Veridex

An AI-powered strategy red-team. Paste a strategy document and four independent critics return a structured critique in parallel.

Live at veridex.fyi.

---

The four lenses

• Pre-Mortem — It's 18 months from now and the strategy failed. Why?
• Unit Economics — Does the math work at every stage?
• Adversarial Competitor — A well-funded rival wants to kill this. How?
• Execution Risk — Assume the strategy is directionally right. What breaks in shipping?

Each lens returns a short synthesis and 2–7 ranked flaws. Each flaw has a title, severity (low/medium/high/critical), a concrete description, and the sharpest question the author must answer next.

Output is streamed over Server-Sent Events so the four passes surface as soon as each one completes.

---

Architecture

Frontend — Next.js on Vercel. Single page. Textarea, live progress panel, structured results view, copy-to-clipboard, download-as-Markdown.

Backend — FastAPI on Render. One endpoint: POST /api/v1/critique/stream. Runs all four critic passes in parallel via asyncio.gather and streams pass_started / pass_completed / error / done events.

Model — OpenAI gpt-5-mini via a provider-agnostic wrapper (run_critic_pass(lens, input_text, model="openai:gpt-5-mini")).

Validation — Every pass is validated against a strict Pydantic schema. One retry on schema failure with a corrective nudge, then CriticValidationError — no silent fallback.

---

Project structure

veridex
├── backend
│   └── app
│       ├── api/v1/routes_critique.py    # SSE endpoint + rate limiter
│       ├── services/critics.py           # Lens prompts, schema, runner
│       ├── llm/client.py                 # OpenAI client
│       └── main.py                       # FastAPI app, CORS
├── frontend
│   └── app/page.tsx                      # Single-page UI
└── README.md

---

Running locally

Backend

cd backend
py -m venv .venv
.venv\Scripts\Activate
pip install -r requirements.txt
uvicorn app.main:app --reload

Backend runs on http://127.0.0.1:8000.

Frontend

cd frontend
npm install
npm run dev

Frontend runs on http://localhost:3000.

---

Environment variables

Backend (backend/.env)

OPENAI_API_KEY=sk-...
CORS_ALLOWED_ORIGINS=                   # optional, comma-separated, for preview deploys

Frontend (frontend/.env.local)

NEXT_PUBLIC_API_URL=http://127.0.0.1:8000

In production (Vercel), set NEXT_PUBLIC_API_URL to the live Render backend URL (e.g. https://veridex-backend.onrender.com). No trailing slash.

---

Rate limiting

5 critiques per IP per hour, in-memory on the backend. Resets on process restart. Good enough for v1 alpha; will be replaced with a persistent store before paid tier.

---

Non-goals (v1)

• No accounts, no signup, no user data at rest.
• No multi-turn conversation or "redirect mid-critique" — red-teams don't get steered.
• No "improvement suggestions" — Veridex challenges, it does not generate.
• No PDF export (deferred to v1.1).

---

Author

Mahesh Reddy Pagadala