Skip to content

local-dev.md

Latest commit

 

History

History
91 lines (69 loc) · 3.27 KB

local-dev.md

File metadata and controls

91 lines (69 loc) · 3.27 KB
Raw

Local Development

The recommended way to run HASTE locally is the Docker Compose stack, which brings up the entire platform — storage emulator, tile server, API, queue workers, and UI — with a single command.

This page is a **summary**. The complete, step-by-step instructions live in two files at
the repository root, which are the source of truth:

- **[`QUICKSTART.md`](https://github.com/microsoft/haste/blob/main/QUICKSTART.md)** — a
  phased, verify-gated runbook (also used to drive AI coding agents).
- **[`docker/README.md`](https://github.com/microsoft/haste/blob/main/docker/README.md)** —
  the comprehensive human-facing prose guide, including prerequisites, memory tuning, and
  troubleshooting.

When any doc disagrees with the compose file, `docker/docker-compose.yml` is ground truth.

What you're standing up

The local stack runs entirely from docker/docker-compose.yml:

Layer Services Notes
Storage emulator azurite, data-init Azure Blob/Queue/Table emulator; data-init seeds it once and exits
Edge api-proxy (nginx), titiler CORS proxy on :7071; COG tile server
API hastefuncapi, hastefuncqueues Azure Functions (Python 3.11); the queue worker spawns GPU jobs
UI ui Vite + React via SWA CLI on :4280
Build-only training_image, imageryprep_image Spawned on demand by the LocalRunner via the Docker socket

Prerequisites

  • Docker with Compose v2 (docker compose version)
  • ~100 GB free disk for the full image set
**Apple Silicon:** the images are amd64-only, so you must export
`DOCKER_DEFAULT_PLATFORM=linux/amd64` before every `build`/`up` command. Docker Desktop
then runs them under emulation — functional, just somewhat slower.

**GPU** acceleration is Linux-only. On macOS or Windows the stack runs CPU-only, so
training and inference jobs are slow or impractical — the rest of the app (UI, project and
imagery management, tiling) is fully usable for evaluation.

Bring it up

From the repository root:

# 1. Configure host + docker socket GID (see QUICKSTART §2 for the full recipe)
cat > docker/.env <<'EOF'
HOST_IP=localhost
DOCKER_GID=<gid-of-your-docker.sock>
HASTE_ENABLE_GPU=0
EOF

# 2. Build the images (first build is ~15–30 min)
docker compose -f docker/docker-compose.yml build

# 3. Start the stack
docker compose -f docker/docker-compose.yml up -d

Once running:

  • UI — http://<HOST_IP>:4280
  • API — http://<HOST_IP>:7071/api/
  • Azurite blob — http://<HOST_IP>:10000
For Docker Compose, the SWA mock-login form pre-fills **User's roles** with
`administrators`. Enter a user ID and username, confirm the role is present,
then select **Login**.

For platform-specific profiles (Apple Silicon, GPU vs. CPU), the health-check gates, the
critical traps (nginx upstream caching, `data-init` re-seeding), and full troubleshooting,
follow [`QUICKSTART.md`](https://github.com/microsoft/haste/blob/main/QUICKSTART.md)
end to end.

Ready for production?

The local stack is for development and evaluation only. To deploy a production-grade environment on Azure, see the Deployment Guide and the Configuration Guide.