dsh-llm-vision-bridge
A DSH LLM bridge that turns image attachments into vision-model descriptions before sending them to a text-only DeepSeek provider.
Install
$ dsh plugin --profile web add github:Einskyle/dsh-llm-vision-bridgePlugin 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 the `deepseek-vision` LLM provider with text and image input modalities.
- Routes image blocks to a configured vision provider and replaces them with text descriptions before fallback LLM processing.
- Passes image-free requests directly to the configured fallback provider.
- Caches image descriptions using an LRU cache for identical image-and-prompt inputs.
- Retries retryable vision failures including 503, 429, and timeout errors according to configuration.
- Provides configurable placeholder or error behavior after a final vision failure.
- Supports pi-ai routes and other OpenAI-compatible vision endpoints.
Useful For
- Enable image-based chat questions while retaining a text-only DeepSeek model for final responses.
- Connect a local llama.cpp Qwen3-VL route to DSH's web chat workflow.
- Use cloud OpenAI-compatible vision endpoints with a DeepSeek fallback provider.
- Avoid repeated image interpretation during replay or session compaction.
Who It Fits
- DSH web GUI users running text-only DeepSeek providers.
- Users with configured pi-ai, llama.cpp, or OpenAI-compatible vision model routes.
- Developers configuring DSH model providers and settings.
Documented Limitations
- A vision-capable provider route must be configured separately.
- To submit images, the active session model must be `deepseek-vision`; `autoRoute` cannot bypass DSH image admission checks.
- The vision provider cannot be `deepseek-vision` itself due to the recursion guard.
- Cloud vision routes require credentials when their provider declares `apiKeyEnv`.
- Vision response time can include a reported 10–40 second Qwen3-VL cold load.
- The package requires Node.js `^22.19.0 || >=24.0.0`.
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, 4 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
- Einskyle/dsh-llm-vision-bridge
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- 5a8e317 · 2026-09-13
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
- 3GitHub stars
- Forks
- 0
- Open issues
- 1
- Last commit
- 2026-08-14
- Last release
- No release detected
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 →Local-first AI agent runtime and DSH plugin bundle providing sandboxed sessions, MCP tools, audit logs, and session replay.
View plugin →Preview, create, and edit spreadsheets, docs, slides, and canvases in DeepSeek Harness powered by Univer.
View plugin →Vision routing and pixel-level visual tool suite for DeepSeek Harness agents with built-in free fallbacks.
View plugin →