If you use a desktop AI workspace for real coding or office automation, the annoying part is rarely the chat UI. It is model routing: one task wants a long-context model, another wants a reasoning model, and another just needs a fast OpenAI-compatible chat endpoint.
This guide walks through a practical setup for using Ace Data Cloud as an OpenAI-compatible provider inside WorkBuddy, the desktop AI workspace from the Tencent CodeBuddy family. The goal is not to replace WorkBuddy's built-in models. It is to add your own custom model entries so you can switch between model families from the same WorkBuddy model picker.
What you can do
WorkBuddy supports custom models through an "OpenAI compatible protocol API" flow. Ace Data Cloud exposes chat-completion endpoints that match that shape, so the setup is mostly a matter of entering the right interface address, API key, and model name.
The document lists these interface addresses:
| Model family | Interface address | Example model IDs |
|---|---|---|
| Claude / general entry | https://api.acedata.cloud/v1/chat/completions |
claude-opus-4-8, claude-sonnet-4-6, claude-haiku-4-5-20251001
|
| GPT | https://api.acedata.cloud/openai/chat/completions |
gpt-5.6-sol, gpt-5.6-luna, gpt-5.5, gpt-5.5-pro, gpt-5.2, gpt-4o, o3
|
| Gemini | https://api.acedata.cloud/gemini/chat/completions |
gemini-3.1-pro, gemini-3.0-pro, gemini-3.5-flash
|
| Grok | https://api.acedata.cloud/grok/chat/completions |
grok-4, grok-3
|
| Kimi | https://api.acedata.cloud/kimi/chat/completions |
kimi-k3, kimi-k2.6, kimi-k2.5
|
| GLM | https://api.acedata.cloud/glm/chat/completions |
glm-5.1, glm-4.7, glm-4.6
|
| DeepSeek | https://api.acedata.cloud/deepseek/chat/completions |
deepseek-r1, deepseek-v3, deepseek-v4-flash
|
The most convenient path is the general entry:
https://api.acedata.cloud/v1/chat/completions
According to the guide, /v1/chat/completions can route to different model IDs, not only Claude. If you prefer separate entries per model family, use one of the family-specific addresses above.
How it works in WorkBuddy
Open WorkBuddy and go to Settings → Models. In the Custom Models area, click Add Model. The add-model dialog is where the important mapping happens:
| WorkBuddy field | Value to enter |
|---|---|
| Provider | Custom |
| Interface Address | For example, https://api.acedata.cloud/v1/chat/completions
|
| API KEY | Your Ace Data Cloud API Token |
| Model Name | A real model ID, such as claude-opus-4-8, gpt-5.5, or gemini-3.1-pro
|
The interface address should include the full /chat/completions path. The guide notes that WorkBuddy validates the path according to OpenAI standards when Custom Protocol is off, and Ace Data Cloud's entry is already the complete standard path.
A minimal OpenAI-style request against the general endpoint looks like this:
curl -X POST "https://api.acedata.cloud/v1/chat/completions" \
-H "Authorization: Bearer YOUR_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-opus-4-8",
"messages": [
{
"role": "user",
"content": "Review this function and suggest a safer implementation."
}
]
}'
The exact request is useful for sanity-checking the same endpoint and model name you put into WorkBuddy. If the model name is wrong, the document says the Ace Data Cloud gateway will reject the request and WorkBuddy will surface a 4xx error.
Choosing the advanced switches
WorkBuddy exposes several capability switches for custom models. These are not decoration; they tell the client how to treat the model in the UI.
Tool Calling declares that the model supports function calling or tools. The guide says mainstream Ace Data Cloud models such as Claude, GPT, Gemini, Grok, Kimi, and DeepSeek support this, so it can be enabled when your chosen model supports tools.
Image Input is for multimodal image input. Enable it only for model IDs that support multimodal input, such as the examples named in the document: claude-opus-4-8, gpt-5.5, or gemini-3.1-pro.
Inference Mode is for models with a visible reasoning or thinking process. The guide names o3, deepseek-r1, gemini-3.1-pro, and grok-4 as examples where this may apply.
Custom Protocol should stay off for this setup. With it off, WorkBuddy validates and uses the standard /chat/completions behavior; Ace Data Cloud's endpoint already matches that expectation.
For context-window dropdowns, the document recommends leaving both Input and Output as Use Provider Default, so WorkBuddy uses the upstream model's window configuration.
Adding more than one model
A useful pattern is to add multiple WorkBuddy custom model entries that share the same API key but use different model IDs.
For example, you might create:
Provider: Custom
Interface Address: https://api.acedata.cloud/v1/chat/completions
API KEY: YOUR_API_TOKEN
Model Name: claude-opus-4-8
Then repeat the same flow for:
Provider: Custom
Interface Address: https://api.acedata.cloud/openai/chat/completions
API KEY: YOUR_API_TOKEN
Model Name: gpt-5.5
After saving, both models appear under WorkBuddy's Custom Models group. That makes the model decision a per-task choice instead of a one-time workspace decision.
Troubleshooting checklist
If WorkBuddy returns an error, check these items first:
- The interface address includes
/chat/completions. - Custom Protocol is off.
- The model name exactly matches a model ID supported by the selected endpoint.
- The API key is the Ace Data Cloud API Token from your console.
- Capability switches match the selected model; for example, do not enable image input for a text-only model.
WorkBuddy's own documentation says custom model configuration, including the API key, is stored locally in workbuddy/models.json and is not uploaded to the cloud. WorkBuddy acts as the communication link that forwards your input to the configured API address.
If you want the source configuration table and screenshots, the full Ace Data Cloud WorkBuddy setup guide is here: https://platform.acedata.cloud/documents/development_workbuddy
Top comments (0)