Skip to content

DEPLOYMENT.md

Latest commit

 

History

History
181 lines (126 loc) · 7.21 KB

DEPLOYMENT.md

File metadata and controls

181 lines (126 loc) · 7.21 KB

ComplAI Deployment Guide

Deploy the backend on Railway and the frontend on Vercel. Set up Supabase first so the backend has a database.

Prerequisites

  • GitHub repo with this codebase
  • Supabase project (create one if needed)
  • Google AI API key (for LLM)

0. Supabase setup

Do this before deploying the backend. The backend needs these tables and seed data to run.

0.1 Create a Supabase project

  1. Go to supabase.com and sign in.
  2. New project → choose org, name, database password, region.
  3. Wait for the project to be ready.
  4. In Settings → API: copy Project URL and anon public key. You will use these as SUPABASE_URL and SUPABASE_KEY in Railway and when running the seed script.

0.2 Create tables (run SQL migration)

  1. In the Supabase dashboard, open SQL Editor.
  2. Open the file backend/scripts/create_tables.sql from this repo and copy its full contents.
  3. Paste into the SQL Editor and click Run.
  4. Confirm there are no errors. This creates: profiles, baselines, risk_state, transactions, compliance_state, rulebooks, new_regulations, compliance_drafts, agent_traces, agent_steps.

0.3 Seed the database (run seed script)

The seed script reads from backend/data/ and requires SUPABASE_URL and SUPABASE_KEY. Run it once from your machine (or any environment that has the repo and env vars).

  1. From the repo root:

    cd backend

  2. Create a .env in backend/ with:

    SUPABASE_URL=https://your-project.supabase.co
SUPABASE_KEY=your-anon-key

    Use the same URL and anon key from step 0.1.

  3. Install dependencies and run the seed script:

    pip install -r requirements.txt
python scripts/seed_supabase.py

    Or with a virtualenv:

    python -m venv .venv
.venv\Scripts\activate   # Windows
# source .venv/bin/activate   # macOS/Linux
pip install -r requirements.txt
python scripts/seed_supabase.py

  4. The script expects these files under backend/data/ (they are in the repo):

    • users.json — user profiles
    • baselines.json — user baselines
    • initial_risk_state.json — optional; initial risk state
    • historical_transactions.json — optional; historical transactions
    • compliance/malta.json, compliance/uae.json, compliance/cayman.json — compliance state and rulebooks
    • compliance/new_regulations/malta_v2.json, uae_v2.json, cayman_v2.json — new regulations to push

    If a file is missing, the script skips that part and continues.

  5. When it finishes, you should see messages like “Seeded N profiles”, “Seeded compliance state…”, etc. The app will then work with real data once the backend is deployed.

Summary: Run backend/scripts/create_tables.sql in the Supabase SQL Editor once, then run backend/scripts/seed_supabase.py once from your machine (with backend/.env set). After that, you only need to deploy Railway and Vercel; no need to run these again unless you reset the database.

0.4 Reset database to initial state

To undo experiments (data injection, anomaly detection, regulatory pushes) and restore everything to the initial seed state:

From the repo root:

cd backend
python scripts/reset_supabase_regulatory.py

Requires SUPABASE_URL and SUPABASE_KEY in backend/.env. The script:

  1. Clears agent_steps and agent_traces
  2. Clears all transactions
  3. Clears compliance_drafts and rulebooks
  4. Re-seeds profiles, baselines, risk_state (all users set to CLEAN/0), historical_transactions, compliance_state and v1 rulebooks from backend/data/
  5. Sets all new_regulations.is_pushed to false

After the reset, users, baselines, risk scores, transactions, and regulatory state match the initial seed.

1. Railway (Backend)

  1. Go to railway.app and sign in with GitHub.

  2. New Project → Deploy from GitHub repo → select your ComplAI repo.

  3. Set root directory
    In the service: Settings → Source → Root Directory → set to backend.

  4. Environment variables
    In the service: Variables → add:

    Variable Value
    LLM_API_KEY Your Google AI API key
    LLM_MODEL gemini-2.0-flash
    LLM_BASE_URL https://api.openai.com/v1
    SUPABASE_URL Your Supabase project URL
    SUPABASE_KEY Your Supabase anon key
    FRONTEND_URL (Set after Vercel deploy)

  5. Deploy
    Railway uses Nixpacks, installs from requirements.txt, and runs the Procfile. PORT is set automatically.

  6. Public URL
    Settings → Networking → Public Networking → Generate Domain.
    Copy the URL (e.g. complai-backend-production.up.railway.app). You will use it for the frontend and for FRONTEND_URL after Vercel is live.

2. Vercel (Frontend)

  1. Go to vercel.com and sign in with GitHub.

  2. Add New → Project → import your ComplAI repo.

  3. Root Directory
    Set to frontend (or leave default and configure build to use frontend).

  4. Environment variables
    Add:

    Variable Value
    NEXT_PUBLIC_API_URL https://<your-railway-domain> (no slash)

    Example: https://complai-backend-production.up.railway.app

  5. Deploy
    Vercel runs next build and deploys. Copy your Vercel URL (e.g. https://complai.vercel.app).

  6. CORS
    In Railway, set Variables → FRONTEND_URL to your Vercel URL (e.g. https://complai.vercel.app).
    Redeploy the backend or wait for the next restart so CORS allows the frontend origin.

3. Recommended order

  1. Supabase: Create project → run backend/scripts/create_tables.sql in SQL Editor → run backend/scripts/seed_supabase.py locally with backend/.env set.
  2. Deploy backend on Railway (use same SUPABASE_URL and SUPABASE_KEY).
  3. Copy Railway public URL.
  4. Deploy frontend on Vercel with NEXT_PUBLIC_API_URL = Railway URL.
  5. Copy Vercel URL.
  6. In Railway, set FRONTEND_URL to Vercel URL and redeploy (or let it pick up on next deploy).

4. Post-deploy checks

  • Backend: Open https://<railway-domain>/api/health — should return JSON with "status": "ok".
  • Frontend: Open your Vercel URL, go to Regulatory Hub, and confirm it loads compliance data (no CORS errors in the browser console).
  • Data: If Regulatory Hub or Monitor show no data, ensure Supabase setup (section 0) was completed: tables created and seed_supabase.py run successfully.

5. Local development

  • Backend: From repo root, cd backend then uvicorn main:app --reload --port 8000.
    Use a .env in backend/ (see backend/.env.example).

  • Frontend: From repo root, cd frontend then npm run dev.
    Set NEXT_PUBLIC_API_URL=http://localhost:8000 in frontend/.env.local (or leave unset to default to localhost).