Bestony AI Provider for Tencent Cloud Token Hub

This plugin adds Tencent Cloud Token Hub’s OpenAI-compatible Chat Completions API to the WordPress AI Client.

• The model list is fetched live from GET /models .
• Only models with status: online are exposed.
• Text generation, chat history, tool calling, streaming and structured output are supported.
• Confirmed vision models declare text and image input.
• Image, video and 3D generation models are excluded because this provider implements Chat Completions only.
• A request with more than four messages is rejected before it reaches the API.

Configuration

The provider ID is tencentcloud_tokenhub .

The API key is owned by the AI Client credential registry. You can enter it in Settings → Connectors as connectors_ai_tencentcloud_tokenhub_api_key , or define TENCENTCLOUD_TOKENHUB_API_KEY as an environment variable or PHP constant supported by the SDK. The provider does not read the Connector option directly.

Optional environment variables or PHP constants:

• TENCENTCLOUD_TOKENHUB_DEFAULT_MODEL — model preferred by the AI plugin. Default: hy4-preview .
• TENCENTCLOUD_TOKENHUB_REQUEST_TIMEOUT — request timeout in seconds. Default: 120 .
• TENCENTCLOUD_TOKENHUB_CONNECT_TIMEOUT — connection timeout in seconds. Default: 10 .
• TENCENTCLOUD_TOKENHUB_STRUCTURED_OUTPUT — json_schema (default), json_object , or none .

The API base URL is fixed at https://api.lkeap.cloud.tencent.com/plan/v3 . The key is sent only to this host. The plugin does not provide a custom host setting.

Model capabilities

Token Hub’s /models response does not include capability details. Unknown online models are therefore treated as text-only chat models. The provider maintains a conservative list of confirmed vision IDs, including the documented DeepSeek V4 Flash Vision, GLM-5V, HY-Vision, Hunyuan vision, and YT-VITA families. The list can be updated when Tencent adds documented vision models.

Image, video, 3D, embedding, speech and other non-chat families are filtered out. No image, video or 3D generation endpoint is implemented.

Message limit

Token Hub rejects a messages array longer than four entries. This provider throws an AI Client parameter exception before sending a request when the prompt would exceed that limit.