S E N T I N E L

Autonomous Competitive Intelligence Platform
Stop Googling. Start Strategizing.

---

What Sentinel does

Sentinel is a multi-agent orchestrator that replaces hours of manual competitive research with minutes of autonomous analysis. Give it a target company (or a strategic question) and it will:

1. Classify your intent and generate high-signal search queries.
2. Hunt live web data through Google Search grounding.
3. Scrape + clean noise into a dense fact sheet.
4. Run a Pro-model SWOT with scored dimensions (Innovation, Market Share, Pricing Power, Brand, Velocity - each 0-100).
5. Compile a C-level markdown report.
6. Optionally draft a viral LinkedIn post.
7. Let you chat with the report (RAG over the fact sheet + live web).

Plus Battle Mode: pick any two saved reports to see a head-to-head radar chart + tale-of-the-tape comparison.

Live

sentinel.adityaai.dev

Stack

layer	tech
frontend	React 19, TypeScript 5.8 strict, Vite 6
styling	Tailwind CSS (Dark Sentinel theme), custom SVG radar charts
AI	Gemini Pro Latest (analyst + chat), Gemini Flash Latest (router/hunter/scraper/reporter/social), Google Search grounding
backend	Vercel Node Functions
data	Supabase Postgres with row-level security
realtime	React-Markdown + GFM for report rendering

Agent chain

Router  - Flash  - classify intent, extract target, draft search queries
Hunter  - Flash  - Gemini + Google Search, pull high-authority sources
Scraper - Flash  - consolidate into a dense fact sheet (<4k tokens)
Analyst - Pro    - SWOT + 5-dimension scoring (0-100 each)
Reporter- Flash  - C-level markdown report
Social  - Flash  - optional LinkedIn post
Chat    - Pro    - RAG over the saved report

Security

• API key server-side only. No GEMINI_API_KEY in the client bundle.
• Per-IP rate limits: 20 analyze/hour, 200 chat/hour, 30 per-agent/hour.
• Supabase RLS on every table.
• IP hashing (SHA-256 + salt) before any persistence.
• CORS allowlist, HSTS, X-Frame-Options DENY, Permissions-Policy.

See SECURITY.md.

Local dev

nvm use                      # node 22 from .nvmrc
npm install
cp .env.example .env.local   # fill in keys
npm run dev                  # http://localhost:3000

For full vercel dev + edge functions, use vercel dev instead.

Deploy

One command:

vercel --prod

Full first-time setup: docs/DEPLOYMENT.md.

API

Every agent is an HTTP endpoint. See docs/API.md.

Architecture

See docs/ARCHITECTURE.md.

Browser requirements

• Chromium 120+, Firefox 125+, Safari 17+
• Works offline for already-loaded reports (service worker shell to come)

Contributing

See docs/CONTRIBUTING.md.

License

Apache 2.0 - see LICENSE.

---

Built by Aditya. Powered by Gemini + Supabase + Vercel.