Basic Concepts

LLM (Large Language Model) is Verity's "brain", which is responsible for understanding what you say and generating responses. The ability to talk, remember and so on is driven by it. Common LLM service providers are DeepSeek, Doubao, etc.

Use of LLM requires registration and creation with the service provider API Key, and then send the message to the model through the interface and get the reply.

TTS (Text-to-Speech) is responsible for turning text into speech - it's what you get when you hear Verity talking. The tone, intonation and other effects are determined by the model provided by the service provider.

STT (Speech-to-Text), in contrast to TTS, is responsible for converting the speech you utter into text - it is what recognizes what you say when you press and hold the V key for speech input.

Base Url is the base address of the API service, usually in the form of https://api.example.com/v1 URL. The module will send a request to this address to use the corresponding service.

“OpenAI-compatible format” means that the request and response structure of the interface follows a common specification developed by OpenAI - currently the vast majority of LLM service providers follow it, so you can switch service providers simply by replacing the Base Url and Key.

Key is a "pass" issued to you by the service provider to verify your identity and billing. After the service provider's console is created, it can be filled into the module or the official new website.

The same service provider usually provides multiple models, for example, DeepSeek has different dialogue models. They differ in ability, responsiveness, and price: strong models respond smarter but more costly and slower, and lightweight models respond the opposite.

Selecting the Model in the "Proxy Service" is choosing which model to use to drive Verity. If you are not sure how to choose, you can use the default recommendation of the service provider first.

The LLM interface has been generally unified into an OpenAI-compatible format; however, The interface between TTS and STT is not yet unified, the realization of each manufacturer varies greatly.

Therefore, this module is only individually adapted toXiaomi MiMo 's TTS and STT; it also supports the TTS/STT interface of the OpenAI-compatible format, which facilitates access to the platform that provides voice services in this format.