Meal Planner

AI-powered meal planning app. Under development. Running at https://gsganden--meal-planner-web.modal.run/.

graph LR
    User(("User"))

    Modal["Modal (Hosting)"]

    subgraph "Meal Planner Application"
        direction LR

        direction TB
        Web_Routing_Layer["Web Request Routing<br/>(routers/)<br/>FastHTML"]
        
        UI_Components["UI Components<br/>(ui/)<br/>MonsterUI"]

        subgraph "Business Logic (services/)"
            direction TB
            Webpage_Text_Extractor_Service["Webpage text extraction<br/>(extract_webpage_text.py)<br/>URL fetching, HTML cleaning"]
            Recipe_Processing_Service["Recipe processing<br/>(process_recipe.py)<br/>Data cleaning & standardization"]
            LLM_Service["LLM interactions<br/>(call_llm.py)<br/>Google Gemini, Instructor"]
        end

        API_Layer["Recipe CRUD API<br/>(api/recipes.py)<br/>FastAPI"]

        Web_Routing_Layer -- "Renders" --> UI_Components
        
        Web_Routing_Layer -- "Calls" --> Webpage_Text_Extractor_Service
        Web_Routing_Layer -- "Calls" --> Recipe_Processing_Service
        Web_Routing_Layer -- "Calls" --> LLM_Service
        Web_Routing_Layer -- "Internal API Call" --> API_Layer
    end

    subgraph "External Resources"
        direction TB
        Database[("Database (SQLite)")]
        External_Web_Pages["External Web Pages/URLs"]
        Google_Gemini_Cloud["Google Gemini Cloud API"]
    end

    User --> Modal
    Modal --> Web_Routing_Layer

    Webpage_Text_Extractor_Service -- "Fetches content" --> External_Web_Pages
    LLM_Service -- "AI Tasks" --> Google_Gemini_Cloud
    API_Layer --> Database

Loading

Setup

uv sync --all-groups

Get a Gemini API key and assign its value to a GOOGLE_API_KEY environment variable inside a dotenv file.

Install pre-commit hooks:

pre-commit install --hook-type pre-push -f

Run App

Run locally:

uv run modal serve deploy.py

Merging a PR into main triggers deployment as part of a CI/CD process. THat is our usual deployment flow but we can deploy directly from local in a pinch, e.g. for an urgent hotfix:

uv run modal deploy deploy.py

Run Alembic Commands

The app's database is in the Modal environment, so we need to run Alembic CLI commands through Modal like so:

uv run modal shell deploy.py::alembic_env -c "alembic history"

Run Tests

Skip tests that make slow LLM calls:

uv run pytest

Run the tests that make slow LLM calls with a chance to retry once as a way to handle atypical nondeterministic failures:

source .env && uv run pytest tests/test_ml_evals.py --runslow --reruns=1

Check test coverage with minimal LLM calls:

./run_fast_coverage.sh