Tencent’s WorkBuddy desktop app can connect to B.AI custom models through an OpenAI Chat Completions–compatible configuration, giving users a way to run selected third-party AI models inside a local desktop workflow. The connection requires a B.AI API key, an exact model ID supported by B.AI’s Chat Completions endpoint, and careful matching of the app’s base URL with its custom-protocol setting.
The setup is local to each computer. WorkBuddy stores model configurations in .workbuddy/models.json, meaning a user who signs in on a second device will need to add the B.AI model again and repeat the connection tests. That design keeps credentials and settings from being automatically carried between machines, but it also makes local backups and API-key management more relevant for users operating across several workstations.
Desktop support is limited to Windows and macOS
WorkBuddy’s desktop release supports Windows 10 or later and macOS 12.0 or later. The Windows installer is offered in x64 format and is compatible with ARM64 systems, while macOS users must choose between separate Apple silicon and Intel versions.
Linux is not included in the desktop support described. The workflow also applies specifically to the WorkBuddy desktop client rather than mobile applications or web-based interfaces.
Before beginning, users need three items: a WorkBuddy account, a B.AI API key and a valid B.AI model ID. The model ID should be confirmed through B.AI’s GET /v1/models endpoint or the provider’s console model list. A visible display name, an alias found in an older guide or a model supported only by another API format may not work.
B.AI’s API key is displayed only once when it is created through the service’s API management pages. Users who do not save it at that point may need to generate a replacement key rather than retrieve the original one later.
URL setting determines whether requests reach the API
In WorkBuddy, users add the connection through Settings, then Models, then Add model. The provider should be set to “Custom” before entering the API key, model ID and connection details.
The most common configuration risk involves the relationship between the base URL and WorkBuddy’s “Custom protocol” toggle. When the toggle is off, WorkBuddy automatically adds /chat/completions to the supplied address. When it is on, the application sends requests to the full URL entered by the user.
That means a base URL and toggle setting must be paired correctly. Entering a URL that already ends with /chat/completions while leaving the automatic behavior active can produce a malformed route such as:
/v1/chat/completions/chat/completions
A request sent to that duplicate path will generally return a 404 error. Users updating a previous B.AI connection are advised to use the edit function or create a new model entry rather than overwrite an unrelated configuration.
WorkBuddy’s advanced settings should be treated as capability declarations. Enabling a switch does not add functions a B.AI model or endpoint does not actually offer. If the model’s supported input length, output limit or tool-use features cannot be confirmed, the recommended approach is to retain the provider defaults.
Saving the model only confirms that WorkBuddy has stored the configuration. It does not establish that the API key is valid, that the selected model is available or that an actual request can reach B.AI.
Basic chat test should come before file access
The first live test should use a new WorkBuddy task and directly select the newly configured B.AI model. Users should avoid the “Auto” option during this stage, since automatic routing could send the prompt to another model and obscure whether the custom connection is working.
A short plain-language chat prompt is the recommended first test because it avoids dependencies on tools, folder access and file permissions. If that request fails, troubleshooting should begin with the currently selected model, followed by the API key, base URL and protocol toggle, the exact model ID, and then account quota or permissions.
The supplied error codes point to different parts of the connection:
- A 401 response usually points to an API-key issue.
- A 403 response can indicate an authorization or access restriction.
- A 404 response often signals an incorrect URL, an incorrectly matched protocol setting or an unavailable endpoint.
- A 429 response points to rate limits or quota constraints.
For a “model not found” error, the model ID should be copied directly from B.AI’s GET /v1/models response or console listing. The selected model must also be accessible through Chat Completions; a model available only through Anthropic Messages or OpenAI Responses would not meet the stated requirement.
File-reading tests depend on tools and permissions
Once basic chat succeeds, users can test WorkBuddy’s tool-calling workflow by granting the task a workspace folder containing one or two non-sensitive text files. A read-and-summarize prompt can then test whether the model is able to invoke tools and access local files.
A successful response in this second stage depends on several layers working together: WorkBuddy must have permission to access the workspace, the selected B.AI model must support tool calling, the endpoint must accept those tool calls, and the relevant capability setting must be enabled in the configuration.
When chat works but file reading fails, the suggested order of checks is the model’s capability settings, the tool-calling toggle, the selected workspace and file permissions. Testing with ordinary text files is preferable to beginning with protected folders, system locations or large datasets.
Users can also compare the time of their test with B.AI console usage or call logs, checking the model ID, number of requests and token consumption. Matching records provide a stronger confirmation that WorkBuddy reached the intended B.AI model rather than another route or locally selected fallback.
Connectivity problems should be narrowed down by confirming access to https://api.b.ai, then reviewing the URL entry, proxy settings or enterprise network policies, B.AI service availability and unusually large inputs that could delay responses. The guidance recommends changing one setting at a time and rerunning the basic chat prompt before returning to file-based tests.
Need to validate your WorkBuddy–B.AI integration? Use Toobit’s API testing tools to simulate and debug requests.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
