← DeepSeek Harness Plugins
DeepSeek Harness PluginManifest Valid

dsh-tool-visual-primitives

Enables visual reasoning for text-only models in DeepSeek Harness by routing images to external vision models and generating coordinate-based visual primitives.

VisionUI & ProductivityTerminal & TUIDeveloper Tools
2GitHub Stars0ForksUpdated2026-09-29

Install

$ dsh plugin --profile web add dsh-tool-visual-primitives@latest

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-tool-visual-primitives bridges multimodal visual analysis to text-only models within DeepSeek Harness (DSH). Inspired by DeepSeek's 'Thinking with Visual Primitives' research, it analyzes images via an OpenAI-compatible vision endpoint and converts visual content into structured textual evidence with normalized bounding boxes and coordinates (0-999). This allows non-multimodal models to interpret images, screenshots, and diagrams either through explicit tool calls (`vision_analyze`) or integrated `[vision]` conversation model aliases.

Key Capabilities

  • Provides explicit `vision_analyze` tool calling for local image paths and remote URLs
  • Creates `[vision]` chat model aliases to automatically intercept and analyze uploaded images for pure-text models
  • Supports 11 automatic vision analysis modes including captioning, object inventory, grounding, spatial relations, UI analysis, and document parsing
  • Outputs structured spatial visual primitives with normalized 0-999 bounding boxes (<box>) and reference points (<point>)
  • Offers 3 detail levels (brief, standard, verbose) and session-level visual evidence caching
  • Includes a native DSH web settings UI for managing API keys, testing connections, and searching model catalogs

Useful For

  • Enabling text-only LLMs in DSH to understand uploaded screenshots, charts, and diagrams
  • Extracting coordinate-based spatial data and UI element layout for automated code generation or UI replication
  • Analyzing document visual structures and relationships without requiring native multimodal support in the primary reasoning model

Who It Fits

  • DeepSeek Harness users relying on text-only reasoning models
  • Developers building visual grounding, diagram parsing, or UI replication workflows in DSH
  • Teams needing cost-effective multimodal pipelines by pairing specialized vision APIs with standard text LLMs

Documented Limitations

  • Requires DeepSeek Harness 0.1.2-rc.1 or higher on the web profile
  • Requires an external OpenAI-compatible visual model API endpoint and corresponding API credentials
  • Enforces a 10MB maximum image size limit and restricts remote URLs from accessing localhost or private networks

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.

Dependency Counts

package.json declares 1 runtime, 1 development, 1 peer, and 0 optional dependencies.

info
Dsh Bundle Declared

package.json declares DSH bundle metadata.

info
License Declared

GitHub reports the repository license as MIT.

info
Package Manifest Available

A root package.json was captured and can be inspected by the registry.

info
Source Available

Public GitHub source metadata is available for this registry snapshot.

info

Source & Registry Notes

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

Source repository
InkshadeWoods/dsh-tool-visual-primitives
Registry source
GitHub · dsh-plugin topic
Registry classification
Plugin
Source checked
0b29046 · 2026-10-04

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
2GitHub stars
Forks
0
Open issues
0
Last commit
2026-09-29
Last release
2026-09-29
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