fix: spoof ollama streaming responses for tool calls

src/routes/ollama.ts

@@ -1,20 +1,20 @@
 import { FastifyInstance, FastifyPluginAsync } from 'fastify';
-import { forwardChatRequest } from '../proxy/forward';
+import { forwardAndSpoofOllamaStreamRequest, forwardChatRequest } from '../proxy/forward';
 import { logger } from '../utils/logger';
 
 const ollamaRoutes: FastifyPluginAsync = async (server: FastifyInstance) => {
   server.post('/api/chat', async (request, reply) => {
     try {
       const body = request.body as any;
-      
-      // Currently only supporting non-streaming requests in this proxy MVP
+
       if (body?.stream === true) {
-        // As per requirements: return clear error or pass through without rewriting
-        // We'll return a clear error for now, because stream parsing is out of scope for MVP
-        reply.status(400).send({ 
-          error: "Streaming is not supported by this proxy MVP. Please set stream=false." 
-        });
-        return;
+        const spoofedStream = await forwardAndSpoofOllamaStreamRequest(body, request.headers.authorization);
+
+        reply.raw.setHeader('Content-Type', 'text/event-stream');
+        reply.raw.setHeader('Cache-Control', 'no-cache');
+        reply.raw.setHeader('Connection', 'keep-alive');
+
+        return reply.send(spoofedStream);
       }
 
       const response = await forwardChatRequest(body, request.headers.authorization);