# Adding New Models to the Registry
This document explains how to add new AI models to the MCP server's model registry.
## Model Configuration Structure
Each model is defined by a JSON configuration file in `./image_gen_mcp/resources/models/`. The file should be named `{model_id}.json` where `model_id` is the unique identifier for the model.
## Configuration Fields
```json
{
"model_id": "unique-model-identifier",
"name": "Human-readable Model Name",
"version": "1.0",
"capabilities": [
"List of model capabilities",
"Each capability as a string"
],
"pricing": {
"pricing_type": "price description",
"another_pricing_type": "another price description"
},
"rate_limits": {
"default": "rate limit description",
"configurable": "configuration info"
},
"size_options": [
"1024x1024",
"1536x1024 (landscape)"
],
"quality_levels": [
"auto",
"high",
"medium",
"low"
],
"formats": [
"PNG",
"JPEG",
"WebP"
],
"features": {
"feature_name": "feature description",
"another_feature": "another description"
},
"best_practices": [
"Best practice recommendations",
"Usage guidelines"
],
"examples": [
"Example usage patterns",
"Sample API calls"
]
}
```
## Adding a New Model
1. Create a new JSON file in `./image_gen_mcp/resources/models/`
2. Name it `{model_id}.json` (e.g., `dalle-3.json`)
3. Fill in all required fields
4. The model will be automatically available via the resource URIs
## Resource URIs
Once configured, models are accessible via:
- `model-info://{model_id}` - Get detailed documentation for a specific model
- `models://list` - List all available models
## Examples
### GPT-Image-1 (existing)
```
model-info://gpt-image-1
```
### DALL-E 3 (example)
```
model-info://dalle-3
```
### List All Models
```
models://list
```
## Custom Documentation
For more detailed documentation, you can also create a `{model_id}.md` file alongside the JSON configuration. This will be used instead of the auto-generated documentation.
## Programmatic Registration
You can also register models programmatically using the `model_registry`:
```python
from image_gen_mcp.resources.model_registry import model_registry, ModelInfo
new_model = ModelInfo(
model_id="my-model",
name="My Model",
version="1.0",
capabilities=["capability1", "capability2"],
pricing={"generation": "$0.01 per image"},
rate_limits={"default": "100 requests per minute"},
size_options=["1024x1024"],
quality_levels=["standard", "high"],
formats=["PNG"],
features={"feature1": "description"},
best_practices=["practice1", "practice2"],
examples=["example1", "example2"]
)
model_registry.register_model(new_model)
```
