LLM Model Management

SoCMate uses LLM models for entity extraction, query generation, evidence assessment, and report generation. The platform supports multiple LLM providers and enables hot-swapping models without redeployment.

All model management endpoints require the admin role. Changes to the model configuration take effect immediately for new investigations.

How It Works

SoCMate’s model gateway routes requests to the configured LLM provider. This decouples the platform from any specific provider and allows you to switch models at any time.

Listing Models

View all models currently configured:

curl -X GET https://api.socmate.yourcompany.com/api/admin/models \
  -H "Authorization: Bearer <admin_token>"

Response:

{
  "models": [
    {
      "model_name": "gpt-5.3",
      "model_info": {
        "id": "model_abc123",
        "provider": "azure",
        "model": "gpt-5.3"
      }
    },
    {
      "model_name": "gpt-5-mini",
      "model_info": {
        "id": "model_def456",
        "provider": "azure",
        "model": "gpt-5-mini"
      }
    }
  ],
  "total": 2
}

Adding a Model

Add a new LLM model:

curl -X POST https://api.socmate.yourcompany.com/api/admin/models \
  -H "Authorization: Bearer <admin_token>" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "gpt-5.3",
    "provider": "azure",
    "model_id": "gpt-5.3",
    "api_base": "https://your-openai-resource.openai.azure.com",
    "api_key": "your-azure-openai-api-key",
    "api_version": "2025-01-01-preview"
  }'

Response:

{
  "message": "Model added successfully"
}

Model Configuration Fields

Field	Required	Description
`model_name`	Yes	Display name used in SoCMate (e.g., `gpt-5.3`)
`provider`	Yes	LLM provider: `azure`, `openai`, `anthropic`, etc.
`model_id`	Yes	Provider-specific model identifier
`api_base`	No	Provider API base URL (required for Azure OpenAI)
`api_key`	Yes	API key for the provider
`api_version`	No	API version (required for Azure OpenAI)

Testing a Model

Send a test prompt to verify the model is working:

curl -X POST https://api.socmate.yourcompany.com/api/admin/models/gpt-5.3/test \
  -H "Authorization: Bearer <admin_token>" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Say hello in one word."
  }'

Response:

{
  "model": "gpt-5.3",
  "response": "Hello.",
  "latency_ms": 320,
  "tokens": {
    "prompt": 6,
    "completion": 2,
    "total": 8
  }
}

If no prompt is provided, the default test prompt "Say hello in one word." is used.

Setting the Default Model

The default model is used for all investigation stages unless a specific model is configured per task.

View Current Configuration

curl -X GET https://api.socmate.yourcompany.com/api/admin/models/config \
  -H "Authorization: Bearer <admin_token>"

Response:

{
  "default_model": "gpt-5.3",
  "available_models": ["gpt-5.3", "gpt-5-mini", "kimi-k2.5", "grok-4.1"],
  "updated_at": "2026-03-27T10:00:00Z",
  "updated_by": "admin@example.com"
}

Update Default Model

curl -X PUT https://api.socmate.yourcompany.com/api/admin/models/config \
  -H "Authorization: Bearer <admin_token>" \
  -H "Content-Type: application/json" \
  -d '{
    "default_model": "gpt-5-mini"
  }'

The model name must match an existing model. The change takes effect immediately for new investigations.

Changing the default model affects all new investigations across the platform. Test the model thoroughly before setting it as default. In-progress investigations continue using the model they started with.

Deleting a Model

Remove a model:

curl -X DELETE https://api.socmate.yourcompany.com/api/admin/models/{model_id} \
  -H "Authorization: Bearer <admin_token>"

Do not delete the model that is currently set as the default. Set a different default model first, then delete the old one.

Health Check

Check the model gateway health:

curl -X GET https://api.socmate.yourcompany.com/api/admin/models/health \
  -H "Authorization: Bearer <admin_token>"

Returns the health status of the model gateway and its configured models.

Supported Providers

SoCMate supports multiple LLM providers:

Provider	Configuration Notes
Azure OpenAI	Requires `api_base` and `api_version`. Most common for enterprise deployments.
OpenAI	Standard OpenAI API. Requires `api_key` only.
Anthropic	Claude models. Requires `api_key` only.
Google	Gemini models via Vertex AI or AI Studio.
Custom	Any OpenAI-compatible API endpoint via `api_base`.

Best Practices

Test before defaulting — Always run the test endpoint on a new model before setting it as the platform default
Monitor quality — Different models produce different quality queries and reports; evaluate investigation output after switching models
Keep a fallback — Maintain at least two configured models so you can switch quickly if one provider has issues
Use appropriate models — Larger models (GPT-5.3) produce better investigation reports; smaller models (GPT-5-mini) are faster and cheaper for simpler tasks

​How It Works

​Listing Models

​Adding a Model

​Model Configuration Fields

​Testing a Model

​Setting the Default Model

​View Current Configuration

​Update Default Model

​Deleting a Model

​Health Check

​Supported Providers

​Best Practices

How It Works

Listing Models

Adding a Model

Model Configuration Fields

Testing a Model

Setting the Default Model

View Current Configuration

Update Default Model

Deleting a Model

Health Check

Supported Providers

Best Practices