dsh-reconnect

When running AI agent tasks powered by DeepSeek Harness, have you ever had your workflow interrupted by a flickering network, a disconnected relay proxy, or a temporary outage at the API provider? The dsh-reconnect plugin is built to solve exactly this. Acting as a host-side plugin, it steps in after the built-in retry policies give up, intelligently resending failed requests to overcome transient issues like relay node failures, rate limits, server 5xx errors, and timeouts. This ensures that your workflows and long-running tasks maintain resilience against short-term external instabilities.

At its core, dsh-reconnect offers a smart, tiered retry strategy combined with effortless configuration. It distinguishes between recoverable errors (e.g., dropped connections, timeouts) and permanent ones (e.g., invalid credentials), applying an exponential backoff (from 1s up to a capped maximum) for the former. Through a unified visual settings panel, you can easily tune parameters like the maximum delay per attempt or whether to retry quota errors, with changes taking effect instantly. This makes it an invaluable tool for developers and operators managing complex proxy chains, API gateways, or any scenario where guaranteeing the completion of lengthy, critical processes is a priority.