An AI knowledge hub that speaks WeChat and Markdown

What it does OmniBox is a cross-platform knowledge hub that collects webpages, office documents, audio files, and mobile captures into an indexed store for LLM Q&A and writing. It renders Markdown with diagrams, formulas, and music notation, and includes team permissions, multi-tenancy, and dark mode. The project spans separate repositories for the web frontend, backend, wizard, and browser extension.

The interesting bit The standout capture surface is a WeChat bot that saves files, videos, voice messages, text, and entire chat records—an ingestion vector most Western tools ignore. An iOS “Flash” mode for quick voice-and-text memos and broad file parsing suggest the authors built this for their own hoarding habits first.

Key highlights

• Browser extension strips and saves main webpage content.
• End-to-end parsing and indexing for PDF, Word, PPT, MP3, and more.
• Markdown editor renders formulas, mind maps, flowcharts, sequence diagrams, Gantt charts, and music notation.
• WeChat bot ingests files, videos, voice messages, text, and chat histories.
• iOS “Flash” capture for quick voice memos and text notes.

Caveats

• The README lists “Agent” as a topic and completed roadmap item, but never describes what the agent actually does.
• RSS subscription remains on the roadmap and is not yet implemented.
• The main repository functions as an integration umbrella; the active code lives across four separate sub-repositories.

Verdict Give it a spin if your knowledge workflow is scattered across WeChat threads, browser tabs, and voice memos. Look elsewhere if you want a minimalist, single-repo note-taking tool.