AI Agent Hub
Back to plugins
🧰

dsh-tool-terminal-search

Web Tools Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install ystyle/dsh-tool-terminal-search

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

Install the plugin by running dsh plugin install ystyle/dsh-tool-terminal-search in DeepSeek Harness; full source is available at https://github.com/ystyle/dsh-tool-terminal-search .

About this plugin

When terminal scrollback grows to tens of thousands of log lines, a model can only page backward one chunk at a time from the newest output, and totalLines tells it the buffer size but not where the critical error line lives. dsh-tool-terminal-search closes that gap by letting the model search by content instead of paging blindly.

terminal_search scans the entire retained buffer line by line, matching a literal substring (grep -F style) by default or a JavaScript regular expression when regex is enabled. Matches are returned newest-first, each carrying a newest-relative offset that feeds directly into a follow-up terminal_read call for context. Repeated regex patterns are served from a process-wide bounded FIFO cache so they are never recompiled. The tool composes exclusively over the public ctx.terminals.read seam, requiring no upstream package changes and running on today's published dsh releases.

If you are running long-lived backend processes under DeepSeek Harness and need to pinpoint an error line or stack trace in a massive log, or if you want content-based search capability before the native upstream implementation lands, this plugin offers a drop-in solution.

Use Cases

  • Pinpoint a specific error message among tens of thousands of terminal log lines
  • Batch-match output lines using regular expression patterns
  • Search key log entries by content in long-running backend processes

Best For

  • Developers running long-lived backend services under DeepSeek Harness
  • Analysts who need to locate entries in large terminal scrollback buffers
  • Teams wanting content-based search before the native upstream implementation ships