🥘 Recipe Chat Assistant

A modern, chat-style desktop application that lets you converse naturally with an AI about cooking, recipes, and food preparation. Built with Python 3.8+ and PyQt5, it works with eight LLM providers — local or cloud — and renders beautifully parsed, interactive recipe cards.

---

✨ Key Features

Category	Highlights
Conversational UI	Chat bubbles, an animated "cooking…" indicator, and rich recipe cards
Real chat memory	Full multi-turn context — follow up with "make it vegan" and it remembers
Token streaming	Watch answers arrive live for OpenAI-compatible, Anthropic and Ollama
8 providers	OpenAI, Anthropic, Google Gemini, HuggingFace, Cohere, LM Studio, Ollama, Custom
Live dark / light theme	One click re-themes the whole app, chat history included
Recipe library	Save, favorite, search and re-open recipes (local SQLite)
🧺 Pantry planning	Track what you have → "Cook from pantry", plus one-click shopping lists
Serving scaler	Rescale any recipe (×0.25–×12); amounts and servings recompute instantly
▶️ Cooking Mode	Hands-free, step-by-step view with optional text-to-speech
📎 Multimodal	Attach a dish photo for vision models (OpenAI, Anthropic, Gemini)
Structured output	JSON-mode + a Pydantic schema validate/normalise every recipe
Export	Copy to clipboard or export any recipe as Markdown / JSON / text
CLI + REST API	The same core runs headless — script it or serve it to a web/mobile app
Persistent settings	Provider, model, timeout and theme are remembered between launches
Robust networking	Sane configurable timeouts, retry/backoff, and clear error messages

---

🧩 Module Breakdown

File	Responsibility
models.py	Data models: APIConfig, Message (with image), ParsedRecipe (Markdown export), provider metadata
theme.py	Central ThemeManager with dark/light palettes, live switching, palette + app stylesheet
config.py	Load/save settings to config.json; .env and env-var API-key fallback; build_api_config
paths.py	Cross-platform app-data directory for config, logs and the database
logging_setup.py	Console + rotating-file logging
response_parser.py	Extracts JSON recipes (code fences, trailing commas, prose); plain-text fallback
schema.py	Pydantic schema — validates & normalises structured recipe output
api_client.py	One client, eight providers: streaming, JSON-mode, vision, retries, timeouts
recipe_store.py	SQLite recipe library (save / search / favorite / delete)
pantry_store.py	SQLite pantry of on-hand ingredients
scaling.py	Serving scaling and shopping-list computation (pure functions)
tts.py	Text-to-speech for Cooking Mode (native Qt, graceful fallback)
engine.py	UI-agnostic RecipeEngine facade — the reusable core
cli.py	Headless command-line interface
server.py	Stdlib REST API over the engine (for web/mobile front-ends)
widgets.py	Themed UI: recipe cards, bubbles, cooking mode, pantry, library, settings, toasts
main.py	Main window: streaming worker, memory, pantry/library, scaling, multimodal, theming
tests/	74-test pytest suite across parser, models, config, stores, scaling, schema, engine, client

---

🚀 Quick Start

# 1. Clone
git clone https://github.com/Roialfassi/Recipe-Chat-Assistant-Application.git
cd Recipe-Optimizer

# 2. Create & activate a virtual environment
python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # macOS / Linux

# 3. Install dependencies
pip install -r requirements.txt   # or: make install

# 4. Run
python main.py                    # or: make run

Open ⚙️ Settings, choose a provider, enter an API key (not needed for LM Studio / Ollama), pick or refresh a model, and Save. Then start asking for recipes.

Tip: You can also supply keys via environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY, …) or a local .env file — no need to type them into the UI.

Where your data lives

Settings, logs and the recipe database are stored in your OS app-data folder (e.g. %APPDATA%\RecipeAssistant on Windows). API keys are only written to disk if you tick "Remember API key", and even then they're obfuscated rather than stored in the clear.

---

🤖 Prompt Format

The assistant requests recipes in a strict JSON envelope so they can be rendered as rich cards:

{
  "name": "Spaghetti Carbonara",
  "description": "A classic Roman pasta.",
  "prep_time": "10 minutes",
  "cook_time": "15 minutes",
  "servings": "4",
  "difficulty": "Medium",
  "ingredients": [
    { "amount": "200 g", "item": "spaghetti" },
    { "amount": "100 g", "item": "pancetta, diced" }
  ],
  "instructions": ["Cook spaghetti until al dente.", "Fry pancetta until crispy."],
  "tips": ["Use freshly grated parmesan."],
  "tags": ["italian", "quick", "comfort-food"],
  "nutrition": { "calories": "600", "protein": "25g" }
}

If valid JSON isn't found, the parser falls back to rendering the reply as text.

---

🖥️ Headless: CLI & REST API

The desktop app is just one front-end over a UI-agnostic core (engine.py).

# CLI
python cli.py ask "quick vegan pasta for two"      # prints Markdown
python cli.py ask "dinner" --pantry --save          # cook from pantry, save it
python cli.py ask "carbonara" --json --stream       # stream JSON to stdout
python cli.py list --search taco                    # browse the library
python cli.py pantry add "olive oil"                # manage the pantry
python cli.py serve --port 8765                     # start the REST API

# REST API (same core; CORS-enabled for web clients)
curl -X POST localhost:8765/api/chat -d '{"message":"quick soup"}'
curl localhost:8765/api/recipes
curl localhost:8765/api/pantry

---

🧪 Development

make dev       # install runtime + dev tooling
make test      # run the pytest suite (74 tests)
make lint      # ruff
make format    # black + ruff --fix

---

🛣️ Roadmap

Recently shipped

• Multimodal photo input for vision models
• Hands-free Cooking Mode with text-to-speech
• Pantry-aware planning, serving scaling & shopping lists
• Structured output via JSON mode + Pydantic schema
• Reusable core engine with CLI & REST API

Next

• Inline generated step photos (image output)
• Speech-to-text voice input
• Cloud sync of favourite recipes
• Keyring-backed secret storage
• Companion web / mobile front-end on the REST API

---

📜 License

Distributed under the MIT License.

🙏 Acknowledgements

• PyQt5 — GUI framework
• LM Studio & Ollama — local model serving

Bon appétit 👨‍🍳