Overview
dsh-quicksight
Two-tier image reading for text-only models in DeepSeek Harness: fast local OCR (RapidOCR, offline) first, then a vision model via modlens as fallback.
README / EN
Package documentation
Registry summary
Two-tier image reading for text-only models in DeepSeek Harness: fast local OCR (RapidOCR, offline) first, then a vision model via modlens as fallback.
dsh.pub verifies the pinned bundle contract, runtime facts, and distribution semantics. The complete README remains in the source repository.
Read the full README on GitHubLIMITATIONS
Known limitations
- 免费视觉引擎（如 Nvidia 免费额度）存在速率限制：连续调用可能返回 429/404，稍后自动恢复；Tier-2 失败时工具会降级报错并提示配置。 - Tier-1（RapidOCR）只提取文字，不分析颜色/布局/图表/人脸；此类需求由 Tier-2 视觉模型处理。 - 支持常见图片格式（PNG/JPG/BMP/TIFF/WebP）；不支持 PDF（需先逐页转图）。