Model Configuration

Create Model Configuration

Click the Create New Configuration button to open custom and mainstream model provider templates for selection:

Note that each configuration corresponds to only one model. Under the same provider, you can click the copy button to quickly create a new configuration.

Configuration

Custom Name

Used to distinguish different model configurations. Can be filled in arbitrarily and can be duplicated.

BaseURL

Note that you only need to configure up to the version number, e.g., https://api.openai.com/v1. The suffix will be added automatically. The same applies to other providers. Please do not fill in chat/completions, /embeddings, etc.

API Key

Fill in your corresponding key here. If using Ollama, LM Studio, etc., you can fill in anything.

Model

Generally, after BaseURL and API Key are configured correctly, Model will be automatically obtained and presented as a dropdown list. If not obtained, you can fill it in manually.

Model Type

Select the model type to distinguish different models. Selecting the wrong model type may result in inability to use normally.

Currently supported: Chat, Voice, Embedding, Rerank.

• Chat: The most common model type, used for dialogue, writing, and other scenarios.
• Voice: Used for speech recognition (STT) and text-to-speech (TTS) functions, configurable in audio settings.
• Embedding and Rerank models: Suitable for scenarios using knowledge base.
• Other model types will be gradually supported in future versions.

Advanced Configuration

If you don't understand these configurations, you can ignore them and use default values.

Temperature

What sampling temperature to use, between 0 and 2. Higher values (e.g., 0.8) will make the output more random, while lower values (e.g., 0.2) will make the output more focused and deterministic. We generally recommend modifying this or top_p but not both.

Top P

An alternative to temperature sampling, called nucleus sampling, where the model considers the results of tokens with top_p probability mass. So 0.1 means only considering tokens comprising the top 10% probability mass. We generally recommend modifying this or temperature but not both.