Module stream_api 

Streaming chat endpoint — the core 1-hop architecture handler.

Flow: Client POST → SharedState (session + cache lookup) → LLM Worker (HTTP to llama-server) → SSE stream back All state access is in-process via Arc/shared memory. The only network hop is to localhost llama-server.

Structs

ChatAttachment

File attachment reference sent with the chat request.

StreamChatRequest

Request body matching what the frontend sends

Functions

generate_stream

POST /generate/stream — Main streaming chat endpoint