Lead Triage API

A tiny API that classifies an inbound contact-form message with Claude before it lands in an inbox - category (client lead / collaboration / networking / general / spam), urgency, and a one-line summary.

Built to power the contact form on my portfolio site: the form's JS calls this API right before submitting to Formspree, and includes the tags in the email. If this service is ever slow, down, or rate-limited, the form still submits normally - triage is a non-blocking enhancement layered on top, never a hard dependency for the contact form to work.

Why this exists

Most of my other project (doc-qa-tool) is retrieval: search documents, answer from what's found. This one is a different core technique - structured extraction: turn free text into a fixed set of tags a program can act on. That pattern (classify, score, extract fields) shows up constantly in real consulting work - lead scoring, support-ticket routing, sentiment tagging - more often, in practice, than document search does.

How it works

1. POST /api/triage with {"name": "...", "message": "..."}
2. Claude (Haiku 4.5 - cheap, fast, exactly suited to short classification) reads the message and returns JSON: category, urgency, summary
3. The response is parsed defensively (tolerates the model wrapping the JSON in a sentence) and validated against a fixed set of allowed values before being trusted

Guardrails (public endpoint)

• CORS-restricted - only requests from an origin in ALLOWED_ORIGIN (comma-separated; the portfolio site's domain(s)) are permitted from a browser
• Rate limited - 20 requests/hour per caller (flask-limiter)
• Input validated - oversized messages rejected before reaching Claude
• Cheap model - Haiku 4.5, chosen for classification specifically

Local setup

python -m venv venv
.\venv\Scripts\Activate.ps1
pip install -r requirements.txt
copy .env.example .env
# edit .env, paste in your Anthropic API key
python app.py

Deploy to Render (free tier)

1. Push this repo to GitHub
2. New Web Service on render.com, pointed at this repo
3. Build command: pip install -r requirements.txt
4. Start command: gunicorn app:app (the Procfile sets this too)
5. Environment variables: ANTHROPIC_API_KEY (required), ALLOWED_ORIGIN (optional - defaults to the portfolio's real domain already)

Set a monthly spend cap in the Anthropic Console regardless - same reasoning as doc-qa-tool: this is a public endpoint, and rate limiting bounds normal traffic but a cap is the real safety net.