Unable to generate code via Ollma qwen2.5-coder

[vishnu-kantliwala]

Describe the bug

The app generates a basic structure in the workspace.
When given an instruction to add a specific feature, it returns the code in the response instead of modifying the workspace.

Link to the Bolt URL that caused the error

https://github.com/stackblitz-labs/bolt.diy

Steps to reproduce

1. Generate a basic structure in the workspace.
2. Provide an instruction to add a specific feature.
3. Observe that the app returns code in the response instead of modifying files in the workspace.

Expected behavior

The app should modify the workspace directly when instructed to add a specific feature.

Screen Recording / Screenshot

Bolt-new-demo.webm

Platform

• OS: Ubuntu 22.04.5 LTS (Jammy)
• AI Model Used: qwen2.5-coder

Provider Used

ollama

Model Used

qwen2.5-coder

Additional context

Ensure that when a feature is requested, the code is directly added/modified within the workspace instead of being returned in the response.

Are there any specific logs/errors related to this issue?
Is this behavior consistent for all features or only for certain ones?

[Anybody2007]

Hi @vishnu-kantliwala ,
I am curious to know what are the parameters and all you used to make the
qwen2.5-coder-bolt-2:latest.

[vishnu-kantliwala]

Hi @vishnu-kantliwala , I am curious to know what are the parameters and all you used to make the qwen2.5-coder-bolt-2:latest.

Here is the Modelfile.
FROM qwen2.5-coder PARAMETER num_ctx 32768