Configuration

Evaluation arena configuration endpoints

Get Evaluation Configuration

Retrieve the current evaluation arena configuration. Admin only.

The evaluation arena allows users to compare models side-by-side and provide feedback on which performs better for specific tasks.

Response

ENABLE_EVALUATION_ARENA_MODELS · boolean

Whether the evaluation arena feature is enabled

EVALUATION_ARENA_MODELS · array

Array of model configurations available in the evaluation arena

Show Model configuration properties

id · string

Unique model identifier

name · string

Display name for the model

info · object

Additional model information and metadata

cURL

curl -X GET "https://your-instance.com/api/evaluations/config" \
  -H "Authorization: Bearer $OPENWEBUI_API_KEY"

{
  "ENABLE_EVALUATION_ARENA_MODELS": true,
  "EVALUATION_ARENA_MODELS": [
    {
      "id": "gpt-4",
      "name": "GPT-4",
      "info": {
        "provider": "openai",
        "capabilities": ["chat", "code"]
      }
    },
    {
      "id": "claude-3",
      "name": "Claude 3",
      "info": {
        "provider": "anthropic",
        "capabilities": ["chat", "analysis"]
      }
    },
    {
      "id": "llama-3",
      "name": "Llama 3",
      "info": {
        "provider": "meta",
        "capabilities": ["chat"]
      }
    }
  ]
}

---

Update Evaluation Configuration

Update the evaluation arena configuration. Admin only.

Use this endpoint to enable/disable the arena feature or modify the list of models available for comparison.

Request Body

ENABLE_EVALUATION_ARENA_MODELS · boolean

Enable or disable the evaluation arena feature

EVALUATION_ARENA_MODELS · array

Array of model configurations to make available in the arena

Show Model configuration properties

id · string · required

Unique model identifier (must match an existing model ID in your system)

name · string · required

Display name for the model

info · object

Additional model information and metadata

Response

Returns the updated configuration with the same structure as the GET endpoint.

cURL - Enable Arena

curl -X POST "https://your-instance.com/api/evaluations/config" \
  -H "Authorization: Bearer $OPENWEBUI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ENABLE_EVALUATION_ARENA_MODELS": true
  }'

{
  "ENABLE_EVALUATION_ARENA_MODELS": true,
  "EVALUATION_ARENA_MODELS": [
    {
      "id": "gpt-4",
      "name": "GPT-4",
      "info": {"provider": "openai", "version": "latest"}
    },
    {
      "id": "claude-3-opus",
      "name": "Claude 3 Opus",
      "info": {"provider": "anthropic", "version": "latest"}
    },
    {
      "id": "llama-3-70b",
      "name": "Llama 3 70B",
      "info": {"provider": "meta", "parameters": "70B"}
    }
  ]
}

Configuration Notes

Arena Feature

The evaluation arena enables users to:

• Compare multiple models side-by-side on the same prompt
• Rate which model performed better (creating feedback entries)
• Tag evaluations with topic keywords (e.g., “coding”, “creative writing”)
• Contribute to the overall leaderboard rankings

Model Selection

When configuring EVALUATION_ARENA_MODELS:

• Model IDs must correspond to models available in your Open WebUI instance
• Include a diverse set of models to enable meaningful comparisons
• Consider organizing models by capability (e.g., general chat, coding, reasoning)
• The info object can include any metadata useful for your use case

Disabling the Arena

To disable the evaluation arena:

{
  "ENABLE_EVALUATION_ARENA_MODELS": false
}

This will hide the arena UI while preserving existing feedback data.