Refactor Model Providers to Use Shared Proxy Package + validation for OpenAI like providers

[sanjay920]

This PR consolidates and standardizes the model provider implementations by introducing a shared proxy package. Key changes:

• Created a shared proxy package from OpenAI provider's server code
• Migrated Groq, DeepSeek, xAI, Ollama, and vLLM providers from their individual implementations to use the shared proxy
• Added consistent validation support across all relevant providers. (6 birds, 1 stone WRT Add validation to openi-like model providers obot#1158)
• Removed duplicate server implementations
• Converted Python-based providers (Groq, Ollama) to Go
• Added model filtering support for provider-specific model types

The shared proxy package reduces code duplication, improves maintainability, and ensures consistent behavior across all model providers.

[iwilltry42]

I like the logic consolidation and the refactoring - just have some confusion around the validation - it seems like it can never return an error (but I may be wrong here).

[iwilltry42]

You can use json.NewDecoder to decode right from the body (reader) into the struct without reading it first.

[iwilltry42]

Some nits and a suggestion for validation handling - overall I like the new merged approach 👍

[sanjay920]

The flow is:

1. Tool is invoked with validate argument
2. Config is created with provider-specific settings (host, API key, etc.)
3. Validate() is called which:
   • Makes HTTP request to provider's models endpoint
   • If any error occurs (bad credentials, network error, etc.):
     • Logs the error (for debugging)
     • Prints JSON-formatted error (for tool consumption)
     • Returns error to trigger exit(1)
4. Main program exits with status 1 if validation failed

[thedadams]

nit: so we don't have to process the body twice.

Suggested change

	body, err := io.ReadAll(resp.Body)
	if err != nil {
	return handleValidationError(toolPath, "Invalid Response")
	}
	var modelsResp api.ModelsResponse
	if err := json.Unmarshal(body, &modelsResp); err != nil {
	return handleValidationError(toolPath, "Invalid Response Format")
	}
	var modelsResp api.ModelsResponse
	if err := json.NewDecoder(resp.Body).Decode(&modelsResp); err != nil {
	return handleValidationError(toolPath, "Invalid Response Format")
	}