WorkBuddy

Tencent's desktop AI agent workbench: local file processing, Skills, MCP and task automation. Its custom-model entry speaks the OpenAI-compatible protocol, so it connects directly to ByteSeek.

Install

Platform	Method
macOS (Apple Silicon)	Download the `arm64` DMG from the international site or the China site
macOS (Intel)	Same sites, `x64` DMG
Windows 10+	Same sites, Windows x64 installer (ARM64-compatible)

On macOS drag WorkBuddy.app into Applications. First launch prepares a local runtime — stay online and wait for the welcome screen; do not force-quit.

Prerequisites

WorkBuddy installed
An OpenAI-format group key from the API Keys page

Add a ByteSeek model

Open WorkBuddy's model settings, pick Custom at the bottom of the provider list, and fill in:

Setting	Value
Provider	`Custom`
Endpoint	`https://tokens.byteseek.ai/v1/chat/completions`
API Key	Your `sk-` key
Model name	An exact ID from the model plaza, e.g. `gpt-5.6-sol`
Tool calling	Enable if the model supports it
Image input	Enable if the model supports vision
Thinking mode	Enable for reasoning models
Custom protocol	Keep off (ByteSeek is standard OpenAI-compatible)

Enter the full endpoint path

WorkBuddy does not append /v1/chat/completions automatically — enter the full address ending in /chat/completions, not just the domain.

Configuration is stored locally in ~/.workbuddy/models.json and never uploaded. With a composite key, model names need the group prefix, e.g. GPT/gpt-5.6-sol.

Verify

Back in the conversation view, select the newly added model and send a message; verify the charge on the usage page.

Common issues

Symptom	Fix
Timeouts / format errors	Endpoint missing the full path — add `/v1/chat/completions`
401	Key pasted incompletely — re-paste
403	Model not in group or format mismatch, see Error Codes
Tool calling inert	The switch is off, or the model lacks tool-calling support

#WorkBuddy

#Install

#Prerequisites

#Add a ByteSeek model

#Verify