How lucasnovack/youmightlike works
youmightlike is a local-first, cross-media recommendation product: cultural data is ingested into DuckDB, transformed into a graph, and used by a Rust engine to serve search and personalized recommendations to a web interface. The principal path runs from user-selected taste seeds through HTTP service orchestration, graph diffusion and domain-specific candidate extraction to results. The README also describes UI capabilities including live neighborhood inspection and printable taste receipts. Some Python web-to-engine wiring and unsampled route details are not shown, so those relationships are left unverified.
Main entry points: service.rs HTTP API · database.py DuckDB Feature Store · mod.rs Seed Orchestration · graph.rs CSR Affinity Graph
Components 21
Data Ingestion
- Batch Ingestion
python/src/youmightlike/pipeline.py - Data Connectors
python/src/youmightlike/connectors - Dataset Ingestors
python/src/youmightlike/storage
Rust Recommendation Engine
- HTTP API
crates/engine/src/service.rs - Catalog Search
crates/engine/src/search.rs - Catalog & State
crates/engine/src/models.rs - Graph Construction
crates/engine/src/recommendation/graph_builder.rs - CSR Affinity Graph
crates/engine/src/graph.rs - Seed Orchestration
crates/engine/src/recommendation/mod.rs - PageRank Diffusion
crates/engine/src/ppr.rs - Media Extractors
crates/engine/src/recommendation/extractors - Track Recommendations
crates/engine/src/recommendation/tracks.rs - Recommendation API
crates/engine/src/recommender.rs
Web Experience
- Web Server
python/src/youmightlike/web/server.py - Discovery Panel
python/src/youmightlike/web/static/js/app.js - Cultural Receipt
python/src/youmightlike/web/static/js/receipt.js - Offline Cache
python/src/youmightlike/web/static/sw.js
Python Recommendation
- Python Recommender
python/src/youmightlike/recommendation/engine.py - Seed Resolver
python/src/youmightlike/resolution/resolver.py
Other components
- User
- DuckDB Feature Store
python/src/youmightlike/storage/database.py
Connections 26
- Batch Ingestion → Data Connectors: uses
python/src/youmightlike/pipeline.py - Batch Ingestion → DuckDB Feature Store: initializes
python/src/youmightlike/pipeline.py - Data Connectors → DuckDB Feature Store: writes
python/src/youmightlike/connectors/music.py - Dataset Ingestors → DuckDB Feature Store: writes
python/src/youmightlike/storage/ingest_music.py - DuckDB Feature Store → Graph Construction: supplies edges
- HTTP API → CSR Affinity Graph: loads
crates/engine/src/service.rs - HTTP API → Catalog & State: loads
crates/engine/src/service.rs - HTTP API → Catalog Search: routes search
crates/engine/src/service.rs - HTTP API → Recommendation API: routes recommendations
crates/engine/src/service.rs - Graph Construction → CSR Affinity Graph: builds
crates/engine/src/recommendation/graph_builder.rs - Seed Orchestration → PageRank Diffusion: diffuses seeds
crates/engine/src/recommendation/mod.rs - Seed Orchestration → Media Extractors: extracts candidates
crates/engine/src/recommendation/mod.rs - Seed Orchestration → Track Recommendations: delegates tracks
crates/engine/src/recommendation/mod.rs - Recommendation API → Seed Orchestration: delegates
crates/engine/src/recommender.rs - Catalog Search → Catalog & State: reads metadata
crates/engine/src/search.rs - Catalog Search → CSR Affinity Graph: checks connectivity
crates/engine/src/search.rs - Media Extractors → Catalog & State: uses metadata
crates/engine/src/recommendation/extractors/music.rs - Media Extractors → CSR Affinity Graph: scores neighbors
crates/engine/src/recommendation/extractors/games.rs - User → Discovery Panel: selects taste
- Discovery Panel → Web Server: requests API
- Web Server → HTTP API: forwards requests
- Web Server → Cultural Receipt: provides receipt data
- Discovery Panel → Offline Cache: uses offline cache
- Python Recommender → Seed Resolver: resolves seeds
python/src/youmightlike/recommendation/engine.py - Python Recommender → DuckDB Feature Store: queries
- Python Recommender → HTTP API: calls Rust service