dsh-llm-qodersdk

Configuring local large language model providers in AI development often involves managing API keys and settings, adding unnecessary complexity. This plugin solves that by allowing DeepSeek Harness users to directly reuse their local Qoder CLI login session—no configuration required—seamlessly routing LLM interfaces to Qoder CLI.

Its core strength lies in the intelligent adapter layer: it offers a plug-and-play experience with dual routing (separately managing built-in and custom models), and efficiently handles dialogue continuation and tool calls through persistent sessions. The plugin bridges host tools via MCP protocol for transparent tool-result relay, while actively fetching the Qoder CLI model catalog. It also manages context windows, compaction thresholds, and overflow recovery, ensuring synergy with Harness's compaction engine to avoid session-state conflicts.

This solution is ideal for users who already have Qoder CLI installed and logged in—whether individual developers, AI tool enthusiasts, or teams needing quick multi-model integration. It streamlines workflows, letting you focus on AI applications themselves rather than underlying configurations, enabling smoother exploration and use of Qoder models within DeepSeek Harness.