How okapitales/automotive-sales-knowledge works
ASK is a FastAPI-based automotive knowledge assistant. Its demonstrated workflow is: an advisor submits a question, the API cleans and classifies it, then runs a retrieval-and-generation chain over Chroma and returns an answer. The repository also has a separate embedding/search service and vehicle reference documents. The README describes VIN-aware answers, clarification, and dealer handoffs, but the sampled implementation does not show those flows; ingestion and the processed-data producer are also not inspectable here.
Main entry points: assistant_service.py Assistant API · ingest Document ingestion · embed_service.py Embedding service · 2024-volkswagen-id.4.pdf Vehicle references
Components 12
Question handling
- Assistant API
api/assistant_service.py - Query gateway
api/slm_gateway.py - Clarification questions
README.md - Dealer handoff
README.md
Knowledge and retrieval
- Embedding service
embeddings/embed_service.py - Document ingestion
ingestion/ingest - Vehicle references
2024-volkswagen-id.4.pdf - Chroma index
Answer generation
- RAG chain
api/assistant_service.py - Language model
- VIN-aware answers
README.md
Other components
- Advisor or customer
Connections 11
- Advisor or customer → Assistant API: submits question
README.md - Assistant API → Query gateway: cleans and classifies
api/assistant_service.py - Assistant API → RAG chain: runs query
api/assistant_service.py - RAG chain → Chroma index: retrieves context
api/assistant_service.py - RAG chain → Language model: generates answer
api/assistant_service.py - Embedding service → Chroma index: creates or searches
embeddings/embed_service.py - Document ingestion → Chroma index: prepares knowledge
- Vehicle references → Document ingestion: supplies references
- Advisor or customer → Clarification questions: may clarify
README.md - Advisor or customer → VIN-aware answers: may request
README.md - Assistant API → Dealer handoff: can generate
README.md