MaxMode

A no-BS, self-hosted weight and calorie tracker built with FastAPI + Jinja templates and offline-first frontend behavior.

Architecture

• Backend: main.py serves full templates and HTMX partials.
• Frontend bootstrap: static/js/bootstrap.mjs (exports window.MaxMode API compatibility layer).
• Frontend modules:
  • static/js/modules/storage.mjs: localStorage-backed state snapshot (loadState) and mutations.
  • static/js/modules/data-utils.mjs: pure weight/chart/date utilities.
  • static/js/modules/charts.mjs: dashboard/weights SVG chart rendering.
  • static/js/views/dashboard-ui.mjs: dashboard metric rendering.
  • static/js/views/weights-ui.mjs: weights list/chart/modal/delete UI.
  • static/js/views/profile-ui.mjs: profile + avatar rendering.
• HTMX compatibility layer: static/vendor/htmx.min.js (self-hosted).
• PWA worker: service-worker.js with safe caching rules and update handling.

Run

uv run uvicorn main:app --reload

Open: http://127.0.0.1:8000

AI Meal Parsing

• The calorie tracker uses LiteLLM for provider-agnostic meal estimation.
• Current dev/test default: set OPENAI_API_KEY in .env.
• Optional provider override: set MEAL_AI_MODEL (for example anthropic/..., openai/..., or vertex_ai/...).

Test

Backend route/rendering tests:

uv run python -m unittest discover -s tests -v

Frontend pure utility tests:

node --test tests_frontend/*.test.mjs

Regression Checklist

1. Load /, /weights, /profile online.
2. Add/edit/delete weight entries and verify chart + dashboard + profile updates.
3. Switch chart ranges (7D, 30D, 90D, All) and verify chart redraw.
4. Refresh with service worker active and confirm app still loads offline.
5. Verify HX-Request partial swaps update only #main-content.