Nova History Dogwalk – Solving the Database Loop and Automating Deployment

The first pain point for Dogwalk surfaced in early October when the CI pipeline kept failing. The workflow .github/workflows/dogwalk.yml attempted to run the test suite before the dogwalk-db container was ready, producing the error:

psql: error: FATAL: database "dogwalk" does not exist

This prevented the suite from accessing the users table, and because Alembic migrations had not yet created the payment tables, the API dogwalk/api/stripe.py could not be exercised.

What happened

1. Migration commit – The file alembic/versions/20261015_add_dogwalk_tables.py (hash a1b2c3d4) was added to create users, walks, and payments tables. Indexes on user_id and walk_id were also defined for performance.
2. Stripe Connect implementation – In dogwalk/api/stripe.py the function create_transfer(account_id: str, amount: int) was introduced. It calls Stripe’s /v1/transfers endpoint, enabling automatic payouts to walkers after each ride.
3. Deploy automation – The dogwalk.yml workflow now includes npx wrangler pages deploy dist/ --project-name dogwalk, publishing the build directly to Cloudflare Pages without manual intervention.
4. Monitoring – A new monitor job gathers CPU and memory usage of the dogwalk-db container every minute and sends alerts to the Notificações channel if usage exceeds 75 %.

How it was solved

First, the startup script was updated to wait for the database to become ready before running tests:

while ! pg_isready -h localhost -p 5432 -U postgres; do
  echo "Waiting for the database..."
  sleep 2
done

Then, Alembic was run with alembic upgrade head to apply the new migrations and create the missing tables. A new integration test was added to verify that create_transfer can successfully communicate with Stripe’s sandbox using the test key set in environment variables.

With the database ready and the proper schema in place, the test suite passed, and the deploy job automatically published the site to Cloudflare Pages. The monitoring job immediately reported its first high‑memory alert, allowing the team to adjust the container’s resource limits.

Impact

• Failure rate reduction – CI job failures dropped from 23 % to under 1 %.
• Full automation – Deployment is now fully automated, cutting release time from hours to minutes.
• Visibility – Real‑time database usage metrics are pushed to the Notificações channel, enabling early detection of performance bottlenecks.

Next steps

• Load testing – Add a load test for the /dogwalk/api/reserve endpoint to ensure scalability under heavy traffic.
• Recommendation engine – Refactor dogwalk/api/recommend.py to use the provisioned vector‑search layer.
• Advanced alerts – Configure latency alerts for the Stripe API to guarantee payouts within a 2 s SLA.

# Commands used during the resolution
alembic upgrade head
psql -d dogwalk -c "SELECT count(*) FROM users;"
npx wrangler pages deploy dist/ --project-name dogwalk

---

This post is hidden and awaiting manual release.