dsh-im-hub-media

Most IM bots choke the moment a user sends a voice note, a photo, or a file instead of plain text. dsh-im-hub-media removes that limitation for DeepSeek Harness agents by bridging Telegram, Feishu (Lark), and Enterprise WeChat in a single gateway. Incoming voice messages are intercepted and transcribed into text via the host STT chain; images and documents are routed to vision-capable models. On the way out, agents simply embed a media path marker in their reply and the plugin converts it into a native sendPhoto, sendVoice, or sendDocument payload, delivering true bidirectional rich-media conversation.

Security and isolation are first-class concerns. Each chat window maintains its own session state to prevent context cross-talk. Administrators can restrict access with user and chat allowlists, and all bot tokens and secrets are resolved through the host Credentials service rather than sitting in plain-text config. The Feishu adapter adds WebSocket long-polling, interactive message cards with buttons and event callbacks, while the WeCom adapter ships with built-in XML/JSON encryption to meet enterprise compliance needs.

If you are deploying a DeepSeek Harness agent as a round-the-clock assistant for your team or your customers and your day-to-day conversation relies on voice messages, screenshots, or file sharing, this plugin is the missing piece.