Skip to content

Releases: c-ling/dsh-plugin-vision-bridge

Release list

v1.0.1

Choose a tag to compare

@c-ling c-ling released this 17 Aug 11:19

v1.0.1

更新内容(中文)

  • 修复:设置页「视觉识别」中的「查看日志」按钮改为紧凑尺寸,避免撑高卡片 header。

Changes (English)

  • Fixed: Made the "View logs" button on the Vision Recognition settings page compact so it no longer stretches the card header.

v1.0.0

Choose a tag to compare

@c-ling c-ling released this 17 Aug 11:05

v1.0.0

更新内容(中文)

  • 新增:DeepSeek Harness 视觉识别插件,支持在会话输入框粘贴/拖拽/按钮上传图片。
  • 新增:发送时自动通过 OpenAI 兼容视觉 API 生成英文图片描述,DeepSeek 只接收纯文本。
  • 新增:describe_image 工具,模型可重新观察已上传图片,支持单图/多图与自定义 prompt。
  • 新增:设置页(Base URL、模型、只写 API Key、提示词模板、超时、图片限制、查看日志)。
  • 新增:完整中英文界面,随 Settings → General → Language 实时切换。
  • 新增:视觉 API 失败时保留输入草稿并明确报错,不静默降级。
  • 文档:中文 README.md + 英文 README-en.md,包含安装/更新/卸载/验证说明。

Changes (English)

  • Added: Vision bridge plugin for DeepSeek Harness; paste/drop/upload images directly in the composer.
  • Added: Automatic English image descriptions via an OpenAI-compatible vision API before DeepSeek receives the text-only prompt.
  • Added: describe_image tool for the model to re-inspect uploaded images, including single/multi-image calls and optional prompt override.
  • Added: Settings page (Base URL, model, write-only API key, prompt template, timeout, image limits, log viewer).
  • Added: Full bilingual UI that switches live with Settings → General → Language.
  • Added: Vision API failures restore the composer draft and surface an explicit error — no silent degradation.
  • Docs: Chinese README.md + English README-en.md with install/update/uninstall/verify instructions.