Enable multiple llms

sia/web/token_websocket.py

@@ -0,0 +1,51 @@
+from aiohttp import web, WSMsgType
+from typing import Dict, Set
+import asyncio
+import json
+
+from .util import wrap_async
+from ..web_agent import WebAgent
+
+class TokenWebSocket:
+    """
+    WebSocket handler for LLM token streaming.
+    Broadcasts new tokens to all connected clients.
+    """
+    
+    def __init__(self, agent: WebAgent):
+        self._agent = agent
+        self._clients: Set[web.WebSocketResponse] = set()
+        self._agent.add_token_handler(wrap_async(self._handle_new_token))
+        
+    async def _broadcast_message(self, message: Dict):
+        """Broadcast message to all connected clients."""
+        disconnected = set()
+        for ws in self._clients:
+            try:
+                await ws.send_json(message)
+            except ConnectionResetError:
+                disconnected.add(ws)
+        self._clients -= disconnected
+
+    async def _handle_new_token(self, llm_name: str, token: str):
+        """Handle new tokens from the WebAgent."""
+        await self._broadcast_message({
+            "llm": llm_name,
+            "token": token
+        })
+
+    async def handle_connection(self, request: web.Request) -> web.WebSocketResponse:
+        """Handle new WebSocket connections."""
+        ws = web.WebSocketResponse(heartbeat=30)
+        await ws.prepare(request)
+        
+        self._clients.add(ws)
+        
+        try:
+            async for msg in ws:
+                if msg.type == WSMsgType.ERROR:
+                    print(f"WebSocket connection closed with error: {ws.exception()}")
+        finally:
+            self._clients.remove(ws)
+            
+        return ws