AI analysis grounded in the code graph — computed facts, not vibes · 2026-09-23T03:03:03Z
WeKnora is a Go-based (with a Python MCP server component) knowledge platform that ingests raw documents — PDF, Word, Excel, images, and more — and exposes them through three mechanisms: a RAG pipeline for direct Q&A, a ReAct-style agent that orchestrates knowledge retrieval, MCP tools and web search for multi-step tasks, and a "Wiki mode" that has the agent generate and maintain interconnected markdown knowledge bases and knowledge graphs from source documents. It ships a CLI (weknora — see cli/cmd/root.go and subcommands like chat, agent, session, chunk, mcp), a router-based backend (internal/router/router.go), and integration surfaces for IM platforms (WeChat Work, Feishu, Slack, Telegram) and external knowledge sources (Feishu, Notion, Yuque, RSS). It targets teams that want a self-hosted, modular alternative to closed RAG/agent stacks.
The 1,302-star weekly gain lines up with a steady release cadence (v0.6.1 → v0.6.2 → v0.6.3 → v0.7.0 → v0.7.1 → v0.7.2 → v0.8.0 across roughly three months) and the README's Trendshift badge, suggesting external discovery/promotion rather than a single viral event. Recent commit activity is dense and feature-heavy (browser task preview, sandbox live shell, MCP OAuth2, chat UI overhaul), indicating the project is under active, visible development, which typically sustains star momentum on trending pages. The graph facts alone don't show marketing or launch events, so some of the growth is likely attributable to factors outside the repo (e.g. Trendshift listing, WeChat ecosystem ties) rather than purely code changes.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.