Highlights
- Import PDF requirement documents and technical designs, including scanned image-only PDFs. The browser converts each page into extractable text and a bounded JPEG, then sends both to the selected Harness AI model for Markdown PRD or technical-design generation.
- Choose whether imported content replaces or appends to the editor, with staged progress, cancellation, sampling and truncation warnings, and no implicit Project save or planning action.
- Delete Projects with complete cleanup of Project-owned execution and collaboration records while preserving shared Agents, Squads, Runtimes, and Harness Workspaces.
- Support the project through the new WeChat Pay and Alipay links in the README.
Security and reliability
- PDF imports enforce canonical base64, image admission limits, model image capability, same-origin JSON requests, bounded concurrency, request cancellation, and a three-minute timeout.
- PDF content is treated as untrusted evidence and cannot enable tools or override the requirement-analysis contract.
- Failed child-record cleanup keeps the Project record available for a safe deletion retry.
Verification
- TypeScript typecheck
- Documentation smoke checks
- 96 automated tests
- npm package smoke test for the 58-file release artifact
- Desktop and mobile Chrome verification of the live PDF conversion and upload payload
Full Changelog: v1.3.4...v1.3.5