Retrieval-Augmented Generation
Retrieval, embeddings, and crawling tools that ground LLM answers in your data.
Retrieval-Augmented Generation
Retrieval-augmented generation connects a language model to your documents, databases, or the web before it answers. This topic covers crawlers, vector search libraries, and complete RAG servers.
Use this comparison to assemble a pipeline for ingestion, embedding, retrieval, and answer generation, or to adopt an end-to-end service with an API.
15 projects
- open-webui/open-webui — Self-hosted web UI for chatting with Ollama and OpenAI-compatible models offline — Document ingestion: Yes; Vector database support: Yes; Hybrid search: No; HTTP API: Yes; Source citations: Yes; Agentic retrieval: No
- langchain-ai/langchain — Python framework for building agents and LLM applications from composable components — Agentic retrieval: Yes
- infiniflow/ragflow — RAG engine that turns PDFs and office documents into a citable context layer for LLMs — Document ingestion: Yes; Vector database support: Yes; Hybrid search: Yes; HTTP API: Yes; Source citations: Yes; Agentic retrieval: Yes
- CherryHQ/cherry-studio — Desktop AI workspace for models, agents, knowledge bases and images — Document ingestion: Yes; HTTP API: No; Agentic retrieval: Yes
- topoteretes/cognee — AI memory platform that turns documents, code and conversations into a searchable knowledge graph — Document ingestion: Yes; Vector database support: Yes; Hybrid search: Yes; HTTP API: Yes
- unclecode/crawl4ai — Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper. Don't be shy, join here: https://discord.gg/jP8KfhDhyN — Document ingestion: Yes; Vector database support: No; Hybrid search: Yes; HTTP API: Yes; Source citations: Yes; Agentic retrieval: No
- run-llama/llama_index — Framework for building RAG and agent applications over your own documents — Document ingestion: Yes; Vector database support: Yes; Hybrid search: Yes; HTTP API: No; Source citations: Yes; Agentic retrieval: Yes
- yifanfeng97/Hyper-Extract — CLI that turns documents into knowledge graphs and hypergraphs with LLMs — Document ingestion: Yes; HTTP API: No; Source citations: Partial; Agentic retrieval: No
- typesense/typesense — Typo-tolerant in-memory search engine with faceting, geo search and vector search — Document ingestion: No; Vector database support: Yes; Hybrid search: Yes; HTTP API: Yes; Source citations: No; Agentic retrieval: No
- dappros/ethora — Chat and messaging SDKs with AI bots for web, mobile and WordPress.
- memvid/memvid — Single-file memory layer for AI agents with instant retrieval and long-term recall — Vector database support: No; Hybrid search: No; Source citations: No; Agentic retrieval: No
- kelindar/search — Go library for embedded vector search and semantic embeddings using llama.cpp — Document ingestion: Yes; Vector database support: No; Hybrid search: No; HTTP API: No; Source citations: Yes; Agentic retrieval: No
- FrigadeHQ/trench — Open-Source Analytics Infrastructure. — Document ingestion: No; Vector database support: No; Hybrid search: No; HTTP API: No; Source citations: Yes; Agentic retrieval: No
- raksbisht/sitechat — Open-source, self-hosted AI chatbot trained on your website and docs. RAG, embeddable widget, no vendor lock-in. — Document ingestion: Yes; Vector database support: Yes; Hybrid search: No; HTTP API: Yes; Source citations: Yes; Agentic retrieval: No
- SciPhi-AI/R2R — SoTA production-ready AI retrieval system. Agentic Retrieval-Augmented Generation (RAG) with a RESTful API. — Document ingestion: No; Vector database support: No; Hybrid search: No; HTTP API: No; Source citations: No; Agentic retrieval: Yes