WorkBuddy
WorkBuddy can connect directly to Fusion through a custom model using either Chat Completions or Responses, without CC Switch. Choose a protocol and use its matching full request URL and Custom protocol setting below.
Connection parameters
Use a FUMO Dashboard key for an account registered and funded online.
| Parameter | Value |
|---|---|
| Chat Completions (full URL) | https://capi.fumolab.ai/api/fusion/v1/chat/completions |
| Responses (full URL) | https://capi.fumolab.ai/api/fusion/v1/responses |
| Available model IDs | fusion-max-2.0, fusion-code-1.0 |
| API format | OpenAI Chat Completions / Responses |
| Recommended model | fusion-max-2.0 |
| Input window / output limit | 373000 / 131072 tokens |
| Max 2.0 reasoning effort | Supports low, medium, high, xhigh, max; recommended: high |
Setup
1. Install the latest version and get an API key
Download the latest version from the WorkBuddy website↗. Create a FUMO API key using Getting started, or see Enterprise onboarding for enterprise accounts.
2. Add a custom model
- Open Settings → Models and click Add model. You can also open the model menu in the conversation composer and select Configure custom models.
- Set Provider to Custom.
- Set Endpoint to the full request URL for your chosen protocol. For Chat Completions, use:
https://capi.fumolab.ai/api/fusion/v1/chat/completions
For Responses, use:
https://capi.fumolab.ai/api/fusion/v1/responses
Leave Custom protocol off for Chat Completions. For Responses, turn it on under Advanced configuration so the client uses the exact URL without appending /chat/completions.
- Enter the FUMO API key for the selected access channel.
- Set Model name to:
fusion-max-2.0
This field takes a complete request URL. If you change the access channel at the top of this page, check both the URL and the key. FUMO does not support Anthropic Messages (/messages).
3. Configure capabilities and reasoning effort
Under Advanced configuration, use these settings for Fusion Max 2.0:
| Setting | Recommended value |
|---|---|
| Tool calling | Checked |
| Image input | Checked |
| Thinking mode | Checked |
| Thinking-only mode | Checked |
| Allow disabling thinking | Unchecked |
| Custom protocol | Off for Chat Completions; on for Responses |
| Default reasoning effort | High (high) |
| Supported reasoning efforts | Check Low, Medium, High, Extra high, Max |
| Input | 373000 |
| Output | 131072 |
The five levels map to low, medium, high, xhigh, and max. Supported reasoning efforts declares which levels are available; Default reasoning effort selects the default. Set both separately. Auto (use request-level default) does not select high: FUMO's server default is medium.
Fusion Max 2.0 does not support disabling thinking. Enable Thinking-only mode and leave Allow disabling thinking unchecked.
These numbers configure the client's input window and per-response output limit. Actual output length depends on the request and remaining context. 373000 is the recommended client setting, rather than the model's maximum context. Set the default reasoning effort to High (high) for either protocol.
4. Test the connection and select the model
- Click Test connection, review the result, then click Save. For Responses, also verify an actual conversation as described below.
- Start a new task, open its model menu, and select
fusion-max-2.0under Custom models. Saving a configuration does not ensure that every existing task switches to it. - Start with a simple question, such as “Reply with only the result of 17 × 23,” and check that the response is
391. - Then ask the agent to read a test file in a test workspace, and check both the tool call and the follow-up response.
Connection successful means the client's connection check passed. Verify ordinary conversations, tool calling, and image input separately. The Auto model menu option uses WorkBuddy's automatic selection; explicitly select your Fusion model.
The current Test connection sends a Chat Completions probe, so its result alone does not establish whether Responses works. After saving a Responses configuration, test a conversation and tool calling in an actual task. If those fail, check the full URL, Custom protocol, and the error details.
5. Optional: add Fusion Code 1.0
Create a separate model entry with the same endpoint and an authorized key. Set its model name to:
fusion-code-1.0
Use the same tool calling, image input, and input / output settings. Code 1.0 does not support adjustable reasoning effort. Leave the five effort levels and default effort unconfigured.
Troubleshooting
The endpoint returns 401, 403, or 404
Check that the endpoint matches the key's access channel, the key is complete, and the model ID is authorized. Use /chat/completions with Custom protocol off for Chat Completions, or /responses with it on for Responses. Avoid duplicate paths. If only the connection check fails, also verify an actual conversation as described above. See First-request troubleshooting.
A saved model is not being used by a task
Open that task's model menu and explicitly select fusion-max-2.0. Existing tasks can retain their previous model; an entry in Settings alone does not confirm the active model.
High is missing or only Auto is available
Edit the model, enable Thinking mode, check all five Supported reasoning efforts, set Default reasoning effort to High, and save. These controls are not needed for Code 1.0.
