CommunitAI: The AI Chief of Staff for Communities

Empowering community leads with fine-tuned AI on DigitalOcean Gradient to automate meeting notes, track action items, and monitor community health in one production-ready dashboard.

Comment

The Problem

Community leaders — HOA boards, campus clubs, sports committees — spend hours every week manually taking meeting notes, chasing action items, and following up with members. Most of that work never gets done properly, and communities suffer for it.

What We Built

CommunitAI is an AI-powered Chief of Staff for community leaders. You upload a meeting recording (or record live), and CommunitAI handles everything else:

  • Transcription — converts audio to text automatically
  • AI Summary — clean, structured meeting summary in seconds
  • Action Item Extraction — every task, owner, and deadline pulled out automatically
  • Sentiment Analysis — health score per meeting so you know when morale is dipping
  • AI Follow-up Drafts — one click generates a personalised nudge for every task owner
  • Multi-workspace — manage your HOA, campus group, and sports club all in one place

How We Built It

  • Frontend: Next.js 15, Tailwind CSS v4, Framer Motion
  • Backend: FastAPI, SQLAlchemy (async), PostgreSQL, Alembic
  • AI Agent: DigitalOcean Gradient AI (serverless inference + ADK), Groq (Whisper transcription)
  • Storage: DigitalOcean Spaces (audio files, transcripts)
  • Deployment: DigitalOcean App Platform (backend + agent), Vercel (frontend)

The agent pipeline runs sequentially: transcribe → extract action items → sentiment analysis → summarise. Each step is fault-tolerant with retry-from-failure support.

What We Learned

  • Building reliable async pipelines with proper connection pool management (asyncpg + SQLAlchemy)
  • Structuring multi-tenant data cleanly across workspaces, meetings, and action items
  • How much time community leaders actually waste on admin — the problem is very real

Challenges

  • Getting the Gradient ADK package name right (gradient-adk, not gradient-ai-adk) — cost us a few failed deploys
  • Handling stale DB connections on DigitalOcean App Platform (solved with pool_pre_ping and pool_recycle)
  • Making the frontend feel fast and polished while keeping the codebase minimal

Built With

  • digitalocean-gradient-ai
  • digitalocean-spaces
  • fastapi
  • llama-3
  • next.js
  • python
  • react
  • tailwind-css
Share this project:

Updates