DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A standalone Cordis plugin that mounts an AWS Kiro LLM provider onto the DeepSeek Harness, reusing existing kiro-cli or Kiro IDE credentials with silent token refresh and 15-model support.
Sync cc-switch provider profiles to DSH routes to bootstrap with zero models configured.
Adds image understanding, OCR, and persistent visual evidence to DeepSeek text-only models via the see_image tool.
A DeepSeek Harness plugin leveraging macOS Vision for local OCR, extracting image text without network or API keys.
Langfuse observability for DeepSeek Harness, tracking sessions, token usage, and subagent lineage.
Configures independent processing chains for images, videos, and audio to convert them into text prompts for session models.
Syncs live model lists from providers to DeepSeek Harness, tracking 5h/7d usage windows and API balance.
Intercepts images for OCR via vision models, auto-sending recognized text, supporting local file reading and multi-turn dialogue.
Automatically populates context window, token limits, and multimodal features for custom providers in DeepSeek Harness, resolving missing model specifications.
Provides local vision capabilities for text-only models with zero API costs and full privacy, featuring structured evidence extraction.
Integrates Xiaohongshu and social media creator AI workflows into DeepSeek Harness for content production and knowledge management.
A diagramming agent skill for DeepSeek Harness that generates architecture, workflow, sequence, dataflow, and lifecycle diagrams from JSON specs, outputting standalone HTML with validation.