bloo-chat-demo.MP4
Github repo
- API endpoints
- Services (have costs/free plan)
- Typesense as a service for hybrid search (full text search + vector search) for retrieval of documents based on a query e.g. user’s question or user’s image items
to store vector embeddings of
wasteType based on the fields item, material,
to perform full text search
- Data cleaning (refer to
datacleaning.ipynb)
- Evaluation of vision model accuracy
- Deployment (Google Cloud Run)
- Cloud firestore database to store NEA items, feedback for each response
- Seeding the cleaned data (NEA items) into Cloud Firestore Database
- Why use large multi-modal models (LMMs) over image classification APIs
- Observability into Langchain Chains via Langsmith
- Tech stack