📡 RepoAlpha — Open Source M&A Intelligence

Bright Data AI Agents Hackathon 2026 · Bright Data · AI/ML API · Speechmatics · LangChain · Supabase · Streamlit

RepoAlpha turns VC and M&A teams into "God Mode" investors by detecting corporate developer adoption before it becomes public news. A "Software Engineer at Nvidia" starring your repo is a 15-point signal. An anonymous account is noise.

---

🤝 Partner Integrations

Partner	Usage
Bright Data	Web Unlocker scrapes GitHub stargazer profiles & trending pages
AI/ML API	llama-3.3-70B analyzes READMEs, generates hype scores & commercial summaries
Speechmatics	Real-time TTS voice alerts narrate BUY signals aloud
LangChain	Orchestrates multi-step agent pipelines
Supabase	PostgreSQL + pgvector for storage, embeddings, and audit trails
Streamlit	Bloomberg Terminal-style dashboard

---

🧠 How It Works (Quick View)

1. GitHub trending repos harvested every hour
2. Every stargazer's profile scraped via Bright Data
3. Company extracted → scored (Nvidia=15pts, Google=12pts, …)
4. AI/ML API (Llama 3.3 70B) generates hype score + commercial summary
5. BUY / HOLD / SELL rating assigned by aggregate corporate score
6. Slack · Discord · Email · Voice alerts fired on threshold breach

---

🆕 Enterprise-Grade Features

Feature	Detail	Cost
Pydantic v2 models	All pipeline data validated before Supabase writes	Free
Circuit breaker	Auto-opens on sustained API failures, heals after 2 min	Free
Exponential retry	tenacity wraps every external call — no silent failures	Free
Structured logging	loguru JSON lines + Supabase pipeline_runs audit trail	Free
TTL in-process cache	Spare Supabase reads within a pipeline run	Free
Watchlist	Pin repos, personal tracking across sessions	Free
Score history	Auto-snapshotted by Supabase trigger → sparkline charts	Free
pgvector semantic search	all-MiniLM-L6-v2 embeddings, ANN search via match_repos()	Free
Multi-channel alerter	Slack + Discord + Resend email + Speechmatics TTS voice	Free
FastAPI REST layer	Full CRUD + CSV/JSON export, Swagger at /docs	Free
GitHub Actions CI/CD	Hourly pipeline + lint + type-check, no paid scheduler	Free
Render.com deploy	API on free web service + cron worker	Free
pytest suite	20+ unit tests covering models, scoring, cache, license	Free

---

⚡ 30-Minute Setup

1 — Clone & install

git clone https://github.com/shanujans/repoalpha.git
cd repoalpha
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt

2 — API keys

Service	Where	Notes
Bright Data	brightdata.com → Account Settings	API Token + Web Unlocker zone
AI/ML API	aimlapi.com → API Keys	Free tier (Llama 3.1 70B)
Speechmatics	speechmatics.com → API Keys	TTS free tier (4 hrs/month)
Supabase	supabase.com → Settings → API	URL + service_role key
GitHub PAT	github.com/settings/tokens	No scopes — raises limit 60→5000 req/hr
Slack webhook	Your workspace → Apps → Incoming Webhooks	Optional alerts
Discord webhook	Server → Integrations → Webhooks	Optional alerts
Resend	resend.com → API Keys	100 emails/day free

cp .env.example .env
nano .env   # paste all keys

3 — Database (run once in Supabase SQL Editor)

# Paste schema.sql into: https://supabase.com/dashboard → SQL Editor → New Query → Run
# This creates all tables, the pgvector extension, the snapshot trigger,
# the match_repos() RPC, and Row Level Security policies.

4 — See the dashboard immediately (no API credits needed)

python main.py --demo          # seeds 5 repos + 18 signals + score history
streamlit run dashboard.py     # → http://localhost:8501

5 — Run the live pipeline

python main.py                 # one full cycle (all 5 phases)
python main.py --loop          # hourly daemon
python main.py --phase 1       # single phase: 1=harvest 2=enrich 3=analyse 4=alert 5=vector

6 — Start the REST API

uvicorn api.main:app --reload --port 8000
# Swagger UI → http://localhost:8000/docs

7 — Run tests

pytest tests/ -v

---

🌐 Free Deployment Stack

Streamlit Community Cloud  →  dashboard.py        (public URL, unlimited)
Render.com free tier       →  api/main.py          (auto-sleep, wakes on request)
Render.com cron (free)     →  main.py             (hourly pipeline run)
GitHub Actions (free)      →  lint + type-check   (on every push to main)
Supabase free tier         →  PostgreSQL + pgvector (500 MB, unlimited API)

Streamlit Community Cloud:

1. Push repo to GitHub (.env is gitignored ✅)
2. share.streamlit.io → New App → select dashboard.py
3. App Settings → Secrets:

SUPABASE_URL = "https://your-project.supabase.co"
SUPABASE_KEY = "your_anon_key"   # anon key is read-only — safe to expose

Render.com:

1. New → Blueprint → connect GitHub repo → it reads render.yaml automatically
2. Set env vars in Render Dashboard → Environment

GitHub Actions:

• Add all .env values as GitHub repository secrets
• Pipeline runs automatically every hour at :05 past
• Manually trigger any phase from Actions → workflow_dispatch

---

📊 Corporate Scoring Reference

Tier	Score	Companies
🏆 AI Leaders	+15	Nvidia · OpenAI · Anthropic · DeepMind · Mistral
🥇 Hyperscalers	+12	Google · Meta · Microsoft · Apple · Amazon · Tesla
🥈 Notable Tech	+10	Netflix · Stripe · Databricks · Snowflake · Hugging Face
🥉 Enterprise	+6	Salesforce · IBM · Intel · Oracle · AMD
🏅 Fortune 500	+3	JPMorgan · Goldman · Boeing · Samsung · Sony
👤 Generic signal	+1	Any bio mentioning "Engineer/Developer/Researcher"

Ratings: BUY ≥ 60 pts · HOLD 25–59 · SELL < 25

---

🗂️ Project Structure

repoalpha/
├── dashboard.py              # Streamlit War Room UI (596 lines)
├── main.py                   # Enterprise orchestrator — all 5 phases (435 lines)
├── schema.sql             # Full Supabase schema + pgvector + triggers + RLS
├── render.yaml               # Render.com free deployment blueprint
├── requirements.txt          # All dependencies
├── .env.example              # Full env variable template (incl. Speechmatics)
├── .gitignore
│
├── agents/
│   ├── harvester.py          # Phase 1: Bright Data GitHub Trending scan
│   ├── enricher.py           # Phase 2: Web Unlocker corporate signal detection
│   ├── analyst.py            # Phase 3: AI/ML API (Llama 3.3 70B) hype + license + dossier
│   └── alerter.py            # Phase 4: Slack / Discord / Email / Speechmatics voice alerts
│
├── api/
│   └── main.py               # FastAPI REST API (Swagger at /docs)
│
├── utils/
│   ├── models.py             # Pydantic v2 data models — full type safety
│   ├── logger.py             # loguru structured logging + Supabase audit trail
│   ├── retry.py              # tenacity retry + circuit breaker + rate limiters
│   ├── cache.py              # TTL in-process cache (no Redis needed)
│   └── vector.py             # pgvector semantic search embeddings
│
├── tests/
│   ├── test_models.py        # Pydantic model validation tests
│   ├── test_enricher.py      # Corporate scoring engine unit tests
│   ├── test_analyst.py       # License classification tests
│   └── test_cache.py         # TTL cache tests
│
└── .github/
    └── workflows/
        └── pipeline.yml      # GitHub Actions: hourly pipeline + CI lint/typecheck

.streamlit/
├── config.toml               # Dark terminal theme (#0E1117 · #00FFAA)
└── secrets.toml.example      # Streamlit Cloud secrets template

---

💰 Zero-Cost Budget Tracker

Service	Free Limit	RepoAlpha Usage	Status
Bright Data	$250 credit	~$30–50/hackathon	✅ 5× headroom
AI/ML API	Generous free tier	~40 req/cycle (70B) + ~200 req/cycle (8B)	✅ Ample
Speechmatics	4 hrs TTS/month	Voice alerts ~ few sec per alert	✅ Free
Supabase DB	500 MB	~10 MB for 50 repos + history	✅ 50× headroom
Supabase API	Unlimited	Unlimited	✅ No limit
Streamlit Cloud	Unlimited public	1 app	✅ Free
Render API	750 hrs/month	~1 instance	✅ Free
Render Cron	750 hrs/month	24 runs/day	✅ Free
GitHub Actions	2,000 min/month (public)	~5 min/run × 24 = 120 min/day	✅ Free
Resend	100 emails/day	1–5 alerts/day	✅ Free
sentence-transformers	Open source, CPU-only	Local embed, no API	✅ Free

Total infrastructure cost: $0.00/month

---