What is Petrichor?

Petrichor is a self-hosted knowledge platform designed for both humans and AI agents. It allows users to write in Markdown, import documents (PDF, DOCX, etc.), and automatically builds semantic wikis with entity extraction and cross-document relationships. The platform features agentic RAG with a two-stage retrieval process (search/outline then deep read) that produces traceable answers with evidence. It supports publishing content publicly, exposing knowledge via REST API and MCP for external AI agents, and exporting knowledge as OKF, Obsidian vaults, or Agent Skill packages. The stack includes Go/Gin API, React/Vite frontend, PostgreSQL with pgvector, Redis/Asynq for background jobs, and Caddy for HTTPS. It's deployed via Docker Compose with user-managed PostgreSQL, Redis, and S3-compatible storage.

Should you self-host Petrichor?

Petrichor scores 41/100 on selfhost.directory (how scoring works) and ranks #29 of 42 in Knowledge Management. It has 153 GitHub stars.

There was no stable release in the past 12 months, so check the repository's recent activity before you rely on it.

Setup is advanced: plan for several services and some ongoing maintenance. An official Docker image is available. It is Apache-2.0-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is Project N.O.M.A.D. (85/100, 41 133 stars). It scores higher overall, so compare the two before you pick. See Petrichor vs Project N.O.M.A.D. side by side.

Features

Agentic RAG with evidence
Code sandbox and browser automation
Document import and web capture
Full-text, vector, and wiki hybrid search
Jina AI reranker integrationOptional, disabled by default
Knowledge portability (OKF, Obsidian, Agent Skills)
Markdown writing and editing
MCP and REST API for external agents
Multi-model provider support
Pi Agent Runtime with dynamic skills
Public publishing and Q&A
Semantic Wiki generation

FAQ

开源版本使用什么许可证？

Petrichor uses the Apache-2.0 license. You can read, modify, and deploy the source code under the license terms, available in the GitHub repository.

如何开始使用？

You can try the online demo at wl.do, or deploy your own instance following the deployment documentation to manage your own knowledge, storage, and model services.

必须使用某一个 AI 模型吗？

No, Petrichor does not bind to a single AI provider. You can configure different models for chat, embeddings, and knowledge building according to your needs.

可以只作为私人知识库使用吗？

Yes, Petrichor can be used as a private knowledge base. Writing, document management, wiki building, and the built-in assistant all work for personal knowledge. Article sharing and public Q&A are optional publishing features.