Chat Playground

The Chat Playground provides a ChatGPT-style interface for interacting with your scientific knowledge bases.

Overview

• SSE Streaming: Real-time token-by-token responses via Server-Sent Events
• Knowledge Base Selection: Choose one or more knowledge bases for RAG-augmented answers
• Tool Modes: Switch between QA, citation lookup, review outline, and gap analysis
• Rich Rendering: Markdown with LaTeX math (KaTeX), code highlighting, and inline citations

Tool Modes

Mode	Description
QA (default)	Ask questions answered using RAG from selected knowledge bases
Citation Lookup	Provide a text passage, find matching citations from your papers
Review Outline	Generate a structured literature review outline
Gap Analysis	Identify research gaps in the selected knowledge base

API Endpoints

POST /api/v1/chat/stream          # SSE streaming chat
GET  /api/v1/conversations        # List conversations
POST /api/v1/conversations        # Create conversation
GET  /api/v1/conversations/{id}   # Get with messages
DELETE /api/v1/conversations/{id} # Delete conversation

Architecture

Frontend (useChat) → POST /chat/stream → LLMClient (LangChain)
                                              ↓
                                        RAGService (LlamaIndex)
                                              ↓
                                        ChromaDB Vector Store

The chat system uses LangChain's BaseChatModel abstraction, allowing seamless switching between providers (OpenAI, Anthropic, Ollama, etc.) via the Settings page or .env configuration.