Prospex is a self-hosted, open-source platform that combines lead discovery, AI-powered outreach generation, and CRM management in one dashboard.

Think Apollo.io + Instantly.ai — but open-source, self-hosted, and free.

• Node.js 20+
• pnpm 9+
• Docker

git clone https://github.com/asiifdev/business-leads-ai-automation.git
cd business-leads-ai-automation
pnpm install

cp .env.example apps/api/.env
cp .env.example packages/database/.env

Edit apps/api/.env with your values:

DATABASE_URL="postgresql://prospex:yourpassword@localhost:5432/prospex"
OPENAI_API_KEY=sk-your-key-here   # or leave empty for mock AI
OPENAI_MODEL=gpt-4o-mini
# Optional: use OpenRouter or Ollama
# OPENAI_BASE_URL=https://openrouter.ai/api/v1

docker compose up -d         # starts PostgreSQL + Redis
pnpm --filter @prospex/database db:push   # push schema
pnpm --filter @prospex/database exec prisma generate

# Terminal 1 — API (port 3001)
pnpm --filter @prospex/api dev

# Terminal 2 — Dashboard (port 3000)
pnpm --filter @prospex/web dev

Open http://localhost:3000 and you're in.

Swagger API docs: http://localhost:3001/api/docs

prospex/
├── apps/
│   ├── web/              # Next.js 16 dashboard
│   ├── api/              # NestJS REST API
│   └── marketing/        # Landing page
├── packages/
│   ├── database/         # Prisma schema + migrations
│   ├── types/            # Shared TypeScript types
│   └── config/           # Shared eslint/tsconfig
├── docker-compose.yml        # Local dev (PostgreSQL + Redis)
├── docker-compose.prod.yml   # Production stack
└── nginx.conf                # Nginx reverse proxy config

Full interactive docs at /api/docs (Swagger).

Prospex supports any OpenAI-compatible AI provider:

OpenRouter (access Claude, Gemini, Llama, etc.):

OPENAI_API_KEY=sk-or-your-openrouter-key
OPENAI_BASE_URL=https://openrouter.ai/api/v1
OPENAI_MODEL=anthropic/claude-haiku-4-5

Ollama (local, fully offline):

OPENAI_API_KEY=ollama
OPENAI_BASE_URL=http://localhost:11434/v1
OPENAI_MODEL=llama3.2

cp .env.example .env
# Edit .env with production values (strong passwords, real API key)

docker compose -f docker-compose.prod.yml up -d

This starts: PostgreSQL + Redis + NestJS API + Next.js + Nginx.

pnpm install            # install all dependencies
pnpm dev                # start all apps (requires turbo)
pnpm build              # build all apps
pnpm --filter @prospex/database db:studio   # Prisma Studio (DB browser)

• Fix: search queries now combine each campaign's search term with every area in its location field (split on commas) before scraping — previously location was stored but never sent to the scraper, so results could drift to the scraping server's IP-based region instead of the requested cities.
• Feature: leads now capture lat/lng coordinates (parsed from the scraped Google Maps place URL) and store them on the Lead model.
• Feature: campaign detail page has a new "Map" tab with a Leaflet-based lead map (custom priority-colored markers, auto-fit bounds, popups linking to lead detail) alongside the existing "List" view.
• Fix: scraping progress no longer gets stuck at 0% — the scraper now reports incremental progress per search query/area combo while scraping runs (previously progress only updated before scraping started and after it fully finished, so long-running scrapes appeared frozen).
• Fix: campaign and lead list polling (every 2s while a campaign is running) no longer flashes the whole page back to a loading skeleton on every refresh — only the initial fetch shows the skeleton, so live progress/lead updates render smoothly.
• Fix: lead scoring/ordering no longer ranks by raw star rating alone (a 5.0★ lead with 1 review could previously outrank a 4.5★ lead with thousands of reviews). Review count is now scraped and stored, and rating contribution to the score uses a Bayesian average (weighted toward a neutral 4.0★ prior by review volume) so low-review, high-star leads no longer unfairly beat well-reviewed ones.

MIT — see LICENSE.