AI-Diagnostic-Imaging-Agent

🚀 Deployment Guide

This guide covers the two supported deployment targets: Docker / docker-compose (local, VPS, or cloud) and Streamlit Community Cloud (free public demo).

1. Docker / docker-compose

Build & run

cp .env.example .env      # set GOOGLE_API_KEY (users can also enter it in the UI)
docker compose up --build -d

App: http://localhost:8501 — health endpoint: http://localhost:8501/_stcore/health

Useful commands

docker compose logs -f            # follow logs
docker compose ps                 # status + healthcheck
docker compose down               # stop
docker compose up --build -d      # rebuild after code changes

Image characteristics

Multi-stage build on python:3.12-slim (small final image, no build tools shipped)
Runs as non-root user appuser
Built-in HEALTHCHECK on the Streamlit health endpoint
Logs go to stdout (docker logs); file log at /tmp/medical_agent.log
Memory limit of 1 GB set in docker-compose.yml (adjust to your host)

Deploying on a VPS / cloud VM (AWS EC2, etc.)

Install Docker + the compose plugin on the server
Clone the repo, create .env with your GOOGLE_API_KEY
docker compose up --build -d
Put a reverse proxy (Nginx/Caddy) in front for TLS, forwarding to port 8501. For Streamlit WebSockets, enable proxy_http_version 1.1 and the Upgrade/Connection headers.

2. Streamlit Community Cloud (free)

Push the repository to GitHub
Go to share.streamlit.io and sign in with GitHub
New app → select the repo, branch main, main file src/app.py
(Optional) In Advanced settings → Secrets, add:
```
GOOGLE_API_KEY = "your_key_here"
```
If omitted, users enter their own key in the sidebar.
Deploy — you get a public https://<app>.streamlit.app URL

Notes:

Python version is read from the platform settings; select 3.12
Dependencies are installed from requirements.txt automatically

3. Configuration reference

Environment variables (via .env, compose environment, or platform secrets):

Variable	Default	Description
`GOOGLE_API_KEY`	–	Gemini API key; if unset, users enter it in the UI
`MODEL_ID`	`gemini-flash-latest`	Gemini model used for analysis
`FALLBACK_MODEL_ID`	`gemini-3-flash-preview`	Model used automatically when the primary is overloaded (503) or out of quota (429)
`ENABLE_WEB_SEARCH`	`true`	Live web search for references; set `false` to use 1 request/analysis (saves quota)
`DEFAULT_LANGUAGE`	`en`	Default UI & report language (`en`/`fr`), switchable in the UI
`MAX_IMAGE_SIZE`	`5242880`	Max upload size in bytes (5 MB)
`MAX_ANALYSIS_TIME`	`120`	Analysis timeout (seconds)
`LOG_LEVEL`	`INFO`	Logging verbosity
`LOG_FILE`	`medical_agent.log`	Optional log file path

4. CI/CD

.github/workflows/ci.yml runs on every push/PR to main:

Lint — ruff check src/ tests/
Tests — pytest with coverage
Docker — image build + container smoke test against the health endpoint

To publish images automatically, add a push step to Docker Hub or GHCR with repository secrets.

⚠️ Reminder

This application is for educational purposes only and must not be used for clinical decision-making without review by qualified healthcare professionals.

This site is open source. Improve this page.