Skip to content

DesireCore v10.0.74

Choose a tag to compare

View all tags
@yi-ge yi-ge released this 06 Jun 14:55
v10.0.74
9f42c64

DesireCore v10.0.74

New Features

  • PDF Document Table of Contents Navigation: When the agent reads a long PDF, it can now first return the document’s table of contents and corresponding page numbers, helping you locate sections more quickly before continuing with in-depth reading.
  • Export Documents to PDF: The agent can now export documents directly to PDF more reliably, without relying on temporary scripts, making the export process more straightforward and dependable.

Fixes

  • Fixed formatting issues when exporting PDFs from the editor, including tables, code blocks, duplicate content panels, heading anchors, and more. Exported layouts are now cleaner and closer to the expected result.
  • Fixed instability in Agent document PDF export, restoring normal availability of the related PDF export functionality.
  • Fixed an issue where reading could get stuck or loop when continuing through long-page or oversized-content PDFs, making large-document reading more stable.

Improvements

  • Optimized the speed and stability of the agent when reading long PDFs. Responses are now faster when revisiting the same document, without repeatedly re-parsing it.
  • Improved the PDF reading strategy by prioritizing extraction of readable text and stopping promptly once the context budget is reached, reducing unnecessary waiting and preventing large documents from filling up the context window.
  • Improved the reading experience for chart-heavy pages, scanned pages, and mixed-content PDFs. Even when the current model is not a vision model, it will still try to return any extractable text first.
  • Optimized handling of very long PDFs. When a request exceeds the page limit, it no longer fails immediately; instead, it reads the first portion and prompts you to continue.

What's Changed

Full Changelog: desirecore/desirecore@v10.0.73...v10.0.74

DesireCore v10.0.74

新功能

  • PDF 文档目录导航：智能体读取长篇 PDF 时，现在可以先返回文档目录和对应页码，帮助你更快定位章节后再继续精读。
  • 文档导出为 PDF：智能体现在可以更稳定地直接把文档导出为 PDF，无需再依赖临时脚本，导出流程更直接可靠。

修复

  • 修复了编辑器导出 PDF 时表格、代码块、重复内容面板、标题锚点等格式异常问题，导出的版式更干净、更接近预期。
  • 修复了 Agent 文档导出 PDF 不稳定的问题，恢复了相关 PDF 导出能力的正常可用性。
  • 修复了长页或超大内容 PDF 在继续读取时可能出现卡住或循环的问题，读取大文档时更稳定。

改进

  • 优化了智能体读取长篇 PDF 的速度与稳定性，同一文档重复翻读时响应更快，不必反复重新解析。
  • 改进了 PDF 阅读策略，优先提取可读文字并在达到上下文预算后及时停止，减少无效等待并避免上下文被大文档撑满。
  • 改进了图表页、扫描页和混合内容 PDF 的读取体验，即使当前模型不是视觉模型，也能尽量先返回可提取的文字内容。
  • 优化了超长 PDF 的处理方式，请求超过页数限制时不再直接拒绝，而是先读取前一部分并提示你继续。

What's Changed

Full Changelog: desirecore/desirecore@v10.0.73...v10.0.74

Download the installer for your platform below. / 请在下方下载对应平台的安装包。
Official Website · All Releases

Contributors

yi-ge and Zxy-y
Assets 2
Loading