Connect WorkBuddy to Nbility

WorkBuddy is Tencent's desktop AI workspace. Its current release supports adding custom models through a graphical interface by entering an endpoint, API key, and model name, without editing configuration files. This guide follows the latest official WorkBuddy model configuration documentation and Tencent Cloud's custom model setup example.

Install and prepare

1. Download the client from the WorkBuddy website.
   • macOS requires macOS 12 or later. Choose the ARM64 or X64 package for your processor; see the official macOS installation guide.
   • Windows requires Windows 10 or later. Windows 7, 8, and 8.1 are not supported; see the official Windows installation guide.
2. Sign in to WorkBuddy.
3. Create a key dedicated to WorkBuddy on Token Management.
4. Confirm the exact model ID on Models.

Add a custom model

1. Start WorkBuddy, select the avatar or account entry in the lower-left corner, and open Settings.
2. Select Models in the sidebar, find Custom Models, and choose Add Model. The model selector in the conversation view also provides a shortcut to this screen.
3. Select Custom from the provider list and enter:

WorkBuddy's current official integration example uses the complete /v1/chat/completions endpoint, so do not enter only the website origin. Add and save a separate configuration for each model.

Configure model capabilities

Open the advanced settings and match each option to the selected model and route:

• Tool calling: required for Agent tasks such as file access and command execution. Enable it only when both the model and upstream route support native tools / function calling.
• Image input: enable it only when both the model and route support vision input.
• Reasoning mode: enable it only for a model that actually supports reasoning mode.
• Custom protocol: Nbility uses the standard Chat Completions path, so this can normally remain off. If your WorkBuddy version rewrites the complete URL incorrectly, enable this option so the client sends requests directly to the URL above.

Capability switches do not add capabilities to a model. Incorrect settings can cause request errors, failed tools, or unreadable image input.

Verify the connection

Save the model, return to the conversation view, and select it from the model picker. Send a short task such as:

List the files in the current working directory and summarize its structure in three sentences. Do not modify any files.

Confirm that the model replies normally. If tool calling is enabled, also confirm that the read-only tool action succeeds. Then open Usage Logs and verify the model name, request status, and token usage.

Troubleshooting

• 401: paste the key again and confirm that it has no extra whitespace, is active, and has permission.
• 404: use https://api.nbility.ai/v1/chat/completions. If WorkBuddy produces an invalid path after automatic completion, enable Custom protocol and try again.
• model not found: use the exact ID from Models and confirm that the key's group can access it.
• 429 or insufficient quota: check the account balance, key quota, concurrency limit, and request rate.
• Chat works but tools fail: enable Tool calling and use a model and route with native tool support.
• Images are not understood: enable Image input and use a model and route with vision support.

Key and data safety

The official WorkBuddy documentation states that custom model settings and API keys are stored locally and are not uploaded to WorkBuddy's cloud. Conversation content is forwarded to the configured third-party model service. Use a dedicated key, set a reasonable quota, review logs regularly, and never expose the key in screenshots, messages, or shared configuration. When the integration is no longer needed, delete the model in WorkBuddy and revoke its Nbility key.