LLM stacks \ 4.4 Human input (deep dive)
25.0717 (0626) Lab notes (Gdrive), Git
- 1 Human enters prompt via UI (web interface, CLI, chat app, etc.)
- 2 UI validates and formats the input
- Handles special characters, length limits
- May add context (conversation history, user preferences)
- Example: "User: What is quantum computing?" → formatted JSON
- 3 UI calls Agent API with formatted prompt
- POST /chat endpoint with user message
- May include metadata (user_id, session_id, preferences)
- 4 UI handles Agent response
- Receives streaming or complete response
- Formats for display (markdown, syntax highlighting)
- Updates conversation history
- Example: Your simple_rag.py uses basic CLI input() for this layer
