dsh-vision-proxy
A DeepSeek Harness provider plugin that transcribes attached images with an OpenAI-compatible vision model before sending text to DeepSeek.
Install
$ dsh plugin --profile web add dsh-vision-proxyPlugin 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
- Registers a DeepSeek Harness `deepseek-vision` provider route that declares image input support.
- Transcribes attached images through OpenAI-compatible VLM `/chat/completions` endpoints before forwarding to DeepSeek.
- Supports configurable VLM backends, including DashScope/Qwen, Zhipu, OpenRouter, custom compatible endpoints, and local Ollama.
- Can auto-detect local Ollama and select a local vision model.
- Provides fallback-model chains with per-entry endpoint, model, key, anonymous-mode, and timeout settings.
- Caches image transcription results in memory by SHA-256 content hash.
- Optionally downsamples large images when the optional `sharp` dependency is available.
- Supports configurable failure behavior, including placeholder text so a failed transcription does not stop the conversation.
Useful For
- Using DeepSeek text-model conversations with GUI image attachments.
- Having a VLM describe screenshots, images, layouts, or OCR-relevant content before DeepSeek answers.
- Keeping image processing local by using a local Ollama-compatible endpoint.
- Configuring multiple vision backends for fallback when an endpoint is unavailable or rate-limited.
Who It Fits
- DeepSeek Harness users who want image attachments in DeepSeek conversations.
- Users with access to an OpenAI-compatible vision-language model endpoint.
- Privacy-conscious users who can run a local Ollama vision model.
Documented Limitations
- Image bytes are sent to the configured VLM endpoint unless a local endpoint such as Ollama is used.
- The plugin requires DeepSeek Harness >=0.1.0-rc.6 and Node.js >=22.19.0.
- Without an API key or detected local Ollama vision model, image transcription cannot succeed.
- Anonymous endpoints are capped at 20 seconds and may fail quickly on HTTP 429.
- Image transcription quality, including small text in dense UI screenshots, depends on the selected vision model.
- The transcription cache is in-process only and is not persisted.
- pnpm >=10 may require allowing build scripts for this package and optional `sharp` before installation completes.
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 1 runtime, 0 development, 0 peer, and 1 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.
package.json declares a postinstall lifecycle script that may run during relevant package installation workflows.
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
- Flyvhidbwo/dsh-vision-proxy
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- d4ac262 · 2026-09-25
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
- 14GitHub stars
- Forks
- 0
- Open issues
- 0
- Last commit
- 2026-08-26
- Last release
- 2026-08-22
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Compatibility bridge and host ABI enabling unmodified Pi plugins to run natively on DeepSeek Harness.
View plugin →Preview, create, and edit spreadsheets, docs, slides, and canvases in DeepSeek Harness powered by Univer.
View plugin →Local-first AI agent runtime and DSH plugin bundle providing sandboxed sessions, MCP tools, audit logs, and session replay.
View plugin →Vision routing and pixel-level visual tool suite for DeepSeek Harness agents with built-in free fallbacks.
View plugin →