From 4468a23b2defb07dae567fbbf6aa9a8d69191a70 Mon Sep 17 00:00:00 2001
From: Nam Thanh Nguyen Macbook <thanhnn.ict@gmail.com>
Date: Wed, 5 Aug 2026 22:28:55 +0700
Subject: [PATCH] fix: retry without tools when OpenAI-compatible server
 returns empty response
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Some OpenAI-compatible servers (NIM, vLLM, TGI) return content: null
with an empty tool_calls array when tools are provided in the request
but the model decides not to call any tool. This results in blank
responses in the Continue GUI.

Fix: detect empty response (content: null + tool_calls: []) and retry
the same request without the tools parameter. The model then responds
with text content normally.

Affected models:
- nvidia/nemotron-ultra-253b (NIM) — frequently
- deepseek-ai/DeepSeek-V4-Flash (vLLM) — occasionally
- Any OpenAI-compatible server with tool support

Related: https://github.com/continuedev/continue/issues/5508
---
 packages/openai-adapters/src/apis/OpenAI.ts | 24 ++++++++++++++++++++-
 1 file changed, 23 insertions(+), 1 deletion(-)

diff --git a/packages/openai-adapters/src/apis/OpenAI.ts b/packages/openai-adapters/src/apis/OpenAI.ts
index d0f8d30ca3a..97e67de8a35 100644
--- a/packages/openai-adapters/src/apis/OpenAI.ts
+++ b/packages/openai-adapters/src/apis/OpenAI.ts
@@ -134,12 +134,34 @@ export class OpenAIApi implements BaseLlmApi {
       const response = await this.responsesNonStream(body, signal);
       return responseToChatCompletion(response);
     }
-    const response = await this.openai.chat.completions.create(
+    let response = await this.openai.chat.completions.create(
       this.modifyChatBody(body),
       {
         signal,
       },
     );
+
+    // Some OpenAI-compatible servers (e.g. NIM, vLLM) return content: null
+    // with an empty tool_calls array when tools are provided but the model
+    // decides not to call any. This is effectively an empty response that
+    // the GUI renders as blank. Retry without tools so the model responds
+    // with text content instead.
+    const msg = response.choices?.[0]?.message;
+    if (
+      msg &&
+      !msg.content &&
+      (!msg.tool_calls || msg.tool_calls.length === 0) &&
+      body.tools?.length
+    ) {
+      const { tools, tool_choice, ...bodyWithoutTools } = body;
+      response = await this.openai.chat.completions.create(
+        this.modifyChatBody(
+          bodyWithoutTools as ChatCompletionCreateParamsNonStreaming,
+        ),
+        { signal },
+      );
+    }
+
     return response;
   }