Canopy is an open-source Retrieval Augmented Generation (RAG) framework designed for developers and researchers to quickly build and experiment with RAG applications. It simplifies the entire RAG workflow, from data ingestion and embedding to query optimization and context generation, enabling users to augment LLMs with their own data and reduce hallucinations.

How It Works

Canopy implements a full RAG pipeline, comprising a KnowledgeBase for data chunking, embedding, and storage (Pinecone/Qdrant), a ContextEngine for retrieving relevant document chunks and formulating context, and a ChatEngine for managing chat history and generating LLM responses. This modular design allows for flexibility in data management and LLM integration.

Quick Start & Requirements

• Install: pip install canopy-sdk (with optional extras like grpc, torch, transformers, cohere, qdrant).
• Requirements: Pinecone API Key, OpenAI API Key, and an INDEX_NAME. Optional keys for Anyscale, Cohere, Jina, Azure OpenAI, and OctoAI.
• Setup: canopy new to create a Pinecone index, canopy upsert <data> to load documents, canopy start to run the server.
• Docs: Library Documentation

Highlighted Details

• Supports Pinecone and Qdrant as vector databases.
• Integrates with OpenAI, Azure OpenAI, Anyscale, and Cohere for embeddings and LLMs.
• Provides a REST API via FastAPI for easy integration with chat UIs.
• Includes a CLI chat tool for evaluating RAG vs. non-RAG workflows side-by-side.
• Can be migrated to by changing OpenAI API base URL to http://host:port/v1.

Maintenance & Community

The repository is no longer actively maintained by the Canopy team, who recommend the Pinecone Assistant for a managed RAG solution.

Licensing & Compatibility

The license is not explicitly stated in the README, but it is an open-source project from Pinecone. Compatibility for commercial use or closed-source linking would require clarification on the specific license.

Limitations & Caveats

The project is no longer maintained, meaning no future updates or bug fixes are expected. The evaluation chat tool's side-by-side comparison feature is currently only supported with OpenAI.