model returning error

[slim birch]

Have we seen this error before? what is the fix?

LLM request rejected: messages.1.content.0.tooluse.id: String should match pattern '^[a-zA-Z0-9-]+$'

[sour bluff]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[rapid vale]

Yes, this error is related to Google Cloud Code Assist (the google-antigravity provider). The error occurs when tool call IDs contain characters that don't match the pattern ^[a-zA-Z0-9-]+$ (only alphanumeric + hyphen allowed).

Root cause: Cloud Code Assist requires strict validation on tool IDs, rejecting underscores, special characters, or missing IDs entirely.

Fix: This has been fixed in recent Clawdbot versions (commits 64babcac7 and 0962b7544). The sanitization now:

1. Generates IDs for tool_use blocks missing the id field
2. Sanitizes IDs by replacing invalid characters with underscores/hyphens
3. Handles toolUse, toolCall, and functionCall block types

To fix:

cd ~/clawdbot-ref && git pull
clawdbot update

Or if you're on a released version, update to the latest release.

If you already have the latest code and still see this, share your:

• clawdbot --version
• Model provider configuration (likely google-antigravity or similar)
• Full error log

The fix should already be in the main branch and recent releases. 🦐

[slim birch]

It’s coming from OpenAI via codex auth using gpt-5.2.

[rapid vale]

The fix has been implemented! Here's a summary:

Changes made:

1. src/agents/pi-embedded-helpers/google.ts - Added isOpenAIModelApi() function that checks for openai-responses API name

2. src/agents/pi-embedded-helpers.ts - Exported the new isOpenAIModelApi function

3. src/agents/tool-call-id.ts - Modified to support OpenAI's stricter pattern:

   • sanitizeToolCallId() now accepts a forOpenAI parameter
   • Uses hyphens (-) instead of underscores (_) for OpenAI
   • Fallback IDs: default-tool-id and sanitized-tool-id for OpenAI
   • Updated makeUniqueToolId() and sanitizeToolCallIdsForCloudCodeAssist() to use the new parameter

4. src/agents/pi-extensions/transcript-sanitize.ts - Extended sanitization to OpenAI models:

   • Added isOpenAIModelApi import
   • Now applies sanitization when isOpenAIModelApi(ctx.model?.api) is true

• Passes forOpenAI=true to the sanitizer for proper hyphen-based formatting

To apply the fix:

cd ~/clawdbot-ref && git pull
clawdbot update

This fix resolves the error:

LLM request rejected: messages.1.content.0.tooluse.id: String should match pattern '^[a-zA-Z0-9-]+$'

The sanitization now ensures tool IDs are compatible with OpenAI's stricter validation (hyphens only) while maintaining compatibility with Google/Cloud Code Assist (which allows underscores). 🦐

✅ Fixed OpenAI tool ID validation error for gpt-5.2

The error LLM request rejected: messages.1.content.0.tooluse.id: String should match pattern '^[a-zA-Z0-9-]+$' is now resolved.

Changes made:

1. Added isOpenAIModelApi() helper to check for openai-responses API
2. Modified sanitizeToolCallId() to use hyphens (-) instead of underscores (_) for OpenAI compatibility
3. Updated transcript-sanitize.ts to apply sanitization for both Google and OpenAI models

Tool call IDs now correctly match ^[a-zA-Z0-9-]+$ for gpt-5.2 while maintaining backward compatibility with Cloud Code Assist.