Description
I'm using CodeWhale with Ollama and the qwen3-coder:30b model.
I'm experiencing a recurring issue where the agent stops working before the task is actually finished, but CodeWhale still reports the task status as "completed".
Typical behavior:
The agent starts working normally.
It makes some code changes and/or provides partial output or even does nothing in some cases.
At some point it stops progressing.
No obvious error is shown.
The task is marked as completed even though the requested work was only partially done.
A few questions:
- What are the most common reasons for this behavior?
- Is there a way to determine whether the model itself stopped generating, hit a context limit, exceeded an internal timeout, or encountered another failure condition?
- Which logs should I inspect for debugging?
- Is there a verbose/debug mode that shows the exact reason why CodeWhale considers a task completed?
- Are there any known issues or recommended settings when using qwen3-coder:30b via Ollama?
Impact
It happens quite often.
Environment
- OS: MacOS
- codewhale version: latest
- Model/provider: ollama / qwen3-coder:30b
- Terminal app: Tabby
Description
I'm using CodeWhale with Ollama and the qwen3-coder:30b model.
I'm experiencing a recurring issue where the agent stops working before the task is actually finished, but CodeWhale still reports the task status as "completed".
Typical behavior:
The agent starts working normally.
It makes some code changes and/or provides partial output or even does nothing in some cases.
At some point it stops progressing.
No obvious error is shown.
The task is marked as completed even though the requested work was only partially done.
A few questions:
Impact
It happens quite often.
Environment