merged into open-legal-products/mike ๐
PR #466 โ fixed citation order across [[PAGE_BREAK]] segments
quotes were verifying as true even when out of document order โ fixed that
4.3k โญ OSS Legal AI platform
โ https://t.co/OhzCl3nDjp
#opensource#LegalAI#buildinpublic
just got merged into harvard-edge/cs249r_book ๐
PR #1962 โ recent-papers section, StaffML About page
28.4k โญ Harvard CS249r ML Systems textbook
โ https://t.co/YwqLxvzZLs
#opensource#MachineLearning#buildinpublic
I built ContextPulse โ MCP server that tracks your Claude Code context budget in real time ๐งต
โ Live token budget bar
โ Loop detection
โ Run diff engine(compare 2 agent runsโnobody has this)
โ Next.js dashboard + WebSocket
https://t.co/6JVAhuf7mV
#MCP#ClaudeCode#OpenSource
Opened my first PR to @LiveKit's `agents` repo (11k+ stars) ๐๏ธ
Found a bug where realtime transcription events were silently dropping their dedup key before reaching consumers. Small fix, CI green, waiting on review ๐ค
https://t.co/mGX8OtVJgV
8 Claude Code skills for NestJS โ module, TypeORM, BullMQ, WebSocket, JWT auth, Redis caching, Docker, testing.
Drop into ~/.claude/skills/ and Claude Code auto-discovers them.
https://t.co/skV1k3WOOP
#NestJS#TypeScript#ClaudeCode
Self-hosted MCP server registry and proxy โ TypeScript/NestJS.
The official MCP registry is Go-only and cloud-hosted.
This runs in your own infra with one docker compose up. ๐งต
Why I built this:
Every time I changed a prompt, I had no idea what broke.
No history. No comparison. No rollback.
prompt-vault fixes that โ version control for AI prompts, like Git but for LLMs.
Self-hosted. Your prompts stay yours.
๐ https://t.co/mfxnmQah5Dโญ
I built Prompt Vault โ self-hosted "Git for AI prompts" ๐
โ Version control for prompts โ Test runner (latency + tokens) โ A/B testing side by side โ Docker Compose, self-hosted
๐ https://t.co/mfxnmQah5D
#opensource#buildinpublic#nestjs#ai