RagLeap Packages

Open-source Python RAG tooling. Pip-installable, MIT licensed.

pip install --index-url https://packages.ragleap.com/simple/ ragleap-rag
PyPI → Documentation → GitHub →
v0.12.1
Latest
238
Tests
6
Vector Backends
MIT
License

What's in the index

PackageVersionDescriptionLink
ragleap-rag 0.12.1 Core RAG library — pluggable embeddings, vector backends, reranking, evaluation PyPI
ragleap-graph 0.5.0 Knowledge graph extension for RagLeap RAG pipelines PyPI
This index mirrors packages already published on PyPI and proxies any other dependency straight from PyPI, so a normal pip install works with no extra configuration beyond the --index-url flag above.

Highlights

Pluggable embeddings

10+ providers supported — OpenAI, Mistral, Together, Ollama, Cohere, Voyage AI, Gemini, or any custom base_url endpoint.

6 vector backends

FAISS, PgVector, Pinecone, Weaviate, Qdrant, Milvus — swap backends without changing your pipeline code.

Cross-encoder reranking

ONNX Runtime based, ~23MB, no CUDA dependency required.

28-format ingestion

Document, image, audio, and video ingestion out of the box.

Query rewriting

Contextual rewriting, HyDE, and multi-query RAG-Fusion for better retrieval.

No hardcoded defaults

Every call requires an explicit model= parameter — no silent fallback to a model you didn't choose.