[Bug]API Error when calling tools

[cocolinfff]

What happened?

It meet errors when calling function tools.

And sometimes ,it just think for a long long long time and error,and, retry. I have lost a lot of token because of this shit bug.
oh my,it's much more expensive than claude-code because of this bug

What did you expect to happen?

It should read the file and solve my work.

Client information

Details

 ✔  ReadFile config.json  │
 ╰──────────────────────────╯
✕ [API Error: OpenAI API error: 400 <400>
  InternalError.Algo.InvalidParameter: An assistant message with
  "tool_calls" must be followed by tool messages responding to each
  "tool_call_id". The following tool_call_ids did not have response
  messages: message[7].role]

✦ Now let's look at the main entry point of the application:

 ╭──────────────────────╮
 │ ✔  ReadFile main.py  │
 ╰──────────────────────╯
✕ [API Error: OpenAI API error: 400 <400>
  InternalError.Algo.InvalidParameter: An assistant message with
  "tool_calls" must be followed by tool messages responding to each
  "tool_call_id". The following tool_call_ids did not have response
  messages: message[7].role]

another error log:

Details

 ✖  Error flushing log events: AggregateError [ETIMEDOUT]:    │
│        at internalConnectMultiple (node:net:1122:18)         │
│        at afterConnectMultiple (node:net:1689:7) {           │
│      code: 'ETIMEDOUT',                                      │
│      [errors]: [                                             │
│        Error: connect ETIMEDOUT 216.239.32.223:443           │
│            at createConnectionError (node:net:1652:14)       │
│            at Timeout.internalConnectMultipleTimeout         │
│    (node:net:1711:38)                                        │
│            at listOnTimeout (node:internal/timers:583:11)    │
│            at process.processTimers                          │
│    (node:internal/timers:519:7) {                            │
│          errno: -4039,                                       │
│          code: 'ETIMEDOUT',                                  │
│          syscall: 'connect',                                 │
│          address: '216.239.32.223',                          │
│          port: 443                                           │
│        },                                                    │
│        Error: connect ETIMEDOUT 216.239.36.223:443           │
│            at createConnectionError (node:net:1652:14)       │
│            at Timeout.internalConnectMultipleTimeout         │
│    (node:net:1711:38)                                        │
│            at listOnTimeout (node:internal/timers:583:11)    │
│            at process.processTimers                          │
│    (node:internal/timers:519:7) {                            │
│          errno: -4039,                                       │
│          code: 'ETIMEDOUT',                                  │
│          syscall: 'connect',                                 │
│          address: '216.239.36.223',                          │
│          port: 443                                           │
│        },                                                    │
│        Error: connect ETIMEDOUT 216.239.34.223:443           │
│            at createConnectionError (node:net:1652:14)       │
│            at Timeout.internalConnectMultipleTimeout         │
│    (node:net:1711:38)                                        │
│            at listOnTimeout (node:internal/timers:583:11)    │
│            at process.processTimers                          │
│    (node:internal/timers:519:7) {                            │
│          errno: -4039,                                       │
│          code: 'ETIMEDOUT',                                  │
│          syscall: 'connect',                                 │
│          address: '216.239.34.223',                          │
│          port: 443                                           │
│        },                                                    │
│        Error: connect ETIMEDOUT 216.239.38.223:443           │
│            at createConnectionError (node:net:1652:14)       │
│            at afterConnectMultiple (node:net:1682:16) {      │
│          errno: -4039,                                       │
│          code: 'ETIMEDOUT',                                  │
│          syscall: 'connect',                                 │
│          address: '216.239.38.223',                          │
│          port: 443                                           │
│        }                                                     │
│      ]                                                       │
│    } (x9)   

Anything else we need to know?

I'm not sure if I need to use VPN?
Gemini-cli need a clean VPN env, but in qwen-code I'm not sure

[tecgic]

你欠费了 ┭┮﹏┭┮

[cocolinfff]

你欠费了 ┭┮﹏┭┮

absolutely not. I'm sure i still have balance.
And this error happens occasionally , maybe restart it and wait for a moment it works again

---

Wait a moment , just now my balance to 0?? it cost 150￥ just for 4 hours??

[chazmaniandinkle]

Root Cause Analysis & Fix Available

I encountered the same errors and tracked down the root cause. The issue is in the message format conversion process in packages/core/src/core/openaiContentGenerator.ts.

Technical Root Cause

Tool results are being serialized with role: "user" instead of role: "tool" during Gemini-to-OpenAI format conversion. This violates the OpenAI API specification that requires tool responses to have role: "tool" with a tool_call_id.

Evidence

I added debug instrumentation that dumps message payloads to .debug/ files. Here's what shows the bug:

Current (incorrect) serialization:

{
  "role": "user",               // ❌ Wrong - causes the 400 error you're seeing
  "parts": [
    {
      "functionResponse": {
        "id": "call_ikfws9zl", 
        "name": "read_many_files",
        "response": {...}
      }
    }
  ]
}

Expected (correct) serialization:

{
  "role": "tool",               // ✅ Correct
  "tool_call_id": "call_ikfws9zl",
  "content": "{...stringified response...}"
}

Why This Causes Your 400 Error

The OpenAI API validates that every assistant message with tool_calls is followed by corresponding tool messages (with role: "tool" and matching tool_call_id). When tool responses get labeled as role: "user", the API can't match them up and throws:

InternalError.Algo.InvalidParameter: An assistant message with "tool_calls" must be followed by tool messages responding to each "tool_call_id"

Location of Bug

File: packages/core/src/core/openaiContentGenerator.ts
Lines: ~821-858 (fallback role assignment logic)

The issue is in the else-fallback that defaults non-model content to role: "user" instead of properly handling functionResponse parts.

Fix Status

I'm preparing a PR with:

1. ✅ Corrected role assignment for tool results
2. ✅ Regression test to prevent recurrence
3. ✅ Debug instrumentation to help diagnose similar issues

The fix ensures only genuine human input gets role: "user" while tool results properly get role: "tool".

Branch: fix/tool-role-serialization (will reference this issue in PR)

[pomelo-nwu]

This issue has been open for over 6 months with no recent activity. It appears to be related to an early version of Qwen Code.
If you can still reproduce this issue with the latest version, please provide updated reproduction steps and we'll investigate.
Closing for now - feel free to reopen with fresh information.