fix(ollama): fix ollama container for CPU vs GPU mode

[waleedlatif1]

Summary

fix ollama container for CPU vs GPU modem, there was a health check for ollama, which checked the GPU profile and therefore required that the user had a GPU

Fixes #1390 #1315

Type of Change

• Bug fix

Testing

Tested by running manually.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
docs	Ready	Preview	Comment	Sep 20, 2025 6:43pm
sim	Ready	Preview	Comment	Sep 20, 2025 6:43pm

[greptile-apps]

Greptile Summary

This PR fixes critical Docker Compose configuration issues that prevented Ollama from running in CPU-only environments. The main problem was that the simstudio service had a hard dependency on the ollama service being healthy, but when using the --profile cpu flag, only the ollama-cpu service would start, causing dependency resolution failures.

The changes address several interconnected issues:

1. Profile Segregation: Added the gpu profile to the main ollama service (lines 100-101), ensuring proper separation between GPU and CPU deployment modes

2. Dependency Management: Removed the hard dependency on ollama from the simstudio service (lines 25-31), eliminating the deployment failure when running in CPU mode

3. Health Check Improvements: Replaced unreliable HTTP-based health checks with ollama list commands (lines 123, 147), providing more robust service health detection

4. Service Discovery: Added network aliases to the ollama-cpu service (lines 153-156), ensuring proper service discovery and communication in CPU-only deployments

5. Startup Logic: Refactored the model-setup service from hard dependencies to polling-based startup logic (lines 167-171), eliminating race conditions during container initialization

These changes ensure that the Docker Compose configuration works seamlessly in both GPU-enabled and CPU-only environments while maintaining the same functional behavior. The fix resolves issues #1390 and #1315 related to deployment failures on systems without GPU support.

PR Description Notes:

• Minor typo: "CPU vs GPU modem" should be "CPU vs GPU mode"

Confidence score: 4/5

• This PR addresses well-defined deployment issues with targeted fixes that resolve CPU-only environment failures
• Score reflects solid understanding of Docker Compose profiles and service dependencies, with changes that directly address the root causes
• Pay close attention to the dependency removal in simstudio service and the new polling logic in model-setup service

_{1 file reviewed, no comments}

_{Edit Code Review Bot Settings | Greptile}