dsh-image-vision-bridge
A DeepSeek Harness host plugin that converts chat images into vision-model descriptions for text-only main models while preserving original images in the visible transcript.
Install
$ dsh plugin --profile web add dsh-image-vision-bridge # npmPlugin 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.
Key Capabilities
- Detects image blocks in LLM streaming requests, including nested tool-result content.
- Calls a configured vision model, defaulting to mimo-v2.5 through the opencode-go provider.
- Replaces image blocks with generated text descriptions in rewritten model requests.
- Preserves the original images and avoids adding vision-description text to the visible chat transcript or session log.
- Caches descriptions by attachment ID within the process to avoid repeated vision calls.
- Passes through requests without images and requests already targeting the configured vision model.
- Falls back to failure-description text if a vision call fails, without interrupting the conversation.
Useful For
- Enable image-aware conversations with a primary model that receives text descriptions rather than image blocks.
- Keep uploaded images visible in chat while supplying private vision context to the responding model.
- Use a separate configured vision model for image analysis in DSH web profiles.
Who It Fits
- DeepSeek Harness users running the web profile.
- Users with an existing vision-capable model route, such as opencode-go with mimo-v2.5.
- Developers who want configurable image-to-text bridging in the DSH LLM stream pipeline.
Documented Limitations
- The main model must be configured with image input modality in DSH settings so the host API boundary accepts image messages, even though the plugin later converts images to text.
- The configured vision provider and model must already be registered in settings.yaml.
- The web profile requires a dsh web restart because HMR is disabled.
- The plugin rewrites requests entering the llm/stream pipeline; the README states this is not an officially guaranteed framework usage and recommends smoke-testing before DSH upgrades.
- Description caching is in-process only.
- The documented tested environment is DeepSeek Harness and dsh-llm 0.1.0-rc.6 with Cordis 4.0.1.
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 0 runtime, 0 development, 2 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- Icestab/dsh-image-vision-bridge
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- 61b3602 · 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
- 1GitHub stars
- Forks
- 0
- Open issues
- 0
- Last commit
- 2026-08-22
- Last release
- No release detected
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
DeepSeek Harness-native vision toolkit for image Q&A, OCR, grounding, UI restoration, pixel comparison, and visual artifacts.
View plugin →Vision routing and pixel-level visual tool suite for DeepSeek Harness agents with built-in free fallbacks.
View plugin →A DeepSeek Harness vision plugin that turns pasted images into structured OCR, layout, and semantic evidence for supported text-only models.
View plugin →DeepSeek Harness web-search plugin that adds web search, X search, and focused page fetching with structured evidence.
View plugin →