H A S T E

High-speed Assessment and Satellite Tracking for Emergencies

HASTE is an AI-driven framework for rapid disaster assessment using satellite and remote sensing data. It automates geospatial analysis with machine learning to produce accurate disaster maps, and provides a user-friendly web interface so that non-technical users can generate critical insights alongside disaster experts.

Quick Start

The fastest way to run HASTE locally is with Docker Compose, which starts the full stack — API, queue workers, tile server, storage emulator, and UI — with no Azure subscription required.

Prefer to let an AI agent do it? QUICKSTART.md is a phased, verify-gated runbook written for coding agents. Point Claude Code or GitHub Copilot at it with a prompt like "Using QUICKSTART.md, stand up and start a local instance of HASTE" and it will run the steps, handle platform quirks (e.g. Apple Silicon), and stop at each health-check gate.

Prerequisites: Docker and Docker Compose

git clone https://github.com/microsoft/haste.git
cd haste
docker compose -f docker/docker-compose.yml up

Service	URL
UI	http://localhost:4280
REST API	http://localhost:7071/api/
TiTiler tile server	http://localhost:8000
Azurite storage emulator	http://localhost:10000

Note: This stack is for local development and evaluation only — it uses development defaults (in-memory storage emulator, disabled auth, wildcard CORS) that aren't suitable for production. To stand up a production environment on Azure, follow the Deployment Guide and Configuration Guide.

See it in action

Documentation

Full documentation is published at microsoft.github.io/haste:

Using HASTE

• User Guide — create projects, add imagery, label, train, and assess damage end to end

Setup & deployment

• Local Development — run the full stack locally with Docker Compose
• Deployment Guide — deploy to Azure with a single azd up
• Configuration Guide — all deployment settings
• Secure Configuration Guidance — production hardening checklist

Reference

• Architecture — how the components fit together
• API Reference — REST, queue, and tile-server APIs

Source for the docs lives in docs/ and is built with Jupyter Book.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         React UI (Vite)                          │
│  Projects · Labeling Tool · Visualizer · Admin · Model Catalog   │
└───────────────────────────┬─────────────────────────────────────┘
                            │ HTTP
┌───────────────────────────▼─────────────────────────────────────┐
│               Azure Static Web Apps / SWA CLI                    │
└──────┬────────────────────────────────────────────┬─────────────┘
       │ /api/*                                     │ tile requests
┌──────▼──────────────┐                    ┌────────▼─────────────┐
│   hastefuncapi       │                    │   titilerfuncapi     │
│   (41 HTTP routes)   │                    │   (TiTiler/FastAPI)  │
│   Azure Functions    │                    │   COG tile serving   │
└──────┬──────────────┘                    └──────────────────────┘
       │ Queue messages
┌──────▼──────────────┐
│   hastefuncqueues    │
│   (7 queue triggers) │
│   Azure Functions    │
└──────┬──────────────┘
       │
┌──────▼──────────────────────────────────────────────────────────┐
│                    hastegeo core library                          │
│  Config · Models · Processors · Data Layers · Runners · Utils    │
└──────┬───────────┬───────────┬───────────┬──────────────────────┘
       │           │           │           │
  ┌────▼───┐  ┌───▼────┐  ┌──▼───┐  ┌───▼──────────┐
  │ Blob   │  │ Cosmos │  │ Data │  │ Azure Batch  │
  │ Storage│  │ DB     │  │ Lake │  │ (GPU pools)  │
  └────────┘  └────────┘  └──────┘  └──────────────┘

Components

Component	Technology	Description
UI	React + Vite	Single-page app for project management, labeling, and visualization
REST API (hastefuncapi)	Python Azure Functions	HTTP endpoints for CRUD operations
Queue Workers (hastefuncqueues)	Python Azure Functions	Queue-triggered functions for async processing
Tile Server (titilerfuncapi)	TiTiler + FastAPI	Cloud Optimized GeoTIFF tile serving
Core Library (hastegeo)	Python package	Shared models, processors, data layers, and utilities

Contributing

We welcome contributions! Please read CONTRIBUTING.md for coding standards, the pull request process, and the Contributor License Agreement requirement.

• Open an issue to report a bug or request a feature
• Start a discussion for questions or ideas
• Read the security policy before reporting vulnerabilities

License

This project is licensed under the MIT License — see LICENSE.txt for details.

Third-Party Software

This project includes third-party components. See THIRD-PARTY-NOTICES.md for details.