When an upstream LLM provider (in our case LM Studio + Qwen3) returns a Jinja prompt-template rendering error such as: Error rendering prompt with jinja template: "No user query found in messages." OpenClaw's context-overflow-diag classifies it as a context overflow and tells the user to Try /reset. Resetting and replaying the same startup prompt triggers the exact same template error again → user-visible infinite loop. error_message=Error rendering prompt with jinja template: "No user query found in messages." context-overflow-diag should recognize provider responses matching Error rendering prompt with jinja template (and similar template/serialization errors) and classify them as provider-template-error (or similar). The user-facing message should be something like: The classifier seems to fall through to overflow whenever it can't compute a token count, regardless of the underlying error string. Replacing the model's Jinja chat template with a role-trusting ChatML template (does not parse message content, just emits <|im_start|>{role}\n{content}<|im_end|>\n per message) makes the error go away. But the diagnostic misclassification on OpenClaw's side is independent and should still be fixed.

TL;DR

The context-overflow-diag classifier in OpenClaw should be updated to recognize and correctly classify provider responses with Jinja template rendering errors as provider-template-error instead of context overflow.

Guidance

• Review the context-overflow-diag classifier logic to ensure it checks for specific error messages like "Error rendering prompt with jinja template" and classifies them accordingly.
• Update the classifier to handle cases where observedTokens is unknown and compactionAttempts is 0, to prevent misclassification as a context overflow.
• Consider adding a new error type provider-template-error to handle such cases and provide a more informative user-facing message.
• Verify the fix by testing with the provided repro steps and checking the gateway log for the correct error classification and user message.

Example

No code snippet is provided as the issue is more related to the logic of the classifier rather than a specific code snippet.

Notes

The current workaround of replacing the Jinja chat template with a ChatML template may resolve the error but does not address the underlying issue of misclassification by the context-overflow-diag classifier.

Recommendation

Apply a workaround by updating the context-overflow-diag classifier to correctly handle Jinja template rendering errors, as this will provide a more accurate and helpful user experience.