Skip to content

Infrastructure.md

Latest commit

 

History

History
37 lines (22 loc) · 2.3 KB

Infrastructure.md

File metadata and controls

37 lines (22 loc) · 2.3 KB
Raw

Infrastructure

This document describes how the app is run locally, packaged, and how CI/CD is set up. Secrets and exact Azure names live in your environment or GitHub; see .env.example for variable names.

Local runtime

  • API — npm run start:dev (NestJS, default port from PORT, usually 3000).
  • MongoDB — docker compose up -d starts mongo:7 on port 27017 with a named volume for data. Set MONGODB_URI=mongodb://localhost:27017/ai-crate-digger in .env.
  • Seed — npm run seed:discogs needs DISCOGS_USER_TOKEN and uses DISCOGS_SEED_TARGET / DISCOGS_SKIP_EMBEDDINGS for cost control.
  • Optional UI — npm run dev:client runs the Vite dev server pointed at the API.

Container image

Dockerfile (multi-stage):

  1. Builder — npm ci, copy sources, npm run build (Nest dist/).
  2. Runner — production npm ci --omit=dev, copy dist/, CMD ["node", "dist/main.js"], expose 3000.

Build and run are standard Docker workflows; compose in this repo is Mongo-only, not the API (run the API on the host or add a service if you want an all-in-one stack).

CI/CD

.github/workflows/ci-cd.yml runs on pushes and PRs to main:

  • Checkout, Node 22, npm ci, lint, build, test.

A docker-deploy job is commented out but documented in-repo: it would log into Azure, build/push to Azure Container Registry, and update an Azure Container Apps app image. Turning it on requires GitHub secrets such as AZURE_CREDENTIALS, ACR_NAME, ACR_LOGIN_SERVER, ACA_APP_NAME, and ACA_RESOURCE_GROUP, plus the same runtime secrets you use locally (Mongo URI, Azure OpenAI keys, etc.) configured on the container app.

External services

  • Azure OpenAI — Chat, embeddings, and optional TTS; endpoints and deployments are configured via AZURE_OPENAI_* variables in .env.
  • Discogs API — Used only for seeding, via DISCOGS_USER_TOKEN.

Operations notes

  • After rewriting git history or rotating secrets, force-push branches and re-clone collaborators as needed; GitHub Actions and Container Apps should pick up new images from the pipeline once deploy is enabled.
  • MongoDB backups for production are not defined in this repo; use Atlas, managed disk snapshots, or your platform’s backup story if you move beyond local Docker.