DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A CET-4/6 passing score calculator that computes results precisely from given parameters, pure Node, no network dependency.
A pure Node.js plugin that precisely estimates exercise heart-rate zones from given parameters and returns detailed results, with no network dependency (estimation only, not medical advice).
Calculates optimal sleep-onset timing from sleep-cycle parameters and returns detailed estimates. Pure Node, no network dependency. Not medical advice.
Installs the Qoder CN provider adapter for DeepSeek Harness, enabling COSY-signed WAF-encoded streaming inference with PAT-based auth and automatic live model catalog discovery.
A universal AI image-generation plugin for DeepSeek Harness that connects to any OpenAI-compatible gateway with visual settings and one-line prompting.
Bridge the latest Zhipu coding model GLM-5.3 into DSH via the existing zai-coding-cn route, before the pi-ai catalog catches up.
Paces LLM requests per provider with RPM token buckets, concurrency control, bounded FIFO queues, and adaptive cooldown to prevent HTTP 429 rate-limit errors in DSH.
Interactively collect project details and auto-generate a nine-chapter Feishu bid-proposal cloud document covering background, analysis, solution, pricing, team and case sections.
Local, offline semantic code search for DeepSeek Harness with hybrid vector+BM25 retrieval, symbol-aware chunking, and CJK-aware tokenization.
Configures the model-request retry budget for active providers in DSH Settings via the settings API, while backoff and retry execution remain owned by DSH.
A PPT page count suggestion tool that precisely calculates recommended pages based on parameters and returns detailed results, fully local with no network dependency.
A pure-Node, zero-network rental insurance advisor plugin that returns clear coverage recommendations based on input parameters.