Use Ollama running remote server

[myrulezzz]

How can i use ollama running on remote server? Is there any way i can setup base url for ollama?

[saeed-siddiqui]

Yes you can run remotely and you can setup http tunnel localhost to the remote port.

[myrulezzz]

How can i set it to run remotely?

[smileyboy2019]

I don't quite understand, is the URL of Olllam configured in the configuration file or do we need to modify the source code

[ImayaBharathi]

The feature i want to request/ would be able to contribute is to add ollama support for loading Models. Since a thread is already open for ollama, thought of asking here. @PromtEngineer let me know if there is scope to add ollama to this project, i can contribute for it as well.

[PromtEngineer]

@ImayaBharathi Yes that will be extremely helpful and would really appreciate the contribution here.

[ImayaBharathi]

@PromtEngineer awesome, could you assign this issue to me. i will finish and raise a PR

[PromtEngineer]

@ImayaBharathi done.

[ImayaBharathi]

@PromtEngineer I went through the project and here is my understanding
Currently, the project supports local Ollama instances (localhost:11434) but doesn't allow connecting to remote Ollama servers. This addition would be valuable for users who:

• Run Ollama on dedicated GPU machines
• Use shared team Ollama instances
• Deploy Ollama in cloud/server environments

Proposed Changes:

1. Add remote URL configuration via:
   • Environment variable: OLLAMA_HOST
   • Settings page in web interface
2. Enhance error handling for remote connections
3. Add connection status checking and feedback

The implementation will:

• Be minimally invasive to existing code
• Maintain backwards compatibility (defaults to localhost)
• Follow project's error handling patterns
• Add clear documentation

Sample usage would be:

# Via environment variable
export OLLAMA_HOST='http://your-server:11434'
python app.py

# Or configure through web interface in Settings page

does this look good to you, if yes respond with yes then i would go ahead with the implementation. if not could you provide your inputs (If you’re short on time, even minimal feedback would be helpful.)