Show HN: WUPHF — Karpathy-Style LLM Wiki with Markdown + Git as Source of Truth

WUPHF is an open-source collaborative office for AI agents (Claude Code, Codex, OpenClaw, local LLMs via OpenCode) that includes a Karpathy-style wiki layer. The wiki uses Markdown and Git as the source of truth, stored at ~/.wuphf/wiki/, with a bleve (BM25) + SQLite index on top. No vector or graph DB is used yet — the goal is to see how far Markdown + Git can go before adding heavier infrastructure.
Key Features
- Each agent gets a private notebook at
agents/{slug}/notebook/plus shared team wiki atteam/. - Draft-to-wiki promotion flow: notebook entries are reviewed (by agent or human) and promoted to canonical wiki with back-links. A state machine handles expiry and auto-archive.
- Per-entity fact log: append-only JSONL at
team/entities/{kind}-{slug}.facts.jsonl. A synthesis worker rebuilds entity briefs every N facts. - Commits are attributed to a distinct Git identity ("Pam the Archivist") for provenance via
git log. - [[Wikilinks]] with broken-link detection (rendered in red).
- Daily lint cron for contradictions, stale entries, and broken wikilinks.
/lookupslash command + MCP tool for cited retrieval. Heuristic classifier routes short queries to BM25 and narrative queries to a cited-answer loop.
Retrieval Tuning
Current benchmark with 500 artifacts and 50 queries achieves 85% recall@20 on BM25 alone, which is the internal ship gate. If a query class drops below that, sqlite-vec is the pre-committed fallback.
Substrate Choices
- Markdown for durability — the wiki outlives the runtime; users can
git cloneand walk away with every byte. - Bleve for BM25.
- SQLite for structured metadata (facts, entities, edges, redirects, supersedes).
- Canonical IDs are first-class: fact IDs are deterministic (include sentence offset), slugs are assigned once and never renamed (redirect stubs used). Rebuild is logically identical, not byte-identical.
Known Limits
- 85% recall is not a universal guarantee — tuning ongoing.
- Synthesis quality depends on agent observation quality. The lint pass helps but is not a judgment engine.
- Single-office scope; no cross-office federation yet.
Demo & Install
A 5-minute terminal walkthrough is available at asciinema (script at ./scripts/demo-entity-synthesis.sh).
Install with: npx wuphf@latest
Build from source: git clone https://github.com/nex-crm/wuphf.git; go build -o wuphf ./cmd/wuphf
The wiki ships as part of WUPHF but can be used standalone. MIT license, self-hosted, bring-your-own keys.
📖 Read the full source: HN LLM Tools
👀 See Also

claude-real-video: Free Tool to Make Claude Watch Videos with Perception Layer
claude-real-video adds a local perception layer so Claude sees camera moves, pacing, gestures, and voice emotion. Demo watches NVIDIA GTC 2026 keynote and narrates screen action. Free version available.

Local Trello-style project manager for OpenClaw agents using markdown files
A developer built a local Trello board for managing projects with OpenClaw agents, using Node.js + Express for the API, React + react-trello for the UI, and markdown files with YAML frontmatter as the data layer. The system runs on the OpenClaw machine and is accessed locally, with agents reading/writing card files directly on the filesystem.

Microsoft Teams SDK Adds HTTP Server Adapter for Existing AI Agents
The Microsoft Teams SDK now includes an HTTP server adapter that lets developers connect existing AI agents to Teams without rewriting their code. It works with LangChain chains, Slack bots, and Azure Foundry deployments by injecting a POST /api/messages endpoint into existing Express servers.

LAP: 1,500+ API Specs Compiled for LLM Consumption to Reduce Claude Hallucinations
LAP is a tool that compiles 1,500+ real API specifications into a lean format optimized for LLMs, providing verified endpoints and parameters to prevent AI coding agents like Claude from hallucinating incorrect API calls.