From 2cb9fea514ce58faf7914bfd6c66fc9d17fb3b90 Mon Sep 17 00:00:00 2001
From: Gemini AI <gemini@google.com>
Date: Wed, 24 Dec 2025 21:53:00 +0400
Subject: [PATCH] Add server-side timeout handling to Ollama Cloud streaming

- Added 60 second timeout per chunk in parseStreamingResponse
- Added 120 second timeout to makeRequest with AbortController
- This prevents the server from hanging indefinitely on slow/unresponsive API

This should fix the UI freeze when sending messages to Ollama Cloud models.
---
 .../server/src/integrations/ollama-cloud.ts   | 48 ++++++++++++++++---
 1 file changed, 42 insertions(+), 6 deletions(-)

diff --git a/packages/server/src/integrations/ollama-cloud.ts b/packages/server/src/integrations/ollama-cloud.ts
index b53ad9c..9da6392 100644
--- a/packages/server/src/integrations/ollama-cloud.ts
+++ b/packages/server/src/integrations/ollama-cloud.ts
@@ -338,12 +338,39 @@ export class OllamaCloudClient {
 
     const reader = response.body.getReader()
     const decoder = new TextDecoder()
+    const STREAM_TIMEOUT_MS = 60000 // 60 second timeout per chunk
+    let lastActivity = Date.now()
+
+    const checkTimeout = () => {
+      if (Date.now() - lastActivity > STREAM_TIMEOUT_MS) {
+        reader.cancel().catch(() => { })
+        throw new Error("Stream timeout - no data received for 60 seconds")
+      }
+    }
 
     try {
       while (true) {
-        const { done, value } = await reader.read()
+        checkTimeout()
+
+        // Create a timeout promise
+        const timeoutPromise = new Promise<never>((_, reject) => {
+          setTimeout(() => reject(new Error("Read timeout")), STREAM_TIMEOUT_MS)
+        })
+
+        // Race the read against the timeout
+        let result: ReadableStreamReadResult<Uint8Array>
+        try {
+          result = await Promise.race([reader.read(), timeoutPromise])
+        } catch (timeoutError) {
+          reader.cancel().catch(() => { })
+          throw new Error("Stream read timeout")
+        }
+
+        const { done, value } = result
         if (done) break
 
+        lastActivity = Date.now()
+
         const lines = decoder.decode(value, { stream: true }).split('\n').filter(line => line.trim())
 
         for (const line of lines) {
@@ -371,7 +398,7 @@ export class OllamaCloudClient {
     }
   }
 
-  private async makeRequest(endpoint: string, options: RequestInit): Promise<Response> {
+  private async makeRequest(endpoint: string, options: RequestInit, timeoutMs: number = 120000): Promise<Response> {
     // Ensure endpoint starts with /api
     const apiEndpoint = endpoint.startsWith('/api') ? endpoint : `/api${endpoint}`
     const url = `${this.baseUrl}${apiEndpoint}`
@@ -386,10 +413,19 @@ export class OllamaCloudClient {
 
     console.log(`[OllamaCloud] Making request to: ${url}`)
 
-    return fetch(url, {
-      ...options,
-      headers
-    })
+    // Add timeout to prevent indefinite hangs
+    const controller = new AbortController()
+    const timeoutId = setTimeout(() => controller.abort(), timeoutMs)
+
+    try {
+      return await fetch(url, {
+        ...options,
+        headers,
+        signal: controller.signal
+      })
+    } finally {
+      clearTimeout(timeoutId)
+    }
   }
 
   async getCloudModels(): Promise<OllamaModel[]> {