RAG · Search

Because making AI agents read Twitter, Reddit, or Bilibili usually means wrestling with a dozen scrapers, API keys, and cookie jars.

Builds interactive knowledge graphs that teach you a codebase instead of just flexing its complexity.

Most RAG systems treat PDFs as flat text; RAG-Anything parses them into text, images, tables, and equations so you can query the whole page.

Persistent memory for coding agents that currently forget everything when the session ends.

ScrapeGraphAI lets you extract structured data from websites and documents by describing what you want in plain English, leaving the LLM to wrestle with the markup.

Khoj exists to unify your documents, web search, and chosen LLM into a single self-hosted brain accessible from Emacs to WhatsApp.

MemPalace is a local-first AI memory system that stores conversations verbatim and retrieves them with semantic search backed by published, reproducible benchmarks.

It turns images and PDFs into structured JSON and Markdown so your RAG pipeline doesn't have to squint.

JeecgBoot exists so Java teams can describe an ERP in plain language, generate the Spring Boot and Vue code, and then manually merge it to keep things flexible.

Sim exists so you can orchestrate LLMs, tools, and vector stores on a visual canvas instead of writing another ad-hoc Python script.

Milvus is a Go/C++ vector database built for billion-scale ANN search with Kubernetes-native horizontal scaling.

Qdrant stores neural network outputs as searchable vectors and lets you filter them with SQL-like payload queries, bridging the gap between embedding models and production search.

A privacy-focused AI answering engine that runs entirely on your own hardware so your searches never leave the building.

Chroma is an open-source search backend that handles the messy embedding pipeline so AI applications can store and retrieve documents with a minimal API.

Because stitching together LLM workflows, RAG, agents, and observability by hand is a full-time job.

A CLI and library that turns prompt evaluation and red-teaming into repeatable, automatable engineering instead of vibe-based guesswork.

A maintainer cataloged every Chinese NLP repo they touched into a single, obsessively categorized list so others wouldn’t have to hunt.

Langchain-Chatchat wires LangChain to ChatGLM, Qwen, and Llama for fully offline knowledge-base Q&A with a Streamlit face.

GitNexus turns your codebase into a queryable dependency graph so AI coding agents stop shipping blind edits.

Why rebuild the same RAG pipeline or agent loop from scratch when you can fork a working template instead?