Fix Ollama fetch errors and improve error handling

[bhouston]

Ollama Fetch Error Improvements

Background

This is a follow-up to issue #401 where a user reported "TypeError: fetch failed" errors when using MyCoder with Ollama.

Proposed Changes

1. Update the Ollama npm package to the latest version (currently using 0.5.14, latest is 0.5.15)
2. Improve error handling in the OllamaProvider class to provide more informative error messages:
   • Add specific error handling for network connectivity issues
   • Provide clearer error messages when the Ollama server is unreachable
   • Handle large model loading failures more gracefully
3. Add a health check for the Ollama server before attempting to use it
4. Add documentation on troubleshooting Ollama connectivity issues

Implementation Plan

1. Update the Ollama dependency in packages/agent/package.json
2. Enhance the OllamaProvider class with better error handling:

   try {
     // Make the API request using the Ollama client
     const response: OllamaChatResponse = await this.client.chat({
       ...requestOptions,
       stream: false,
     });
     // Process response...
   } catch (error) {
     // Enhanced error handling
     if (error.message.includes('fetch failed')) {
       throw new Error('Failed to connect to Ollama server. Please ensure Ollama is running and accessible.');
     }
     // Handle other specific errors...
     throw error;
   }

3. Add a health check method to verify Ollama server status
4. Update documentation with troubleshooting tips

Testing Plan

• Test with Ollama server running/not running
• Test with various network configurations
• Test with different model sizes
• Verify error messages are user-friendly and actionable

Related Issues

• "TypeError: fetch failed" when trying to use ollama #401 "TypeError: fetch failed" when trying to use ollama