This note is a reference for the retrieval layer under LLM applications. It starts with HelixDB, an open-source graph-vector database built in Rust on LMDB that keeps graph relationships and vector embeddings in one system. The context store section lists two tools that feed agents fresh material. Context7 is an MCP server that serves current official documentation and code examples to coding assistants. Airweave lets agents run semantic search across apps, databases, and APIs. The RAG section names the databases and frameworks used to build retrieval pipelines, including Weaviate, Milvus, Faiss, Cognita, LLMWare, and Quivr. GraphRAG closes that list by pairing retrieval with knowledge graphs. Use it to pick a store or framework, then follow the see-also links for RAG background and related data tooling.
Tools
- HelixDB is an open-source, high-performance graph-vector database designed for Retrieval-Augmented Generation (RAG) and AI applications. Built in Rust and powered by LMDB, it offers a unified platform for managing both graph relationships and vector embeddings, streamlining the development process for AI-driven systems.
Context Store
- Context7 - https://context7.com: Context7 is an MCP (Model Context Protocol) server developed by Upstash that provides up-to-date, real-time official documentation and code examples to large language models (LLMs) and AI coding assistants such as Cursor, Claude Desktop, VS Code, Windsurf, and others.
- Airweave AI: Airweave is a tool that lets agents run semantic search across any app. It is MCP-compatible and seamlessly connects any app, database, or API, turning their content into agent-ready knowledge.
RAG (Retrieval-Augmented Generation)
- Weaviate: Open-source, cloud-native vector database with semantic search and LLM integration.
- Milvus: Open-source vector database specialized in managing and searching vectors at large scale; distributed and highly scalable.
- Faiss: Facebook AI's efficient vector similarity search library, optimized for speed and memory; not a complete database.
- Cognita: Open-source framework for building RAG applications; simplifies development with a modular architecture that supports a wide range of components and advanced features.
- LLMWare: Open-source framework for building enterprise LLM applications (RAG, agents); provides RAG pipeline tooling, supports many data types, and can run on CPU.
- GraphRAG: A technique that combines RAG with knowledge graphs to improve LLM context, enabling deeper understanding, handling complex queries, improving explainability, and reducing noise.
- Trieve: All-in-one platform for search, recommendations, RAG, and analytics, delivered via API - https://github.com/devflowinc/trieve #RAG #search #vector
- Quivr: A RAG (Retrieval-Augmented Generation) solution for integrating GenAI into applications, supporting many large language models (LLMs) and vector stores - GitHub #RAG #LLM #vector
See also: RAG Overview · Knowledge And Data Management · Zvec