AI Agent Hub
Back to plugins
🖥️

dsh-hermes-bridge

Client Updated 2026.08.24

Run the following command in DeepSeek Harness:

dsh plugin install dsh-pulse/dsh-hermes-bridge

Paste the following prompt into your AI chat to install this plugin:

Install the dsh-hermes-bridge plugin from https://github.com/dsh-pulse/dsh-hermes-bridge in DeepSeek Harness.

About this plugin

Send a WeChat message and a DSH agent starts working; when it finishes, the structured result lands back in your chat. dsh-hermes-bridge turns that once-manual pipeline into a first-class DSH plugin and fixes the pain points that kept it fragile: one-shot headless invocations burn tokens on every cold start, consecutive messages lose session context, and long-running tasks offer no completion notification. By reusing a resident per-cwd agent session (create / resume / followup), token cost drops 15-20x versus headless, and every state transition—accepted, started, done, failed—is pushed to WeChat with structured changes, verification, and leftovers, all sent through hermes send with zero LLM calls.

The plugin persists its task queue to a JSONL file on disk. After a service restart it reloads unfinished tasks, resumes the original agent session so work continues where it left off, and marks a task failed if it has been interrupted more than maxRuns times to prevent infinite loops. Paired with systemd Restart=always, you can submit tasks via WeChat in the evening, let the server restart overnight, and have the queue pick right back up—each finished result is pushed to WeChat as it lands.

Best suited for users who run a self-deployed Hermes Agent gateway and the DSH web profile and want WeChat as their sole interaction entry point while still leveraging the full DSH agent tool set. The plugin binds to 127.0.0.1 only, requires a Bearer token on every request, and provides cwd whitelisting, queue caps, LRU session-pool eviction, and outbound secret redaction. Review the environment-assumptions table in the README before wiring it into your setup.

Use Cases

  • Submit DSH tasks via WeChat at night; the queue auto-resumes after a server restart and pushes each result
  • Follow up across multiple WeChat messages while the same resident agent session retains full context
  • Receive structured changes, verification, and leftovers pushed to WeChat when a long task completes

Best For

  • Developers running a self-deployed Hermes Agent gateway with the DSH web profile
  • Users who want WeChat as the sole entry point for driving full DSH agent workflows
  • Teams focused on token cost and reducing cold-start overhead via session reuse