How kishankumar1328/kirstonai works
KirstonAI combines conversational AI, document-grounded retrieval, autonomous coding, and multimodal utilities in a React/FastAPI application. The typical path is user → routed UI page → API endpoint → service/orchestrator → model or retrieval, with conversation state persisted in PostgreSQL. Separate branches provide coding tools, search, speech synthesis, analytics, and 3D generation. The README documents orchestration and integrations; sampled source confirms app/router registration and selected tool, ingestion, and UI behavior, but not every service-to-service call, so uncertain wiring is omitted.
Main entry points: App.tsx Page router · api.ts API client · main.py FastAPI app · api Feature endpoints
Components 30
User interface
- Page router
frontend/src/App.tsx - Chat UI
frontend/src/pages/ChatPage.tsx - Coding UI
frontend/src/pages/CodingAgentPage.tsx - File chat UI
frontend/src/pages/FileChatPage.tsx - Product pages
frontend/src/pages - API client
frontend/src/services/api.ts
API boundary
- FastAPI app
backend/app/main.py - Chat endpoint
backend/app/api/chat.py - Coding endpoint
backend/app/api/coding_agent.py - Feature endpoints
backend/app/api
AI and retrieval
- Query orchestrator
backend/app/agents/graph.py - Agent pathways
backend/app/agents/nodes.py - LLM provider
backend/app/ai/llm.py - RAG pipeline
backend/app/rag - Document chunking
backend/app/rag/ingestion.py
Product services
- Chat service
backend/app/services/chat_service.py - Coding engine
backend/app/services/coding_agent_service.py - Coding tools
backend/app/services/tools_engine.py - Workspace operations
backend/app/services/workspace_manager.py - Terminal execution
backend/app/services/terminal_executor.py - Speech synthesis
backend/app/services/tts_service.py - 3D generation
backend/app/services/object3d_service.py - Conversation service
backend/app/services/conversation_service.py
Persistence
- Repositories
backend/app/repositories - PostgreSQL
backend/app/database/session.py - Vector store
backend/app/rag/vectorstore.py
Other components
- User
- AI model APIs
- TTS providers
- 3D generation API
Connections 32
- User → Page router: opens
- Page router → Chat UI: routes
- Page router → Coding UI: routes
- Page router → File chat UI: routes
- Page router → Product pages: routes
- Chat UI → API client: requests
- Coding UI → API client: requests
- File chat UI → API client: requests
- Product pages → API client: requests
- API client → FastAPI app: HTTP requests
- FastAPI app → Chat endpoint: mounts
- FastAPI app → Coding endpoint: mounts
- FastAPI app → Feature endpoints: mounts
- Chat endpoint → Chat service: dispatches
- Chat service → Query orchestrator: invokes
- Query orchestrator → Agent pathways: routes
- Agent pathways → LLM provider: requests inference
- Agent pathways → RAG pipeline: retrieves context
- LLM provider → AI model APIs: calls
- RAG pipeline → Vector store: searches
- Document chunking → Vector store: stores chunks
- Coding endpoint → Coding engine: dispatches
- Coding engine → Coding tools: executes tools
- Coding tools → Workspace operations: operates files
- Coding tools → Terminal execution: runs commands
- Feature endpoints → Speech synthesis: dispatches TTS
- Speech synthesis → TTS providers: synthesizes with
- Feature endpoints → 3D generation: dispatches generation
- 3D generation → 3D generation API: calls
- Conversation service → Repositories: uses
- Repositories → PostgreSQL: reads and writes
- Chat service → Conversation service: manages threads