Delectable

A little project to help me learn Python + Temporal + LLM APIs

It's a service takes a list of some of your favourite dishes, finds some similar dishes you might like based on them, then finds restaurants in the specified location that serve them.

The request info is used to create multiple LLM prompts that retrieve the data, which is then processed and returned.

It uses Temporal to orchestrate a workflow.

Setup

1. Install and Start Temporal:

   Download at https://temporal.io/setup/install-temporal-cli

   Open a terminal and run the Temporal development server:

   temporal server start-dev --db-filename dev.db

   Keep the Temporal Web UI open in your browser to see the workflow execution live: http://localhost:8233

2. Start the Application Server:

   In a new terminal, start the app:

   python3 -m venv .venv
   source /venv/bin/activate

   python main.py

3. Start the Temporal Worker:

   In another new terminal, start the Temporal worker that will process the workflows:

   python run_worker.py

Configuration

Before running the application, you need to set the GEMINI_API_KEY environment variable. This is required to authenticate with the Google Gemini API.

export GEMINI_API_KEY="your-api-key"

Usage

1. Submit a Search Request:

   Use curl or any API client to send a POST request to the /search endpoint.

   curl -X POST -H "Content-Type: application/json" -d '{
       "dishes": [
           "Lamb Tikka Dhansak",
           "Bahn mi",
           "Steak and ale pie"
       ],
       "dish_count": 3,
       "location": "London"
   }' http://localhost:8787/search

   The response will be a workflow ID, which you'll use in the next step.

2. Retrieve the Results:

   Once the workflow has had a moment to run (you can watch its progress in the Temporal UI), use the workflow ID from the previous step to make a GET request to the /result/{workflow_id} endpoint.

   Replace {id-from-result} with the actual workflow ID.

   curl http://localhost:8787/result/{id-from-result}

   Once it's finished the response will contain a "success" status a list of dishes and the restaurants where you can find them.