How neeleshpanchumarthi/fashion-recommender-system works
StyleIQ is a fashion discovery product with a React client, a FastAPI search/try-on API, and a separate offline catalog-ingestion worker. The main shopping flow turns a natural-language query into understood filters, retrieves and reranks products, then returns results for display; outfit-aware and multilingual search extend that flow. Separate branches provide asynchronous virtual try-on, review insights, and index preparation. The graph follows documented flows and indexed source references; some frontend/API wiring and unsampled subsystems are linked only where the README documents them.
Main entry points: ingestion_worker.py Ingestion worker · pipeline.py Retrieval pipeline · useSearch.ts Search experience · apiClient.ts API client
Components 27
Shopping frontend
- React app
frontend/src/App.tsx - Search experience
frontend/src/hooks/useSearch.ts - Product results
frontend/src/pages/SearchResults.tsx - Try-on dialog
frontend/src/components/tryon/TryOnDialog.tsx - API client
frontend/src/services/apiClient.ts
API and services
- API routes
app/api/router.py - Search endpoint
app/api/routes/search.py - Search service
app/services/search_service.py - Try-on endpoint
app/api/routes/tryon.py - Try-on jobs
app/services/tryon_service.py - Review insights
app/services/product_enrichment_service.py
Search and domain
- Retrieval pipeline
app/retrieval/pipeline.py - Query understanding
app/retrieval/query_processor.py - Dense retrieval
app/retrieval/retriever.py - Result reranking
app/retrieval/reranker.py - Vector repository
app/repositories/vector_repository.py - Text embeddings
app/clients/embedding_client.py
Catalog ingestion
- Dataset access
app/repositories/dataset_repository.py - Ingestion worker
workers/ingestion_worker.py - Review analysis
app/services/review_analysis_service.py - Batch processing
workers/batch_processor.py - Checkpointing
workers/checkpoint_manager.py
External systems
- Groq LLM
app/clients/llm_client.py - Pinecone index
app/clients/pinecone_client.py - Sentiment model
app/clients/sentiment_client.py - Hugging Face Space
app/clients/tryon_client.py
Other components
- Shopper
Connections 28
- Shopper → React app: uses
README.md - React app → Search experience: renders
- Search experience → API client: requests search
README.md - API client → Search endpoint: calls API
README.md - API routes → Search endpoint: routes to
app/api/router.py - Search endpoint → Search service: calls
app/api/routes/search.py - Search service → Retrieval pipeline: runs
app/services/search_service.py - Retrieval pipeline → Query understanding: processes query
app/retrieval/pipeline.py - Query understanding → Groq LLM: understands query
app/retrieval/query_processor.py - Retrieval pipeline → Text embeddings: embeds query
app/retrieval/pipeline.py - Retrieval pipeline → Dense retrieval: retrieves candidates
app/retrieval/pipeline.py - Dense retrieval → Vector repository: searches vectors
app/retrieval/retriever.py - Vector repository → Pinecone index: queries index
app/repositories/vector_repository.py - Retrieval pipeline → Result reranking: reranks candidates
app/retrieval/pipeline.py - Retrieval pipeline → Search service: returns results
app/services/search_service.py - Search experience → Product results: shows results
- API routes → Try-on endpoint: routes to
app/api/router.py - Try-on endpoint → Try-on jobs: submits or polls
app/api/routes/tryon.py - Try-on jobs → Hugging Face Space: runs try-on
app/services/tryon_service.py - Try-on dialog → API client: uses API
- Ingestion worker → Dataset access: loads catalog
workers/ingestion_worker.py - Ingestion worker → Batch processing: processes batches
workers/ingestion_worker.py - Ingestion worker → Review analysis: analyzes reviews
workers/ingestion_worker.py - Review analysis → Sentiment model: classifies reviews
app/services/review_analysis_service.py - Ingestion worker → Text embeddings: embeds products
workers/ingestion_worker.py - Ingestion worker → Vector repository: upserts vectors
workers/ingestion_worker.py - Ingestion worker → Checkpointing: records progress
workers/ingestion_worker.py - Review insights → Review analysis: uses analysis