ArchivedThis repository is archived and may no longer be actively maintained.
DeepSeek Harness Plugins
DeepSeek Harness PluginManifest Valid

dsh-windows-ocr

OCR attached images locally using the built-in Windows OCR engine before sending text to DeepSeek Harness models.

VisionUI & ProductivityTerminal & TUIDeveloper ToolsSecurity & PolicySkills & WorkflowsRemote Execution
9GitHub Stars1ForksUpdated2026-09-13

Install

$ dsh plugin --profile web add dsh-windows-ocr

Plugin Overview & Capabilities

AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.

source-grounded
dsh-windows-ocr is a DeepSeek Harness plugin that allows text-only language models to process attached images. By hooking into the Cordis llm service and pre-step lifecycle, it performs local optical character recognition (OCR) via Windows.Media.Ocr and substitutes image data with recognized text blocks before outgoing requests leave the machine.

Key Capabilities

  • Performs local OCR processing on attached images via Windows.Media.Ocr without sending image bytes to model providers
  • Provides a capability shim to allow text-only models to accept image attachments at admission time
  • Intercepts agent/pre-step lifecycle to replace image blocks with structured <image_ocr> text blocks
  • Supports configurable language selection via BCP-47 tags and optional vision-model passthrough
  • Implements process-level caching for OCR results keyed by attachment id and automatic temp-file cleanup

Useful For

  • Using text-only LLMs with screenshots, receipts, or document images while maintaining privacy
  • Extracting text from images without paying for or requiring multimodal model vision APIs
  • Preventing sensitive image files from leaving the local machine during model inference on Windows

Who It Fits

  • Windows users of DeepSeek Harness who want local OCR processing
  • Developers working with text-only models who need to extract text from images
  • Privacy-conscious users wanting to prevent raw image serialization to remote providers

Documented Limitations

  • Requires Windows 10 or 11 with appropriate Windows OCR language packs installed
  • Animated GIFs only have their first frame recognized
  • OCR text cache is in-memory per process up to maxCacheEntries (default 200)
  • Fails closed: if the plugin is unloaded or uninstalled, text models will refuse image attachments

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

No automated security signals have been published yet.

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository
maxwell-feng/dsh-windows-ocr
Registry source
GitHub · dsh-plugin topic
Registry classification
Plugin
Source checked
e20ebb6 · 2026-09-14

This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.

Repository Activity

GitHub stars
9GitHub stars
Forks
1
Open issues
0
Last commit
2026-09-13
Last release
2026-09-13
For maintainers

Maintaining this plugin?

This listing is generated from public repository data. If you maintain this project, you can review the information and share this listing with your users if you find it useful.

Add to README
Listed on DSHPlugin.app