CRMind

CRMind is an entity-first, citation-aware CRM intelligence platform. It resolves companies and people to canonical records, retrieves source-backed evidence using hybrid search, runs workflow agents for lead finding and account briefs, and returns structured JSON where claims include citations.

Architecture Overview

• Backend: FastAPI + asyncpg + pgvector
• Agents: LangGraph-compatible workflow runtime with grounded retrieval
• Data: PostgreSQL (companies, people, signals, chunks, agent_runs)
• Frontend: Next.js 14 App Router under frontend
• API docs: http://localhost:8000/docs

Quick Start

docker compose up -d
cp .env.example .env
python scripts/init_db.py
python scripts/seed_demo_data.py
uvicorn backend.main:app --reload
cd frontend && npm install && npm run dev

Demo Queries

• Find senior engineers at Example Inc
• Show recent hiring signals for Example Inc
• Account brief for Example Inc
• Why is the record for example.com stale?

API Reference

OpenAPI docs are available at:

• http://localhost:8000/docs

Deployment

Vercel (frontend)

1. Import only the frontend directory as a Vercel project.
2. Set these Vercel environment variables:
   • NEXT_PUBLIC_API_BASE_URL=https://<your-render-service>.onrender.com/api/v1
   • NEXT_PUBLIC_SUPABASE_URL=https://<your-project>.supabase.co
   • NEXT_PUBLIC_SUPABASE_ANON_KEY=<supabase-anon-key>
3. Keep frontend/vercel.json as-is for cache headers.
4. Deploy with the Vercel Hobby (free) tier.

Render (backend)

1. Create a Render Web Service from render.yaml (Starter free tier).
2. Set required env vars in Render:
   • DATABASE_URL and DATABASE_URL_DIRECT (Supabase pooled/direct URLs)
   • SUPABASE_URL, SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY, SUPABASE_JWT_SECRET
   • OPENAI_API_KEY, GROQ_API_KEY, API_KEY, INTERNAL_SERVICE_API_KEY
   • DB_POOL_MAX_SIZE=5 (important for Supabase free connection limits)
3. Confirm health endpoint at /api/v1/health.
4. Keep free-tier usage low by using scheduled ingest and query caching.

Supabase (database)

1. Create a Supabase free project.
2. Enable required extensions in SQL editor:
   • CREATE EXTENSION IF NOT EXISTS vector;
   • CREATE EXTENSION IF NOT EXISTS pg_trgm;
3. Run migrations with Alembic or python scripts/init_db.py.
4. Use Supabase pooled connection for DATABASE_URL and direct connection for DATABASE_URL_DIRECT.

Free-Cost Deployment Path

1. Host frontend on Vercel Hobby.
2. Host backend API on Render free tier.
3. Use Supabase free Postgres + Auth.
4. Keep DB_POOL_MAX_SIZE=5 and use cached agent responses (CACHE_TTL_SECONDS=3600) to stay in free limits.
5. Use GitHub Actions schedules for batch ingest and score refresh during off-hours.

Resume Bullets

• Built entity-first CRM intelligence platform with grounded RAG and citation-backed responses.
• Implemented async ingestion and hybrid retrieval pipeline (vector + keyword + ranking).
• Designed multi-workflow agent system (lead finder, account brief, enrichment, research, ops debug).
• Added cache-aware agent orchestration and background crawl processing for cost efficiency.