How brunolopaiva/local-llm-interface works
ChatLLM is a local document-chat application: users ask questions in a browser, the server retrieves relevant chunks from a knowledge base, and a local Ollama model streams a grounded response back to the chat UI. A separate training page appends new text and rebuilds the in-memory index. The RAG stages and socket handlers are implemented together in server.js, so several nodes share that path. Relationships shown are supported by the server/client excerpts or README; HTML details and other unsampled source were not inspected.
Main entry points: server.js HTTP server · client.js Chat interface · treino.js Training form · index.html Chat page
Components 15
Web experience
- Chat page
public/index.html - Chat interface
public/client.js - Training page
public/treino.html - Training form
public/treino.js
Application server
- HTTP server
server.js - Socket handlers
server.js
Knowledge and RAG
- Knowledge text
data.txt - Text chunking
server.js - Ollama embeddings
server.js - Memory vector store
server.js - Context retrieval
server.js - RAG prompt
server.js - Local generation
server.js
Other components
- User
- Ollama service
Connections 25
- User → Chat page: opens
README.md - Chat page → Chat interface: loads
- User → Chat interface: asks
README.md - Chat interface → Socket handlers: sends chat
public/client.js - HTTP server → Chat page: serves static
server.js - HTTP server → Training page: serves route
server.js - Training page → Training form: loads
- User → Training form: submits text
README.md - Training form → Socket handlers: sends training
public/treino.js - HTTP server → Socket handlers: hosts sockets
server.js - Socket handlers → Chat interface: streams events
server.js - Socket handlers → Training form: sends status
server.js - HTTP server → Knowledge text: reads at startup
server.js - Socket handlers → Knowledge text: appends text
server.js - HTTP server → Text chunking: splits text
server.js - HTTP server → Ollama embeddings: creates embeddings
server.js - HTTP server → Memory vector store: builds index
server.js - HTTP server → Context retrieval: creates retriever
server.js - HTTP server → RAG prompt: configures prompt
server.js - HTTP server → Local generation: configures chain
server.js - Socket handlers → Context retrieval: queries chain
server.js - Socket handlers → Local generation: streams generation
server.js - Ollama embeddings → Ollama service: embeds locally
server.js - Local generation → Ollama service: generates locally
server.js - Socket handlers → HTTP server: rebuilds RAG
server.js