aura-vision
Run the following command in DeepSeek Harness:
dsh plugin install Ck-epsilon/aura-vision
Paste the following prompt into your AI chat to install this plugin:
On a machine where dsh is already installed, run dsh plugin install Ck-epsilon/aura-vision and restart dsh to see the Recognize button beside the input box; full source is at https://github.com/Ck-epsilon/aura-vision
About this plugin
Vision recognition inside DeepSeek Harness used to mean wiring up a paid multimodal endpoint, hand-rolling tiling and stitching logic, and worrying about how to turn results into shareable documents. aura-vision bundles all of that into a single free channel: it defaults to Zhipu GLM-4V-Flash (free tier), can point to any OpenAI-compatible endpoint, or fall back to Pollinations anonymous inference. A single click beside the input box turns a picture into text, with no extra key-service setup required.
Long documents and large screenshots are where it shines. The free tier caps output at 1024 tokens, so the plugin applies adaptive grid tiling (targeting 1100px per block, up to a 3x3 grid, 8 percent overlap) and transcribes block by block to stay within the limit while keeping content complete. Results can be exported as Markdown, Word, or a tall PNG image; tables are additionally available as Excel. History sessions store thumbnails and originals separately, support favourites, filtering, and one-click self-contained Markdown export. Data migrates with your workspace, keys live only in the local credential store, and images are sent solely to the backend you choose.
If you regularly handle screenshots, scanned contracts, or whiteboard photos inside Harness sessions, or simply want a zero-cost, plug-and-play vision entry point, aura-vision makes image-to-text as natural as copy-paste, as if it had always been there.
Use Cases
- Transcribe scanned contract tables directly within a Harness session
- Convert whiteboard photos into editable Markdown documents
- Export recognition results as Word or Excel for archiving in one click
Best For
- Developers who frequently process images inside DeepSeek Harness
- Individual users seeking a zero-cost vision recognition pipeline
- Privacy-sensitive users who want images routed only to a chosen backend
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.