AI Agent Hub
Back to plugins
🖥️

dsh-session-turn-eta

Client Updated 2026.09.14

Run the following command in DeepSeek Harness:

dsh plugin install 778672151/dsh-session-turn-eta

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

Install the plugin by running dsh plugin install 778672151/dsh-session-turn-eta inside DeepSeek Harness; the source repository is at https://github.com/778672151/dsh-session-turn-eta.

About this plugin

When a DeepSeek Harness turn chains multiple steps it can easily last several minutes, yet the interface offers no progress feedback at all. dsh-session-turn-eta renders a live progress bar under the input composer, showing a ticking percentage and a human-readable time-remaining label, turning blind waiting into something tangible.

The prediction is not a simple linear extrapolation. It blends the current turn observed per-step rate with the session median historical step duration, so one long idle gap cannot skew the estimate, and it re-anchors the bar the moment a turn outruns its own projection instead of freezing at a misleading zero or 100 percent. Once the turn exceeds the estimate the bar fills and pulses with a finishing label; if the turn fails or is interrupted the bar turns red and reads progress interrupted.

It is aimed at developers who regularly fire multi-step, long-running DSH instructions and want to cut down on the anxiety of repeated blind waits during iterative development. The plugin works entirely through read-only session projections, mutates no session state, registers no model-visible capability, and ships prebuilt so installation requires no build step.

Use Cases

  • Multi-step turns take minutes and you want a live remaining-time readout
  • A turn outruns its estimate and you need honest status feedback instead of a misleading zero
  • Iterative development with frequent long-running generations where wait anxiety is a real cost

Best For

  • Developers who frequently run multi-step, long-running DSH instructions
  • Product or front-end engineers who care about the client-side wait experience
  • Technical users triggering complex generation workflows repeatedly during iteration