Models and Providers
After reading this page you will be able to: choose between the HuiAI route and the bring-your-own-key route, then get a model platform configured and verified.
1. Two ways to set up models
Section titled “1. Two ways to set up models”Where: Settings → Model
| Route | Who it is for | What you do |
|---|---|---|
| Use HuiAI | Users with a HuiAI login token | Save the token in Required Setup — the HuiAI provider appears on the Model page automatically, nothing to configure by hand |
| Bring your own key | Users with their own platform API keys | Add platforms and keys manually on this page, see section 3 |
Paste and save your HuiAI login token under Settings → Required Setup. The app auto-creates the HuiAI model provider (deepseek-v4-flash / x-preview-f-free / gpt-5.6-terra) on the Model page and keeps its key updated.
Click Add Model in the top-right corner and choose Add manually: pick a platform, paste your API key, select models — ready in a few minutes. Multiple keys can auto-rotate.
2. Supported platforms
Section titled “2. Supported platforms”The platform dropdown lists platforms in the order below: Custom comes first, with Kimi (Moonshot) pinned right after it, followed by direct-connect official platforms and preset providers. The in-app list is authoritative.
| Platform | Region | Notes |
|---|---|---|
| Custom | Any | No preset endpoint; you enter the API endpoint yourself |
| Moonshot (China) | China | Kimi China |
| Moonshot (Global) | Global | Kimi international |
| New API | Any | Aggregation gateway; pick a request protocol per model |
| Gemini | Global | Google official platform |
| Gemini (Vertex AI) | Global | Google Vertex AI channel |
| OpenAI | Global | GPT family |
| Anthropic | Global | Claude family |
| AWS Bedrock | Global | Amazon managed models |
| DeepSeek | China | — |
| MiniMax | China | — |
| Novita | Global | — |
| OpenRouter | Global | Aggregation gateway |
| Dashscope | China | Alibaba Cloud Qwen |
| Dashscope Coding Plan | China | Qwen coding plan; requires your own plan endpoint |
| SiliconFlow-CN | China | SiliconFlow |
| SiliconFlow | Global | SiliconFlow international |
| Zhipu | China | Zhipu (GLM) |
| xAI | Global | Grok family |
| Ark | China | Volcano Engine Ark |
| Qianfan | China | Baidu Qianfan |
| Hunyuan | China | Tencent Hunyuan |
| Lingyi | China | Lingyi Wanwu |
| Poe | Global | Aggregation platform |
| PPIO | China | — |
| ModelScope | China | ModelScope community |
| InfiniAI | China | — |
| Ctyun | China | China Telecom Ctyun |
| StepFun | China | StepFun |
3. Add a platform (bring your own key)
Section titled “3. Add a platform (bring your own key)”- Open Settings → Model, click Add Model in the top-right corner and choose Add manually (or choose Add via chat and let the Butler guide you)
- Pick a Model Platform; the API Endpoint is filled in for most platforms. If the address looks off, the app suggests base_url auto-fix suggestions
- Fill in the API Key: one per line for multiple keys, the system will auto-rotate
- Pick the models to enable (click the magnifier in the model box to fetch the model list) and adjust the options below as needed
- After saving, the platform collapses into a single row; expand it to see its models
Platform fields explained:
| Field | Notes |
|---|---|
| Model Platform | Determines the preset endpoint, protocol, and icon; switching platforms clears the selected models and resets the API Endpoint to the platform’s preset value |
| API Endpoint | Auto-filled for preset platforms; turn on the Full URL switch to use the address as-is without appending paths |
| API Key | Support multiple API Keys, one per line, system will auto-rotate; you can add more when editing the platform |
| Request Protocol | Shown for New API platforms only: OpenAI / Gemini / Anthropic. Choose the API request format for this model; the new-api gateway will route to the upstream using the selected protocol |
| Vision input | Automatic detection / Enabled by configuration / Disabled by configuration; automatic uses catalog capability detection. Enable or disable only to override it |
While you type, the app automatically detects the protocol and key (you will see “Detecting API protocol…”). With multiple keys it shows a summary: all keys valid, how many valid vs invalid, or all invalid — fix as instructed.
4. Day-to-day management
Section titled “4. Day-to-day management”Each platform collapses into one row; hovering shows “Models (N) | API Keys (N)”, and expanding lists its models one by one.
| Action | Where | Result |
|---|---|---|
| Edit platform | Pencil button on the platform row | Change the name, endpoint, keys (multi-key rotation supported) |
| Add model / remove platform | ”+” and ”−” buttons on the platform row | Removing a platform asks “Are you sure you want to delete all models?” first |
| Enable / disable | Switches on the platform row and each model row | Turn off a few models and keep only the ones you use |
| Configure model | Gear button on the model row | Adjust per-model settings such as vision input |
| Health Check | Heartbeat button on the model row | Shows “Platform - Model: Success (latency ms)” or the failure reason; the dot in front of the model name turns green / red, hover for latency and check time |
| Clear status | ”Clear status” in the top-right corner | Wipes all health check results, showing “Health status cleared” |
Deleting a single model asks “Are you sure you want to delete this model?”; deletions only affect future conversations.
5. One-click provider import (aionui:// links)
Section titled “5. One-click provider import (aionui:// links)”Users of gateways like New API or one-api can click an aionui:// link from the web UI:
aionui://add-provider?base_url=xxx&api_key=xxx— opens with an endpoint and keyaionui://provider/add?v=1&data=…— opens with the whole config packed into one parameter
AIPC UP then opens the Add Model dialog pre-filled; review and save. If the browser asks for permission to open an external app, allow it.
6. HuiAI and Google sign-in
Section titled “6. HuiAI and Google sign-in”Google account OAuth sign-in: not available in this version — there is no in-app Google login entry, and models are always configured with API keys. To use Gemini with a Google account, install the Gemini CLI Agent and finish its account login in a terminal; see Agents.
How to tell it worked
Section titled “How to tell it worked”- The Model page shows your platform card, and expanding it lists the enabled models
- With multiple keys, detection reported “all keys valid” or pinpointed the invalid ones
- A Health Check turns the dot in front of a model green and shows the latency in milliseconds
- A new Aion CLI conversation offers the models you just enabled