V6.0 Configuration FAQs

Model Category Configuration Explanation

About Accessing Domestic Models

In v6, domestic version models have been removed. All models now follow the OpenAI input format. If you need to use domestic models, please use a relay. The system has built-in ultra-low-priced relays available, with dozens of models perfectly adapted to the system. If you want to access domestic models, please choose a suitable relay.

Parameter Description

• Open Status: This does not represent whether the model is started or stopped, but only whether it is displayed on the client side. For some internal tool keys or drawing keys, we can turn them off, but they will still work normally; users just can't select them.

• Chinese Name: This field can be customized to be displayed to the user.

• Associated Model: This is the real model called by the system. We only provide a part of them. If your relay supports it, you can manually enter the model in the model selection box, supporting any model.

• Deduction Type: The system has two types of balances. The deduction category of the selected model distinguishes between ordinary and expensive models.

• Per-call Deduction: The amount deducted for each call.

• Context: Limits the length that users can choose. Setting it lower can control the maximum context length, and the same goes for tokens.

• Custom Sorting: Controls the display order of models on the client side.

• Set as Default Model: Check this model, and it will be used by default when creating a new conversation on the client side. Only one can be checked; checking multiple is invalid.

• Set as Drawing Model: Requires the relay or key to support the gpt-4-e3 model. When this option is checked, the selected model can be ignored, and the system will ultimately call gpt-4-e3.

• Set as Tool Model: The system needs to use this model for some internal operations such as translation, association, etc.

• Support File Upload: Controls the upload button on the client side. When the current model has this option enabled, the upload button will be displayed, and at this time, you can paste and upload files.

• Support Networking Mode: The networking mode is a mode supported by the built-in relay. When called, net- will be added before the model. If your relay supports it, you can enable it; otherwise, it will report an error. It is recommended to use the built-in relay!

• All models support card pools, and configuring multiple keys will call them in an orderly manner according to weight.

How to enable Dall-E drawing

Dall-E drawing calls the gpt-4-e3 model. When your Key supports it, check the "Set as Drawing Model" option, and this model can be called for drawing!

How to use plugins

We have a built-in plugin system, which is not open in the current version. Since different plugins require different model support, the model will only be enabled when the corresponding model is set and there is a key in the card pool. The following are the models corresponding to each plugin:

• Dall-E Text-to-Image ---> gpt-4-dalle
• Midjourney Text-to-Image ---> mj-chat
• SD Text-to-Image ---> stable-diffusion
• Document Analysis ---> gpt-4-all
• Image-to-Text ---> gpt-4-all
• QR Code Generation ---> gpt-3.5-turbo-16k

Note the above models. When you configure a model like the above in the "Associated Model" (and there is at least one key in the card pool), the corresponding plugin will be automatically enabled!

How to modify plugin sample data

We currently do not have a setting open in the backend, but we have still reserved a modification entry for you. You can go to the /dist/common/constants directory in the source code to find the plugin.js file, where you can modify the corresponding examples and some basic configurations!

How to use tts voice

TTS voice calls the official text-to-speech and speech-to-text models. To save space, we do not store files locally or in the cloud. We will read them in real-time streams. You need to manually enter the tts name in the "Associated Model" in the model settings to enable both sides of the voice-to-text conversion. You can refer to the configuration of the demo site. Please make sure your key or relay supports the tts model. After configuration, the voice can be accessed normally!

How to use gpts for site-wide search

The recommended relay supports the search-gpts model. Add this model and configure the key.

How to use networking search

After enabling networking search, net- will be added before the model when calling it. The system's built-in relay already supports this. If your own key does not support it, you can go to the officially recommended relay to purchase it. The price is super low on the market, and it's even cheaper after connecting to the system!

How to use local storage, app logo cannot be uploaded, file upload issues

For local storage, we will store files in the local project. So if you need to use this function, please make sure to set the Backend Service BASE_URL in the backend under Registration Management/Email Login Registration. This parameter is required, otherwise, the file upload will not find the correct directory location!