AI-assisted integration

1. Add Verifai to your AI tool

Cursor: Settings → Indexing & Docs → add your Verifai deployment URL or paste llms.txt / OpenAPI links. Then reference @Docs while coding.

2. Cursor rule (download)

Copy this file into your project at .cursor/rules/verifai-integration.mdc so the agent follows Verifai conventions.

After adding the rule, enable it for your integration task or set alwaysApply: true while building face auth.

3. Copy-paste integration prompt

Paste this into Cursor, Claude, or Codex when starting an integration task in your app repo.

Integrate the Verifai facial recognition API into this application.

Documentation:
- Quickstart: https://verifai.iqtbots.ai/docs/
- OpenAPI schema: https://verifai.iqtbots.ai/api/schema/
- llms.txt: https://verifai.iqtbots.ai/llms.txt

Requirements:
1. Store VERIFAI_API_KEY and VERIFAI_BASE_URL in environment variables (server-side only).
2. Use Authorization: Bearer <api_key> on all /api/v1/ recognition calls.
3. Flow: create collection (once) → enroll user with external_id = our user ID → **verify** on login/check-in (preferred). Use identify only for small unknown-probe watchlists.
4. Upload images as multipart/form-data with field name "image" (JPEG/PNG).
5. Default match threshold: 0.4; check data.matched and data.similarity on verify.
6. Handle error.code: no_face_detected, person_not_found, not_enrolled, gallery_too_large, rate_limit_exceeded, daily_quota_exceeded.
7. Never call Verifai from frontend JavaScript with the API key — proxy through our backend.

Implement a service class and wire it into our existing auth/user flow.

4. Claude / Codex project instructions

Add a short section to your repo's CLAUDE.md or AGENTS.md:

## Verifai facial recognition
- Docs: https://verifai.iqtbots.ai/docs/
- OpenAPI: https://verifai.iqtbots.ai/api/schema/
- llms.txt: https://verifai.iqtbots.ai/llms.txt
- API key in env only; backend proxy for all /api/v1/ calls
- Flow: enroll with external_id = our user id → verify on login

5. Security reminder

• Never put vf_test_ or vf_live_ keys in frontend code.
• Your app receives photos → your server calls Verifai → your server returns pass/fail.
• Use test keys (vf_test_) until you are ready for production.