wrapper

---

Docs

• Usage
• Providers
• Changelog

---

License

MIT © Raghav Singla