DSH Image Vision Bridge
Icestab/dsh-image-vision-bridge · model-adapters
DSH Plugin: Chat hair map automatically to visual model (mimo-v2.5) analysis, text description is passed quietly to pure main text model and chat records are kept as shown.
★ 2GitHub stars
0Forks
2026-08-25Last updated
JavaScriptLanguage
MITLicense
Key features
- The goal is to set up the visual model of the plugin configuration (the call itself, or a session that already uses visual models) so that it can see directly.
- Other cases Construct a rewrite request: the message's image block (with tool-result nesting) is replaced by a description text, which is then called in the actual flow of this request.
- The original requested object and session log are completely unmodified.
- The description results are attached id in the process cache (retest/subsequent rounds are not repeated); visual calls are downgraded to a failed text without interrupting the conversation; and user cancellations are routinely thrown out.
Requirements
- Tested environmentDeepSeek Harness 0.1.0-rc.6 and 0.1.1-rc.2 (@deepseek-ai/dsh, dsh-llm, coldis 4.0.1), 11 offline self-measures were passed.
- Public APIctx.llm.stream, llm/stream water line, contentHasImage/createMessesage/BlockAssembler, attachment image block.
Install command
dsh plugin --profile web add github:Icestab/dsh-image-vision-bridge