[BUG] Tool calling error with latest changes - Cerebras

[brianboysen51]

Problem (one or two sentences)

Seeing an issue since the recent refactor on tool calling. Here is the specific issue I see with Cerebras:

Date/time: 2025-12-24T20:17:23.143Z
Extension version: 3.37.1
Provider: cerebras
Model: zai-glm-4.6

Cerebras completion error: Cerebras API Error (422): Tools with mixed values for 'strict' are not allowed. Please set all tools to 'strict: true' or 'strict: false'

Context (who is affected and when)

All users if using Cerebras on version 3.37.1 (and potentially earlier)

Reproduction steps

1. Using cerebras provider (I am using my code plan)
2. send a message to a mode with cerebras provider selected

Expected result

Valid api calls (not having misaligned tool schemas)

Actual result

Above API error blocking even basic requests

Variations tried (optional)

No response

App Version

3.37.1

API Provider (optional)

Not Applicable / Other

Model Used (optional)

zai-glm-4.6

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

// 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": "2025-12-24T21:59:36.035Z",
    "version": "3.37.1",
    "provider": "cerebras",
    "model": "zai-glm-4.6",
    "details": "Cerebras completion error: Cerebras API Error (422): Tools with mixed values for 'strict' are not allowed. Please set all tools to 'strict: true' or 'strict: false'"
  },
  "history": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "<task>\nhi\n</task>"
        },
        {
          "type": "text",
          "text": "<environment_details>\n# VSCode Visible Files\n\n\n# VSCode Open Tabs\nproject.md\n\n# Current Time\nCurrent time in ISO 8601 UTC format: 2025-12-24T20:16:51.392Z\nUser time zone: America/Chicago, UTC-6:00\n\n# Current Cost\n$0.00\n\n# Current Mode\n<slug>code</slug>\n<name>💻 Code</name>\n<model>glm-4.7</model>\n<tool_format>native</tool_format>\n\n\n# Current Workspace Directory (c:/Users/brian/Downloads/db_cs_handshake) Files\ndatabricks_auth.py\ndatabricks_cli.py\ndatabricks_kernel.py\nproject.md\nREADME.md\nrequirements.txt\nsetup.py\n.vscode/\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": 1766607411443
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "reasoning",
          "text": "The user just said \"hi\". This is a simple greeting. I should respond in a friendly but direct way, as per my instructions to not be conversational but rather direct and to the point. I should acknowledge their greeting and perhaps ask if there's something specific they'd like help with in their project.\n\nLooking at the environment details, I can see they have a project in the workspace directory `c:/Users/brian/Downloads/db_cs_handshake` with several Python files related to Databricks (auth, cli, kernel), plus project documentation files. This appears to be some kind of Databricks integration project.\n\nI'll respond briefly and ask if they need help with anything specific.",
          "summary": []
        },
        {
          "type": "text",
          "text": "Hi! I can see you have a Databricks-related project with Python files for authentication, CLI, and kernel functionality. How can I help you today?"
        }
      ],
      "ts": 1766607415393
    },
    {
      "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\nproject.md\n\n# Current Time\nCurrent time in ISO 8601 UTC format: 2025-12-24T20:17:20.128Z\nUser time zone: America/Chicago, UTC-6:00\n\n# Current Cost\n$0.01\n\n# Current Mode\n<slug>code</slug>\n<name>💻 Code</name>\n<model>zai-glm-4.6</model>\n<tool_format>native</tool_format>\n\n\n# Current Workspace Directory (c:/Users/brian/Downloads/db_cs_handshake) Files\ndatabricks_auth.py\ndatabricks_cli.py\ndatabricks_kernel.py\nproject.md\nREADME.md\nrequirements.txt\nsetup.py\n.vscode/\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": 1766607440169
    }
  ]
}

[roomote]

I've analyzed this issue and here's my proposed implementation plan:

Root Cause:
The error occurs because the base convertToolsForOpenAI method in base-provider.ts sets strict: true for regular tools and strict: false for MCP tools (to preserve optional parameters). When a user has both types of tools, Cerebras receives mixed strict values, which it doesn't allow.

Proposed Fix:
Override the convertToolsForOpenAI method in the CerebrasHandler class to ensure all tools have consistent strict: false values:

1. Add a convertToolsForOpenAI override in src/api/providers/cerebras.ts
2. This override will call the base method, then set strict: false for all tools to ensure consistency
3. Add a test case in cerebras.spec.ts to verify mixed tools get consistent strict values

Why strict: false:

• MCP tools cannot use strict mode because they have optional parameters from the MCP server schema
• Setting all tools to strict: false ensures consistency with the Cerebras API requirement
• The Cerebras handler already has a custom convertToolSchemaForOpenAI that strips unsupported fields, so non-strict mode is appropriate

Please mention @roomote in a comment saying that you approve this plan, and I'll proceed with the implementation.

[jqknono]

https://blog.roocode.com/p/sorry-we-didnt-listen-sooner-native

It looks like Roo Code has chosen to lose some users or vendors.