How nhutminhit/vector-space-model works
Anime Vector Space Model is a semantic-search application with a FastAPI API and web UI, plus a notebook-based analysis workflow. The main user path is a free-text or similar-anime request through a router and controller, then query embedding and Milvus retrieval; a separate TF-IDF baseline provides sparse search. The import path cleans anime data, builds documents, encodes them, and stores vectors. The README also documents model comparison and evaluation. Some notebook and UI internals are not excerpted, so those connections are shown only where documented.
Main entry points: search_service.py Dense search and import · main.py FastAPI app · data_loader.py Anime preprocessing · cli.py Import and stats CLI
Components 18
API and UI
- FastAPI app
apps/main.py - Search routes
apps/router/search_router.py - Import routes
apps/router/import_router.py - TF-IDF routes
apps/router/tfidf.py - Request schemas
apps/router/schemas.py - Web UI
apps/static/app.js
Data and ingestion
- Import controller
apps/controller/import_controller.py - Anime preprocessing
apps/services/data_loader.py - Dense search and import
apps/services/search_service.py
Search methods
- Search controller
apps/controller/search_controller.py - Embedding models
apps/services/embedding.py - Milvus vector store
apps/services/milvus_store.py - TF-IDF search
apps/services/tfidf_search.py
Research and operations
- Retrieval evaluation
apps/services/evaluation.py - VSM analysis
vector-space-model.ipynb - Import and stats CLI
apps/cli.py - Model and settings config
apps/config.py
Other components
- Search user
Connections 29
- Search user → FastAPI app: sends requests
README.md - FastAPI app → Search routes: registers routes
apps/main.py - FastAPI app → Import routes: registers routes
apps/main.py - FastAPI app → TF-IDF routes: registers routes
apps/main.py - FastAPI app → Web UI: mounts UI
apps/main.py - Search routes → Request schemas: uses schemas
apps/router/search_router.py - Import routes → Request schemas: uses schemas
apps/router/import_router.py - Search routes → Search controller: delegates search
apps/router/search_router.py - Import routes → Import controller: delegates import
apps/router/import_router.py - Import controller → Anime preprocessing: loads anime
apps/controller/import_controller.py - Import controller → Dense search and import: ensures import
apps/controller/import_controller.py - Search controller → Dense search and import: runs retrieval
apps/controller/search_controller.py - Search controller → Anime preprocessing: loads similar source
apps/controller/search_controller.py - Dense search and import → Anime preprocessing: loads documents
apps/services/search_service.py - Dense search and import → Embedding models: encodes passages and queries
apps/services/search_service.py - Dense search and import → Milvus vector store: creates and searches collections
apps/services/search_service.py - TF-IDF routes → Anime preprocessing: loads anime
apps/router/tfidf.py - TF-IDF routes → TF-IDF search: runs sparse search
apps/router/tfidf.py - Import and stats CLI → Anime preprocessing: loads anime
apps/cli.py - Import and stats CLI → Embedding models: downloads models
apps/cli.py - Import and stats CLI → Dense search and import: imports data
apps/cli.py - Import and stats CLI → Milvus vector store: reads stats
apps/cli.py - VSM analysis → Retrieval evaluation: evaluates retrieval
README.md - Retrieval evaluation → Embedding models: uses embeddings
apps/services/evaluation.py - Retrieval evaluation → Milvus vector store: uses vector search
apps/services/evaluation.py - Retrieval evaluation → TF-IDF search: compares baseline
apps/services/evaluation.py - Dense search and import → Model and settings config: reads model registry
apps/services/search_service.py - Embedding models → Model and settings config: reads model settings
apps/services/embedding.py - Milvus vector store → Model and settings config: reads store settings
apps/services/milvus_store.py