Creates a new AI model configuration
Initializes a model configuration with the specified parameters and settings. The model configuration defines how the AI model behaves and interacts with the platform.
Configuration data including: - name: Display name (required, unique among public models) - model: Upstream model identifier (e.g. 'gpt-4o', 'claude-3-5-sonnet-20241022') - description: One-line description (required) - api_type: Provider protocol (e.g. 'openai', 'anthropic') - base_url: Provider API base URL (required, encrypted) - api_key: Provider API key (required, encrypted) - prices: Pricing map (values must be numbers) - type: Model category (chat | image | rerank | moderation | language | embedding | contextualized_embedding) - max_context_tokens / max_output_tokens: Token limits
Optionaluri: stringOptional custom URI path for the model endpoint
Basic configuration:
const model = new Model({
name: 'support-gpt',
model: 'gpt-4o',
description: 'Support model fine-tuned on Zendesk tickets.',
api_type: 'openai',
base_url: 'https://api.openai.com/v1',
api_key: process.env.OPENAI_API_KEY!,
prices: { input_per_1k: 0.005, output_per_1k: 0.015 }
});
Advanced configuration:
const model = new Model({
name: 'enterprise-claude',
model: 'claude-3-5-sonnet-20241022',
description: 'Enterprise documentation assistant.',
api_type: 'anthropic',
base_url: 'https://api.anthropic.com/v1',
api_key: process.env.ANTHROPIC_API_KEY!,
prices: { input_per_1k: 0.003, output_per_1k: 0.015 },
type: 'chat',
max_context_tokens: 200000,
max_output_tokens: 8192,
encoding: 'cl100k_base'
}, '/enterprise/model');
Get the chat functionality for this model
This getter provides access to the model's chat capabilities through the Chat class. It enables direct interaction with the model for text generation and completion tasks.
A new Chat instance configured for this model
Basic chat:
const response = await model.chat.completions.create({
messages: [
{ role: 'user', content: 'What is machine learning?' }
]
});
Advanced chat with system prompt:
const response = await model.chat.completions.create({
messages: [
{
role: 'system',
content: 'You are an expert in artificial intelligence.'
},
{
role: 'user',
content: 'Explain neural networks to a beginner.'
}
],
temperature: 0.3,
max_tokens: 1000,
stream: false
});
console.log('Model explanation:', response.choices[0].message.content);
Check if the entity is active
This method checks the active status of the entity. Most entities in the system can be active or inactive, which affects their availability and usability in the platform.
True if the entity is active, false otherwise
Convert model instance to interface data
This method serializes the model instance to a plain object that matches the interface type. This is useful for:
The model data as a plain object matching the interface type
Convert model instance to API payload
This method creates a payload suitable for API requests by:
A clean object suitable for API requests
Update model data with new values
This method updates the model's data and instance properties with new values. It performs a shallow merge of the updates with existing data, allowing for partial updates of the model's properties.
Object containing properties to update
const user = new User({
email: 'old@example.com',
firstName: 'John'
});
// Update multiple properties
user.update({
email: 'new@example.com',
lastName: 'Doe'
});
// Save changes to API
await user.save();
This method only updates the local model instance. To persist changes to the API, call save after updating.
Save the model instance to the API
This method persists the current state of the model to the API using a PUT request. It requires the model to have an ID (existing instance). For new instances, use the collection's create method instead.
The method:
Promise resolving to the updated model data
Delete the model instance from the API
This method permanently deletes the model instance from the API and clears the local data. This operation cannot be undone.
The method:
Optionalparams: objectOptional query parameters for deletion (e.g., { delete: true, deleteS3: true })
Promise that resolves when deletion is successful
Basic deletion:
const user = await users.get({}, 'user-id');
if (user) {
await user.delete();
// User is now deleted and instance is cleared
}
Subscribe to live updates for this entity via the SSE relay.
Opens a Server-Sent Events connection to the Mosaia relay server
(mosaia-sse-relay-server) scoped to this specific document. The
relay watches the underlying MongoDB change stream and forwards every
update to handlers.onUpdate.
Requires the entity to have an id (must be saved first).
Lifecycle callbacks (onStart, onUpdate, onError, onClose).
A subscription handle. Call close() to disconnect.
Rerank documents using this model
Uses this model to rerank documents based on their relevance to a query string.
Rerank request parameters
Rerank request interface
Promise resolving to the rerank response
Generate embeddings using this model
Generates vector embeddings for text input using this model.
Embedding request parameters
Embedding request interface
Promise resolving to the embedding response
Like or unlike this model
Toggles the like status of this model. If the model is already liked, it will be unliked, and vice versa.
Promise that resolves to the updated model instance
Model class for managing AI model configurations
This class represents an AI model configuration in the Mosaia platform. It provides a structured way to define, customize, and interact with various AI models through a unified interface.
Features:
Remarks
Models are the core AI engines that power:
The class supports various model providers and configurations:
Example
Basic model setup:
Example
Using chat capabilities: