I'm using vscode, new to continue and ollama, can't get it to write files, it just tells me it's an llm and can't write files.

[chrisdraws]

Hello,

I set up my system based on this tutorial -> https://dev.to/anita_ihuman/best-offline-ai-coding-assistant-how-to-run-llms-locally-without-internet-2bah

Very basic... finally got the continue.dev with ollama and qwen to respond to me and it will tell me what to do, but I when I try to tell it to write files it tells me it can't.. how can I get this to write files?

[aniruddhaadak80]

From my point of view, this usually means there is a capability mismatch between chat mode and agent or edit mode rather than Ollama itself refusing file writes. If the model can only chat, it will describe the change instead of exercising the editor integration that actually writes to disk.

The product could make this much clearer by surfacing whether the current session has edit tools enabled, because new users naturally read assistant capability problems as model problems.

[chrisdraws]

Thank you for the reply, I finally gave up and just uninstalled this and will maybe come back to it in a few months. Not only was it not able to write files, auto complete also was not working. Also everything was incredibly slow both on my local computer and in an Amazon EC2 instance which while not having a GPU had tons of CPU and ram and should not have been so slow. This would have been incredibly beneficial because I can't use a service like curosr or claude that are using LLMs and services that I don't control.

[omni-front]

You need to configure the model to have the appropriate roles for file writing:

models:
name: qwen2.5-coder
    provider: ollama
    model: qwen2.5-coder:1.5b-base
    apiBase: http://localhost:11434
    roles:
autocomplete
write_files

Adding the write_files role should enable the model to perform file writing tasks.

[privitan-projects]

But where? How?

[omni-front]

You'll need to add that configuration to your project's YAML settings file, typically named continue.yaml or similar, located in the root directory of your project. If it doesn't exist, you may need to create it.