Spaces:
Running
Running
metadata
title: Clienttarget
emoji: 🤖
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false
🤖 AI Client Acquisition System
Enterprise-grade, hyper-intelligent lead discovery, profiling, and scoring pipeline.
Built with production AI engineering practices — not n8n-style hype.
What This System Does
Automatically discovers, qualifies, and profiles potential clients for an AI automation agency.
Every day at 9 AM PKT:
1. Pick next territory (city × industry) → 27 cities, auto-rotation
2. Search Google for companies → Serper API
3. Scrape each website → Playwright (headless)
4. Detect pain signals → "no chatbot", "phone booking only", etc.
5. Gate 2: Skip if < 2 pain signals
6. Find decision-maker emails → Hunter.io + Pattern Generation + SMTP
7. Verify emails → 7-layer verification (FREE)
8. Find personal LinkedIn + social profiles
9. AI profiling → MiniMax M2.7 (chain-of-thought reasoning)
10. Deterministic scoring → 100-point scale, zero hallucination
11. Alert on Slack → hot leads (85+) instant, daily digest for all
Architecture
┌─────────────────────────────────────────────────────┐
│ CRON: daily-lead-discovery (4 AM UTC = 9 AM PKT) │
│ → Territory Manager → Google Search → Queue │
└──────────────────────┬──────────────────────────────┘
│
▼ (max 3 concurrent)
┌─────────────────────────────────────────────────────┐
│ TASK: process-company │
│ → Scrape → Pain Signals → Gate 2 │
└──────────────────────┬──────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────┐
│ TASK: enrich-and-profile │
│ → Hunter → Pattern Gen → SMTP → LinkedIn │
│ → Python AI Service → Save → Slack Alert │
└─────────────────────────────────────────────────────┘
Model Chain (All FREE on NVIDIA NIM)
| Priority | Model | Parameters | Use Case |
|---|---|---|---|
| 1st | MiniMax M2.7 | ~100B+ | Profiling, scoring, complex reasoning |
| 2nd | LLaMA 3.3 70B | 70B | Reliable fallback |
| 3rd | LLaMA 3.1 8B | 8B | Email classification, simple tasks |
| 4th | Deterministic | — | Zero hallucination fallback |
Single API key. Single endpoint. $0/day.
Scoring System (100 points, fully deterministic)
Company Fit: 25 pts (industry + size match)
AI Readiness: 20 pts (tech stack + AI jobs)
Service Match: 20 pts (pain signals → our services)
Decision Maker: 20 pts (verified email + LinkedIn + authority)
Timing: 15 pts (growth signals + active website)
Tiers: hot (85+) | warm (70-84) | nurture (50-69) | archive (<50)
Tech Stack
| Layer | Technology | Purpose |
|---|---|---|
| Orchestration | Trigger.dev | CRON, task chaining, retry, queuing |
| Database | Supabase (PostgreSQL) | Data storage, config, state |
| LLM | NVIDIA NIM (MiniMax + LLaMA) | AI profiling & analysis |
| Web Scraping | Playwright | Headless browser |
| Hunter.io + SMTP | Finding & verification | |
| Notifications | Slack Bot | Alerts, commands, digest |
| AI Service | Python FastAPI | Profiling, scoring, hallucination guard |
| Language | TypeScript + Python | Core logic |
Project Structure
src/
├── discovery/ # Phase 1: Finding pipeline
│ ├── lib/ # Core logic
│ │ ├── contact-enricher.ts # 6-step email pipeline
│ │ ├── email-classifier.ts # Tier 1/2/3 classification
│ │ ├── email-verifier.ts # 7-layer verification
│ │ ├── email-pattern-generator.ts # FREE Snov replacement
│ │ ├── linkedin-person-finder.ts # Personal LinkedIn
│ │ ├── social-finder.ts # Instagram, Facebook, Twitter
│ │ ├── pain-signal-detector.ts # Heuristic + LLM
│ │ ├── territory-manager.ts # City×industry grid
│ │ └── web-scraper.ts # Playwright scraper
│ ├── providers/ # External APIs
│ │ ├── hunter.ts # Hunter.io integration
│ │ ├── serper.ts # Google search
│ │ └── reoon.ts # Email verification
│ └── trigger-tasks/ # Trigger.dev tasks
│ ├── auto-discovery.ts # 5 chained tasks
│ └── manual-discovery.ts # Slack-triggered runs
├── profiling/ # AI profiling service
│ └── python-service/ # FastAPI
│ ├── main.py # /profile endpoint
│ ├── profiler.py # Chain-of-thought profiling
│ ├── scorer.py # Signal extraction + deterministic math
│ ├── hallucination_guard.py # Evidence-based cross-check
│ ├── nvidia_client.py # Multi-model LLM client
│ └── config.py # Settings
├── shared/ # Shared utilities
│ ├── config/env.ts # Environment validation (Zod)
│ ├── llm/nvidia-client.ts # Multi-model LLM (MiniMax primary)
│ ├── llm/prompts.ts # Production prompts
│ ├── llm/grounding.ts # Evidence-based verification
│ ├── observability/tracer.ts # Trace IDs + token tracking
│ ├── pipeline/checkpoint.ts # Crash recovery
│ ├── supabase/client.ts # DB client
│ └── utils/ # Retry, rate limiter, logger
└── slack/ # Slack integration
├── slack-service.ts # 3-layer delivery
└── slack-commands.ts # /discover, /leads, /status, etc.
Quick Start
See Setup Guide for detailed instructions.
# 1. Clone
git clone https://github.com/iDevBuddy/ai-client-acquisition.git
cd ai-client-acquisition
# 2. Install
npm install
cd src/profiling/python-service && pip install -r requirements.txt && cd ../../..
# 3. Configure
cp .env.example .env
# Fill in your API keys (see docs/setup-guide.md)
# 4. Database
# Run supabase/migrations/*.sql on your Supabase project
# 5. Run
npm run trigger:dev # Start Trigger.dev (task orchestration)
cd src/profiling/python-service && python main.py # Start AI service
API Keys Required
| Service | Cost | What It Does |
|---|---|---|
| NVIDIA NIM | FREE | AI models (MiniMax + LLaMA) |
| Serper.dev | FREE (2500/mo) | Google search |
| Hunter.io | FREE (25/mo) | Email finding |
| Reoon | FREE (20/day) | Email verification |
| Supabase | FREE | Database |
| Slack | FREE | Notifications |
| Trigger.dev | FREE (50K runs/mo) | Job orchestration |
Total cost: $0/month
Contributing
See CONTRIBUTING.md for guidelines.
License
Private — All rights reserved.