policy-draft-assistant

Part of the GENIA Americas AI Toolkit — repo #7. Assembles AI governance policy drafts from a fixed clause library.

Honest scope note: this is a deterministic template engine — it assembles pre-written clauses you select, it does not call an LLM to generate novel policy language. If you want actual generative drafting, that would mean integrating a real model call (e.g. the Anthropic API) on top of this, which is not built here.

Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Run

uvicorn app.main:app --reload

Docs at http://127.0.0.1:8000/docs

Test

pytest tests/ -v

11/11 tests pass as of last verified run.

Auth

All endpoints require an X-API-Key header. Keys are configured via the API_KEYS env var as key1:org_one,key2:org_two — each key authenticates its caller as exactly one org, and reads/writes are scoped to that org (a valid key for one org cannot read another org's drafts).

API

• GET /clauses/{policy_type} — list available clause keys/text for a policy type (acceptable_use, data_governance, vendor_risk).
• POST /policies — submit policy type and selected clause keys; org name is derived from your API key, not sent in the request body. Get back the assembled policy text.
• GET /policies/{id} — retrieve one draft belonging to your org.
• GET /health

Extending the clause library

Add entries to app/clause_library.py. Legal/compliance review of any clause before real use is your responsibility — this tool assembles text, it doesn't validate legal accuracy.