qwen 3 coder file edits are appended to first line of existing code

[gompa]

when running nanocoder with the ollama provider and with the qwen3 coder model loaded.
having the model edit files, the "edits" are appended to the first line of the file, leaving the original code after the pasted "edits"

[will-lamerton]

Hey @gompa - thanks for adding this issue. I will explore the problem and come back to you. A first knee-jerk reaction is that it's the fact that Qwen3 Coder when run through Ollama has issues with tool calling but, I will explore and post this thread :)

[will-lamerton]

Hey @gompa - thanks again for submitting this issue. Can you confirm with me a couple of things to further test this?

• Which version of Qwen 3 Coder are you using?
• Which provider are you using it through?
• System details?

[gompa]

hey, its on ollama v0.12.1 and the model: qwen3-coder:30b
system is running on debian trixie

[will-lamerton]

Hey @gompa - looking at this issue for the next release coming over the next few days. Can you confirm this is still an issue. Only reason I ask is that there have been a lot of known issues with Qwen3 Coder 30b locally when running through Ollama. For a long time it didn't support tools properly.

However, they updated the model on Ollama a week ago and testing myself, things seem okay. Please let me know from your end and if you're experiencing issues, any screenshots would be immensely helpful :)

You may have to remove and re-pull the latest model version.

Thanks!

Will