Dishcovery

A blazing-fast recipe search service built with FastAPI, Elasticsearch, and Kubernetes.

Features

• Lightning-fast search across 62,000+ recipes (10-50ms response times)
• Multi-field search across titles, descriptions, ingredients, and directions
• Advanced filtering by cuisine, dietary preferences, prep time, and healthiness score
• Auto-generated API docs at /docs

Quick Start

1. Start Infrastructure Services

docker-compose up -d

2. Run FastAPI Locally (Development)

cd api
pip install uv
uv sync
uv run uvicorn main:app --reload --host 0.0.0.0 --port 8000

Data will be automatically loaded from recipes.json into Elasticsearch on first run.

Note: For development, FastAPI runs locally while Elasticsearch runs in Docker for optimal development experience.

3. Search Recipes

# Basic search
curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "pasta", "size": 5}'

# Filtered search
curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "", "cuisines": ["korean"], "is_vegan": true, "max_prep_time": 30}'

API Endpoints

Endpoint	Method	Description
/	GET	HTML welcome page (Jinja2 + HTMX)
/health	GET	Health check (API + Elasticsearch status)
/search	POST	Search recipes with query parameters
/bulk-load	POST	Load all recipes into Elasticsearch
/docs	GET	Interactive API documentation

Search Examples

Basic Text Search

curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "chocolate", "size": 3}'

Cuisine & Dietary Filters

curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "", "cuisines": ["italian"], "is_vegan": true, "size": 5}'

Time-based Filtering

curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "breakfast", "max_prep_time": 15, "size": 10}'

Complex Search

curl -X POST localhost:8000/search \
    -H "Content-Type: application/json" \
    -d '{"query": "healthy chicken", "cuisines": ["asian"], "max_prep_time": 30, "size": 5}'

Response Format

See example_output.json for a sample search response.

Architecture

See ADR.md for architecture decision records.

Key decisions:

• Server-side rendering with Jinja2 + HTMX (no separate frontend)
• Form-encoded search requests

Roadmap

• Prometheus metrics collection
• Jinja2 + HTMX frontend (in progress)
• Grafana dashboards
• [?] Recipe recommendation engine
• [?] Elasticsearch cluster scaling