Appendices
Supplementary materials for the Conversational AI course.
Chatbot Pipeline Patterns
These appendices examine three progressively sophisticated patterns for building document-grounded chatbots:
- The CVC Pattern — Content → Vector Store → Chatbot. The simplest pipeline: chunk documents, embed them in a vector store, and retrieve by similarity.
- The CCGC Pattern — Content → Concepts → Graph → Chatbot. An NLP pipeline extracts concepts and relationships into a knowledge graph that the chatbot traverses to answer questions.
- The CCGCC Pattern — Content → Concepts → Graph → Compression → Chatbot. A graph compression stage converts subgraphs into token-efficient markdown, treating the context window as a precious resource. A neuro-symbolic design combining deterministic graph traversal with LLM generation.