iGait

📖 Working on the codebase? Head to docs/README.md.

_{Role-aware onboarding (frontend / backend / pipeline / infra), architecture deep-dives, deploy flow, secrets, and kubectl recipes all live there.}

---

iGAIT is an innovative, objective, equitable, widely accessible, easy-to-use, free web-based tool for early autism screening.

Architecture at a glance

iGait is a microservice-based, multi-stage pipeline:

1. Frontend receives the user's video upload.
2. Backend handles user/job creation and the initial submission email, then drops the job onto a Firebase RTDB queue.
3. Pipeline stages (5 of them) each pull from their own queue, transform files via S3, and enqueue for the next stage.
4. The finalize stage accumulates the result or failure and emails the user.

I/O is atomic through Firebase RTDB (queues + job state) and AWS S3 (media files). All services share apps/shared (igait-lib) for domain types and clients. Each stage scales independently — a slow stage just gets more replicas. Deeper architecture notes (stage execution modes, the Firebase client invariant, the queue contract) live under docs/architecture/.

Quick start

Linux or WSL2 only — Windows is strongly discouraged. Install Docker, Nix (with flakes enabled), and direnv (auto-enters the Nix dev shell on cd).

git clone --recurse-submodules https://github.com/igait-niu/igait.git
cd igait
docker compose up   # zero-config; first boot ~5–10min

Then open the frontend at http://localhost:4173. The hermetic stack is env-var-free — no AWS, no Firebase project, no .env required to boot. The only optional passthrough is OPENAI_* in a repo-root .env if you're working on /assistant routes.

For the full topology, port map, troubleshooting, and the why behind every load-bearing config flag, see docs/local-dev/hermetic-stack.md. For everything else — role-aware dev loops, deploy flow, secrets, kubectl recipes — head to docs/README.md.