luobosibing2/dsh-jev-plugin
plugin
Native DeepSeek Harness (DSH) plugin integrating TypeSafe Jev or Decision api like luna as a System One decision layer for agent selection, supervision, corrections, and approvals.
Install
npx @deepseek-ai/dsh plugin --profile jev add ./dist/dsh-jev-plugin-0.1.0.tgzRestart `dsh web` after install. Bundle APIs can change during the developer preview.
README badge
[](https://dshhub.dev/plugins/dsh-jev-plugin)Paste this into your README. The star count updates with every catalog sync.
From the README
Excerpt from luobosibing2/dsh-jev-plugin, cleaned of badges and images.
deepseek-harness-jev
Native DeepSeek Harness (DSH) plugin with Jev or Luna Decisions as its judgment model.
deepseek-harness-jev connects DeepSeek Harness (DSH) to Jev by TypeSafe AI for agent skill and file selection, task supervision, shared-finding corrections, tool-output filtering, single-operation approval assistance, and historical stage navigation. Its 12 features are individually configurable from one Jev settings page and are all disabled by default.
The main model continues to plan, generate answers, and call native tools. Enabled features use the saved Jev or Luna Decisions connection at DSH extension points for skill catalogs, agent lifecycle, tool results, and approvals, then apply results according to each feature. DSH configures the main model separately. Integration uses public Cordis / DSH plugin APIs without modifying the host source.
This is an independent community project, not an official DeepSeek or Jev release. It is an early-stage plugin with a bounded DSH compatibility range; its APIs and model judgments are not a correctness guarantee.
The Chinese feature website explains each DSH integration point, the information sent to the judgment service, and the observed test cases and limits.
The Chinese principles and all features explainer covers all 12 features and includes a complete skill-selection input/output example. Its Markdown source is maintained in this repository.
The quick review directory links to short source guides. The Chinese log-admission interview guide covers generic and test logs, design choices, live-model results and their limits; its Markdown source is kept here.
The Chinese Goal supervision and completion review follows the two current flows from their hooks through judgments, adoption, limited supplements and delivery records. Its Markdown source separates current behavior from historical live-model results.
main supports Jev and Luna Decisions through OpenRouter or OpenAI, with three separate saved connections per profile. The 2026-10-07 public report separates fixed protocol/installation checks, controlled live samples, and real DeepSeek Flash tasks. Official OpenAI calls succeeded in these limited samples; the only live OpenRouter diagnostic returned a provider-access HTTP 403 and was stopped. These results do not establish general judgment accuracy or stable task benefit.
…
