Build a dedicated AI hub: A complete illustrated tutorial for connecting WorkBuddy to B.AI models

In today's fast-paced development environment, WorkBuddy has become an indispensable "productivity hub" for many developers on desktop, thanks to its flexible workflow orchestration and powerful system-level integration capabilities. It not only aggregates fragmented development tools but also serves as a dedicated intelligent middle platform, allowing developers to directly invoke top-tier AI capabilities in an immersive environment without switching contexts—greatly eliminating the friction of daily task execution and returning core focus entirely to high-value thinking and creation.

To help developers more efficiently bring B.AI's high-performance model matrix and system-level infrastructure into their daily development workflows, this article provides a detailed guide on how to add B.AI custom models in the WorkBuddy desktop client. Before configuration, please prepare a WorkBuddy account, a B.AI API Key, and the model ID actually available to your current account. Next, simply follow the straightforward configuration steps in this article to unlock an exceptionally smooth local AI collaboration experience.

1. Before You Begin

Step 1: Install WorkBuddy

Go to the official WorkBuddy download page and select the appropriate installation package for your system environment. Users who have already installed WorkBuddy can skip this step. If you cannot find the "Model" or "Add Model" option, first select "Check for Updates" to upgrade your version.

Windows

The official page currently offers a Windows x64 (compatible with ARM64) installation package, requiring Windows 10 or later. After downloading the installer, double-click it and follow the wizard to complete the installation, then launch WorkBuddy.

Note: If the system blocks the installation, first confirm that the installer comes from the official page, then check the app name and publisher information in the pop-up. Do not bypass the check by disabling Windows security protection.

macOS

The official page currently offers separate .dmg files for Apple silicon and Intel versions, requiring macOS 12.0 or later.

• For M1, M2, M3, M4 and other models, choose the Apple silicon version.
• For Intel processor models, choose the Intel version.

After opening the .dmg, drag WorkBuddy into "Applications," then launch it from the Applications folder. If you are unsure of your chip type, you can check "Chip" or "Processor" in "About This Mac."

Other System Notes

Currently, the WorkBuddy desktop client only supports Windows and macOS; Linux is not yet supported. Also, given the different feature scope of mobile and HarmonyOS clients, all operations in this tutorial are based on the desktop client.

Step 2: Log In to WorkBuddy

On first launch, click "Log In" and complete authentication according to the method displayed in the client. The international version's official documentation lists Google and GitHub OAuth. If your current client displays WeChat QR code scanning or other entry points, the actual options shown in the client shall prevail.

Step 3: Obtain a B.AI API Key

First log in to B.AI, navigate to the API Management or API Key Management page in the left navigation bar, click "Create API Key," and give it an easily recognizable name (e.g., WorkBuddy). Due to B.AI's official mechanism, the complete Key is only displayed once upon successful creation, so be sure to copy and save it immediately after creation.

Important confirmations before configuration:

• Pre-check: Before configuration, confirm that your account has available quota and that the API Key has invocation permissions for the target model.
• Obtaining the Model ID: The model ID should be based on the response from B.AI's GET /v1/models or the current console list—do not rely solely on example names from other tutorials.
• Protocol Compatibility: WorkBuddy custom models use the OpenAI Chat Completions protocol, so the target model must also be open to that endpoint. Configurations that only support Anthropic Messages or OpenAI Responses cannot be entered directly.

Note: Do not expose the complete Key in articles, screenshots, chat logs, or public repositories. If you suspect the Key has been leaked, immediately delete the old Key, create a new one, and update the WorkBuddy configuration.

2. Configure the B.AI API via the Interface

Step 1: Open Custom Model Configuration

Open WorkBuddy, click the account avatar in the lower-left corner, and select "Settings."

Select "Model" on the left, then click "Add Model."

If there are already other custom models in the current list, make sure to click "Add Model" to create a new configuration—do not directly overwrite the parameters of unrelated models. If you need to modify an existing B.AI model, simply click the pencil icon next to that model. If you want to keep the old version while making modifications, you should add a new configuration entry as a completely new model.

Step 2: Fill In the Access Parameters

Select "Custom" from the "Provider" dropdown list, then fill in the relevant information on the pop-up page.

The filling instructions are as follows:

The URL and "Custom Protocol" must be set as a pair:

According to WorkBuddy's official documentation, when "Custom Protocol" is turned off, the client automatically appends /chat/completions to the end of the URL you entered, following standard OpenAI Chat Completions rules; when this switch is turned on, the client will send requests directly to the complete URL you entered without any path concatenation. Note: do not mix the two approaches. Duplicate concatenation resulting in /v1/chat/completions/chat/completions will typically return a 404.

Step 3: Set Model Capabilities

The switches in "Advanced Settings" are only used to declare model capabilities—checking them will not automatically grant capabilities to a model that does not originally support them. Please configure based on actual conditions, referring to the following principles:

When you cannot confirm the input or output limits, keep "Use Provider Defaults." Do not enable all capabilities by default without confirmation from documentation or testing.

Step 4: Save the Configuration

After checking the URL, API Key, and model ID, click "Save." The new model should appear in the "Saved Models" list.

If it does not appear, check in order whether the save pop-up is still open, whether any fields show errors, whether the model selector has refreshed, and the client version. If necessary, completely exit WorkBuddy and restart it. A model appearing in the list only indicates that the configuration has been saved—it does not prove that the B.AI invocation has succeeded. Follow-up verification is still required.

3. Select the Model and Verify the Configuration

Step 1: Select the B.AI Custom Model

Return to "New Task," open the model selector near the input box, and select the model you just added under the custom model group. During testing, do not select "Auto" mode—"Auto" mode may dispatch other models and cannot prove that this request used B.AI.

Step 2: Verify Normal Conversation

Send a simple question that does not rely on tools:

Please reply only with: B.AI normal conversation test successful

Receiving a normal reply indicates that WorkBuddy has read the configuration and that the Key, URL, and model ID can at least complete one text invocation. If it fails, check in the order of "current model → API Key → URL and protocol switch → model ID → quota and permissions." Do not judge whether routing succeeded by asking the model "Who are you"—the model's self-reported identity cannot serve as evidence of successful integration.

Step 3: Verify Tool Calling

First create a new folder used only for testing, place one or two text files without private information into it, then select that folder as the workspace in WorkBuddy, grant only the permissions required to complete the reading task, and finally send the following instruction:

Please read the text files in the current workspace, list the file names, and summarize the content of each in one sentence. Do not modify, move, or delete any files.

Only when WorkBuddy displays a file-reading tool call and correctly returns the file summaries does it indicate that the tool calling chain is available. If it can chat but cannot read files, check tool calling, the model's tool calling capability, B.AI endpoint support, and workspace permissions.

Step 4: Verify B.AI Call Records

If the B.AI console provides usage or call records, go check the recent request times, model IDs, request counts, and Token consumption to ensure the data matches the tests you just performed.

To ensure the entire integration process is complete and usable, please verify the success status of the following three stages one by one:

4. Frequently Asked Questions

1. What should I do if I cannot find the custom model entry point?

It may be that you are not logged in, the client version is outdated, or you are not in the desktop client's model settings. First log in, then go to "Settings" → "Model." If you still cannot find the entry point, select "Check for Updates" to update the client and restart. If the problem persists, submit your version number and a screenshot via "Help & Feedback."

2. What should I do if I cannot install or open it on Windows or macOS?

First confirm that your system version meets the requirements, the installer comes from the official page, and the downloaded architecture is correct. Handle blocking according to your system's provided security settings process—do not disable security protection or switch to unknown mirrors.

3. Why does the model not appear after saving?

First confirm whether the entry exists in "Saved Models," then close settings and reopen the model selector. If it still does not appear, completely exit and restart WorkBuddy, and also check for client updates.

4. What causes 401, 403, 404, and 429 errors, and how can they be resolved?

Only change one item at a time, then retest with a normal conversation to make it easier to pinpoint the cause.

5. How do I resolve a "model not found" prompt?

Call GET /v1/models again or check the current console list, and copy the exact id. Do not enter display names, aliases, or version numbers from other tutorials. Also confirm that the model is open to the Chat Completions endpoint.

6. What should I do if it keeps loading, times out, or fails to connect?

First confirm that your network can access https://api.b.ai, then test with a short question. Next, check the URL, proxy or corporate network policies, B.AI service status, and whether the input file is too large. If the problem persists, record the time of occurrence, WorkBuddy version, complete error message, and request ID, then contact official support.

7. What should I do if it can chat but cannot read files?

Check in the order of "model capability → Tool Calling switch → workspace → file permissions." First test with ordinary text files—do not directly select system directories, protected directories, or sensitive files.

8. Why is the system still calling the built-in model? 

Please check the model selector to make sure it is not in "Auto" state and that the configured B.AI custom model is explicitly selected. Send a short question, then go to the B.AI console to verify the call time and usage data. If you still have