[BUG] Unexpected API Response: The language model did not provide any assistant messages. This may indicate an issue with the API or the model's output.

[PlanBernhard]

Problem (one or two sentences)

Date/time: 2026-05-07T19:43:10.092Z
Extension version: 3.53.0
Provider: roo
Model: anthropic/claude-sonnet-4.6

Unexpected API Response: The language model did not provide any assistant messages. This may indicate an issue with the API or the model's output.

Context (who is affected and when)

me using anthropic/claude-sonnet-4.6 with a custom mode

Reproduction steps

use anthropic/claude-sonnet-4.6 with a custom mode

Expected result

That it works

Actual result

Error

Variations tried (optional)

No response

App Version

Version: 3.53.0 (44f6bba)

API Provider (optional)

None

Model Used (optional)

No response

Roo Code Task Links (optional)

REMOVED SEE ANSWER BELOW

Relevant logs or errors (optional)

[roomote-v0]

I'd like to help investigate this issue, but I need some additional information to narrow down the root cause. Could you please clarify the following:

1. Is this error consistent or intermittent? Does it happen every time you use anthropic/claude-sonnet-4.6 with your custom mode, or only sometimes?
2. Does the same model work in a non-custom mode? For example, does anthropic/claude-sonnet-4.6 work fine in the default Code mode or other built-in modes?
3. Custom mode details: Could you share the configuration of your custom mode (the relevant parts from .roomodes or the settings)? Specifically, what tools are enabled, and does it have a custom system prompt?
4. Error log contents: The attached error_roocode.txt file is referenced but I cannot access its contents directly. Could you paste the key parts of that log (especially any stack traces or API error details) into a comment?
5. Do other models work with the same custom mode? For example, does switching to a different model (like anthropic/claude-sonnet-4) with the same custom mode configuration work correctly?

The error "The language model did not provide any assistant messages" occurs when the API stream completes without returning any text or tool-call content blocks (see Task.ts). This can happen for several reasons -- a server-side issue, model misconfiguration, or the API returning an empty response. Understanding the pattern above will help determine whether this is a client-side bug or a transient API issue.

Please tag @roomote in your response with the answers, and I'll be happy to investigate further or propose a fix.

[chereekana-droid]

显然，是昨天才出现的错误，直至现在
// Please share this file with Roo Code Support (support@roocode.com) to diagnose the issue faster
// Just make sure you're OK sharing the contents of the conversation below.

{
"error": {
"timestamp": "2026-05-08T04:02:29.038Z",
"version": "3.53.0",
"provider": "openai",
"model": "gemma4:31b",
"details": "语言模型返回了一个没有任何文本或工具调用的空响应。这可能是由于 API 问题、速率限制或模型特定问题所导致。系统已自动提示模型重试。"
},
"history": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "<user_message>\n你好\n</user_message>"
},
{
"type": "text",
"text": "<environment_details>\n# VSCode Visible Files\n\n\n# VSCode Open Tabs\n\n\n# Current Time\nCurrent time in ISO 8601 UTC format: 2026-05-08T04:01:29.333Z\nUser time zone: Asia/Shanghai, UTC+8:00\n\n# Current Cost\n$0.00\n\n# Current Mode\ncode\n💻 Code\ngemma4:31b\n\n\n# Current Workspace Directory (c:/Users/1/Desktop/cb-import-site) Files\n404.html\nabout.html\nfaq.html\nindex.html\ninquiry.html\nprivacy.html\nproducts.html\nrobots.txt\nsitemap.xml\nadmin/\nadmin/index.html\nadmin/inquiries-mgt.html\nadmin/posts-mgt.html\nadmin/products-mgt.html\nassets/\nassets/cases/\nassets/icons/\nassets/img/\nassets/products/\nblog/\nblog/index.html\ncss/\ncss/admin-style.css\ncss/style.css\ndata/\ndata/inquiries.json\ndata/posts.json\ndata/products.json\njs/\njs/admin-engine.js\njs/blog-engine.js\njs/inquiry-handler.js\njs/main.js\njs/product-engine.js\nposts/\nYou have not created a todo list yet. Create one with update_todo_list if your task is complicated or involves multiple steps.\n</environment_details>"
}
],
"ts": 1778212889376
},
{
"role": "assistant",
"content": [
{
"type": "text",
"text": "你好！我是 Roo，一名经验丰富的软件工程师。很高兴能为你提供帮助。\n\n我注意到你当前处于 c:/Users/1/Desktop/cb-import-site 项目目录中，这是一个包含产品展示、博客、咨询管理以及后台管理系统的静态站点。\n\n请问有什么我可以帮你实现的？无论是开发新功能、修复 Bug 还是优化现有代码，请随时告诉我。"
}
],
"ts": 1778212892612
},
{
"role": "user",
"content": [
{
"type": "text",
"text": "[ERROR] You did not use a tool in your previous response! Please retry with a tool use.\n\n# Reminder: Instructions for Tool Use\n\nTools are invoked using the platform's native tool calling mechanism. Each tool requires specific parameters as defined in the tool descriptions. Refer to the tool definitions provided in your system instructions for the correct parameter structure and usage examples.\n\nAlways ensure you provide all required parameters for the tool you wish to use.\n\n# Next Steps\n\nIf you have completed the user's task, use the attempt_completion tool.\nIf you require additional information from the user, use the ask_followup_question tool.\nOtherwise, if you have not completed the task and do not need additional information, then proceed with the next step of the task.\n(This is an automated message, so do not respond to it conversationally.)"
},
{
"type": "text",
"text": "<environment_details>\n# VSCode Visible Files\n\n\n# VSCode Open Tabs\n\n\n# Current Time\nCurrent time in ISO 8601 UTC format: 2026-05-08T04:02:18.821Z\nUser time zone: Asia/Shanghai, UTC+8:00\n\n# Current Cost\n$0.00\n\n# Current Mode\ncode\n💻 Code\ngemma4:31b\nYou have not created a todo list yet. Create one with update_todo_list if your task is complicated or involves multiple steps.\n</environment_details>"
}
],
"ts": 1778212938823
}
]
}

[PlanBernhard]

Dear @roomote

I regonized that there are security critical details in the log file which should NEVER be shared. This is in my opinion a CRITICAL second issue WHICH SHOULD BE ADDRESSED.

About your questions:

1. No it just happened this time
2. Yes.
3. Please tell me which part of the log you require to see WITHOUT leaking sensitive data.
4. The custom modes work correctly today.

Thanks

[roomote-v0]

Hi @PlanBernhard, thanks for flagging the security concern about sensitive data in log files. That's a valid and important point -- diagnostic/error files should redact sensitive information like API keys before being shared.

I'll investigate how the error log files are generated and look into adding redaction of sensitive data (API keys, tokens, etc.) from diagnostic output. I'll open a PR to share progress.

Regarding the original bug: since it was a one-time occurrence and custom modes are working correctly now, it was likely a transient API issue.

[roomote-v0]

Opened a draft PR with my attempt at addressing the sensitive data concern: #12290

The PR adds automatic redaction of API keys, tokens, and secrets from diagnostics files before they are written to disk. The diagnostics file header now also informs users that redaction has been applied, while still advising manual review. Feedback welcome!

[PlanBernhard]

@roomote I removed everything that had sensitive data in it. Please tell me when I missed something or when you need some specific information:

// Please share this file with Roo Code Support (support@roocode.com) to diagnose the issue faster
// Just make sure you're OK sharing the contents of the conversation below.

{
  "error": {
    "timestamp": "2026-05-07T19:44:24.026Z",
    "version": "3.53.0",
    "provider": "roo",
    "model": "anthropic/claude-opus-4.7",
    "details": "Unexpected API Response: The language model did not provide any assistant messages. This may indicate an issue with the API or the model's output."
  },
  "history": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "REMOVED"
        },
        {
          "type": "text",
          "text": "REMOVED"
        }
      ],
      "ts": 1778182963092
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "tool_use",
          "id": "toolu_015pKJX8Bqm6KvJNw3z3LxfX",
          "name": "read_file",
          "input": {
            "path": "plans/custom_modes.yaml",
            "mode": "slice",
            "offset": 1,
            "limit": 200
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_01PEAm5j7ysbHUa1UVj9uxCS",
          "name": "read_file",
          "input": {
            "path": "plans/custom_modes.yaml.bk",
            "mode": "slice",
            "offset": 1,
            "limit": 300
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_015Y2ufmA1p3myZtmSBPq9vs",
          "name": "read_file",
          "input": {
            "path": ".othersoftware/tasks/modes-yaml-9f2a/requirements.md",
            "mode": "slice",
            "offset": 1,
            "limit": 100
          }
        }
      ],
      "ts": 1778182971534,
      "reasoning_details": [
        {
          "type": "reasoning.text",
          "text": "Let me analyze the task. I need to create a consolidation plan for `plans/custom_modes.yaml`. First, I need to read the relevant files to understand the current state.\n\nLet me read:\n1. `plans/custom_modes.yaml` - current file\n2. `plans/custom_modes.yaml.bk` - backup template\n3. `.othersoftware/tasks/modes-yaml-9f2a/requirements.md` - requirements\n4. `tasks/modes-yaml-9f2a/mode-profile-overview.md` - if it exists\n\nNo skill applies here - this is a planning task for creating a consolidation plan document.",
          "format": "anthropic-claude-v1",
          "signature": "REMOVED",
          "index": 0
        }
      ]
    },
    {
      "role": "user",
      "content": [
        {
          "type": "tool_result",
          "tool_use_id": "toolu_015pKJX8Bqm6KvJNw3z3LxfX",
          "content": "REMOVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_01PEAm5j7ysbHUa1UVj9uxCS",
          "content": "REMOVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_015Y2ufmA1p3myZtmSBPq9vs",
          "content": "REMOVED"
        },
        {
          "type": "text",
          "text": "REMOVED"
        }
      ],
      "ts": 1778182971548
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "tool_use",
          "id": "toolu_01TQsvz33t1ASVFnddUEhZUe",
          "name": "read_file",
          "input": {
            "path": "plans/custom_modes.yaml.bk",
            "mode": "slice",
            "offset": 301,
            "limit": 255
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_01YWNqTEG1QM9aaD39trCCv4",
          "name": "read_file",
          "input": {
            "path": "plans/custom_modes.yaml",
            "mode": "slice",
            "offset": 201,
            "limit": 101
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_016HYaX6AHEFpkxiZHZVxpdX",
          "name": "list_files",
          "input": {
            "path": "tasks/modes-yaml-9f2a",
            "recursive": false
          }
        }
      ],
      "ts": 1778182978254,
      "reasoning_details": [
        {
          "type": "reasoning.text",
          "text": "REMOVED",
          "format": "anthropic-claude-v1",
          "signature": "REMOVED",
          "index": 0
        }
      ]
    },
    {
      "role": "user",
      "content": [
        {
          "type": "tool_result",
          "tool_use_id": "toolu_01TQsvz33t1ASVFnddUEhZUe",
          "content": "REMOVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_01YWNqTEG1QM9aaD39trCCv4",
          "content": "REMOVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_016HYaX6AHEFpkxiZHZVxpdX",
          "content": "mode-profile-overview.md\nplan.md"
        },
        {
          "type": "text",
          "text": "REMOVED"
        }
      ],
      "ts": 1778182978268
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "tool_use",
          "id": "toolu_01DxVGMVXbjCPGEw5BNmWpEk",
          "name": "read_file",
          "input": {
            "path": "tasks/modes-yaml-9f2a/mode-profile-overview.md",
            "mode": "slice",
            "offset": 1,
            "limit": 120
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_0163CnxdBQXyLoXpyNq1YSrW",
          "name": "read_file",
          "input": {
            "path": "tasks/modes-yaml-9f2a/plan.md",
            "mode": "slice",
            "offset": 1,
            "limit": 80
          }
        },
        {
          "type": "tool_use",
          "id": "toolu_01SgxDtxTP4BLHpNvSyHjiRL",
          "name": "list_files",
          "input": {
            "path": ".roo",
            "recursive": false
          }
        }
      ],
      "ts": 1778182984640,
      "reasoning_details": [
        {
          "type": "reasoning.text",
          "text": "REMOVED",
          "index": 0
        }
      ]
    },
    {
      "role": "user",
      "content": [
        {
          "type": "tool_result",
          "tool_use_id": "toolu_01DxVGMVXbjCPGEw5BNmWpEk",
          "content": "REMVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_0163CnxdBQXyLoXpyNq1YSrW",
          "content": "REMOVED"
        },
        {
          "type": "tool_result",
          "tool_use_id": "toolu_01SgxDtxTP4BLHpNvSyHjiRL",
          "content": "REMOVED"
        },
        {
          "type": "text",
          "text": "<user_message>\nDate/time: 2026-05-07T19:43:10.092Z\nExtension version: 3.53.0\nProvider: roo\nModel: anthropic/claude-sonnet-4.6\n\nUnexpected API Response: The language model did not provide any assistant messages. This may indicate an issue with the API or the model's output.\n</user_message>"
        },
        {
          "type": "text",
          "text": "REMOVED"
        }
      ],
      "ts": 1778183051440
    }
  ]
}

[roomote-v0]

I've seen the debug file you shared -- thanks for providing that context. I've traced through the code and found a bug in the manual retry path for empty API responses that could cause cascading failures. I'll open a PR with my attempt at addressing this. Feedback welcome!

[roomote-v0]

Opened a draft PR with my attempt at addressing this: #12291

Found a bug in the manual retry path -- when user clicks "Retry" after an empty API response, the userMessageWasRemoved flag was not set, causing the retry to skip re-adding the user message to conversation history. This could lead to cascading empty responses. Also added diagnostic logging to help trace empty response causes. Feedback welcome!

[PlanBernhard]

@roomote I just got this error. Maybe it is related?

// Please share this file with Roo Code Support (support@roocode.com) to diagnose the issue faster
// Just make sure you're OK sharing the contents of the conversation below.

{
  "error": {
    "timestamp": "2026-05-08T16:21:15.221Z",
    "version": "3.53.0",
    "provider": "roo",
    "model": "anthropic/claude-sonnet-4.6",
    "details": "This may indicate a failure in the model's thought process or inability to use a tool properly, which can be mitigated with some user guidance (e.g. \"Try breaking down the task into smaller steps\")."
  },
}

[roomote-v0]

@PlanBernhard I've seen your comment -- thanks for sharing the diagnostic. Your error is a different (but related) path from the original issue.

Your error details say "This may indicate a failure in the model's thought process or inability to use a tool properly" -- this is the mistake_limit_reached error, triggered when the model hits the consecutive mistake limit (e.g., from repeated tool errors or failing to use tools). The original issue is about completely empty API responses (no assistant messages at all).

These can be causally linked: if an empty response triggers a retry that corrupts the conversation state (the bug addressed in PR #12291), the model may then fail to use tools properly, cascading into the mistake limit.

I'm opening a follow-up PR that:

1. Includes the userMessageWasRemoved fix from PR fix: add missing userMessageWasRemoved flag in manual retry for empty API responses #12291 (prevents retry state corruption)
2. Resets the consecutiveNoToolUseCount and consecutiveNoAssistantMessagesCount counters when the mistake limit dialog is shown (prevents frustrating retry loops after user provides guidance)

Will share progress shortly.