feat: expose richer model metadata in v1/models

[VedantMadane]

Fixes #143.

This PR expands the model schema returned by the /v1/models\ endpoint to include additional metadata about each model, similar to what other providers like OpenRouter expose.

New fields added to OpenAIModel:

• \context_window: Maximum context length supported by the model.
• \�rchitecture: Technical architecture of the model.
• \parameters: Description of model parameters.
• \quantization: Quantization method used.
• \size: Parameter size (e.g., " 8B).

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[gemini-code-assist]

Code Review

This pull request enhances the /v1/models endpoint by including additional metadata for each model, such as context window size and architecture. The changes look good, but I've found a critical issue where the code could panic if a model's configuration is missing. I've provided a suggestion to handle this case gracefully.

[gemini-code-assist]

The m.Config() method can return a nil config with a nil error. If this happens, the subsequent calls to methods on config (e.g., config.GetContextSize() on line 91) will cause a nil pointer dereference, leading to a panic. You should add a check to handle the case where config is nil.

 	config, err := m.Config()
 	if err != nil {
 		return nil, fmt.Errorf("get config: %w", err)
 	}

 	if config == nil {
 		// If there's no config, return the model with only basic info.
 		return &OpenAIModel{
 			ID:      id,
 			Object:  "model",
 			Created: created,
 			OwnedBy: "docker",
 		}, nil
 	}

 	var contextWindow int32
 	if cw := config.GetContextSize(); cw != nil {
 		contextWindow = *cw
 	}

 	return &OpenAIModel{
 		ID:            id,
 		Object:        "model",
 		Created:       created,
 		OwnedBy:       "docker",
 		ContextWindow: contextWindow,
 		Architecture:  config.GetArchitecture(),
 		Parameters:    config.GetParameters(),
 		Quantization:  config.GetQuantization(),
 		Size:          config.GetSize(),
 	}, nil

[doringeman]

Hey @VedantMadane, why do you need this?
This API you've modified corresponds with the official List models.
Isn't /models what you're looking for? E.g., curl -s http://localhost:12434/models | jq .

[VedantMadane]

Hi @doringeman, good question! The /engines/llama.cpp/v1/models endpoint (OpenAI-compatible List Models) currently returns only id and object fields, while the /models endpoint returns more details but in a non-standard format.

This PR enriches the OpenAI-compatible response with metadata fields like architecture, parameters, quantization, context_window, and size — so consumers using the standard OpenAI SDK can access model capabilities without needing a separate non-standard call.

The idea is to keep the interface OpenAI-compatible while exposing the richer metadata that Docker Model Runner already has.

[ericcurtin]

Missing unit tests for new fields, ContextWindow should be *int32 for proper omitempty behavior.

[doringeman]

Hey @VedantMadane , WDYT about adding all the new extra fields under a common field called "dmr"?
E.g.,

{
  "id": "VAR_chat_model_id",
  "object": "model",
  "created": 1686935002,
  "owned_by": "openai",
  "dmr": {
     "custom-field":"custom-vaalue"
   }
}

[doringeman]

LGTM, thanks! 🙌