feat: add clear error message when llama-server not running

src/models/local/llamacpp.test.ts

@@ -81,4 +81,16 @@ describe('LlamaCppClient', () => {
       usage: { inputTokens: 5, outputTokens: 2 },
     });
   });
+
+  it('throws clear error when server not running', async () => {
+    mockFetch.mockRejectedValue(new TypeError('fetch failed'));
+
+    const client = new LlamaCppClient({
+      endpoint: 'http://localhost:8080',
+    });
+
+    await expect(client.chat({
+      messages: [{ role: 'user', content: 'Hello' }],
+    })).rejects.toThrow('llama-server not running at http://localhost:8080');
+  });
 });

src/models/local/llamacpp.ts

@@ -48,14 +48,22 @@ export class LlamaCppClient implements ModelClient {
       headers['Authorization'] = `Bearer ${this.authToken}`;
     }
 
-    const response = await fetch(`${this.endpoint}/v1/chat/completions`, {
-      method: 'POST',
-      headers,
-      body: JSON.stringify({
-        messages,
-        max_tokens: request.maxTokens ?? 2048,
-      }),
-    });
+    let response: Response;
+    try {
+      response = await fetch(`${this.endpoint}/v1/chat/completions`, {
+        method: 'POST',
+        headers,
+        body: JSON.stringify({
+          messages,
+          max_tokens: request.maxTokens ?? 2048,
+        }),
+      });
+    } catch (error) {
+      if (error instanceof TypeError && error.message.includes('fetch failed')) {
+        throw new Error(`llama-server not running at ${this.endpoint}`);
+      }
+      throw error;
+    }
 
     if (!response.ok) {
       const text = await response.text();