Skip to content

Update si_wei_dao_tu_sheng_cheng_qi to v0.0.23 - #3036

Merged
crazywoola merged 1 commit into
langgenius:mainfrom
zhangyuqz:codex/mind-map-0.0.23
Sep 11, 2026
Merged

Update si_wei_dao_tu_sheng_cheng_qi to v0.0.23#3036
crazywoola merged 1 commit into
langgenius:mainfrom
zhangyuqz:codex/mind-map-0.0.23

Conversation

@zhangyuqz

Copy link
Copy Markdown
Contributor

Plugin Submission

Plugin information

Submission type

  • New plugin
  • Version update

What changed

This updates the published 0.0.22 plugin to 0.0.23.

  • Fix blurry, very tall mind maps: replace the approximately 8,300-pixel long-edge cap with a total-pixel budget, retaining a maximum-side safeguard and remeasuring the layout at the actual drawing scale.
  • Preserve PNG resolution for longer when meeting the requested image-size limit: compression and removal of redundant opaque alpha come first, then gradual RGB precision reduction, then proportional resizing if still necessary.
  • Handle oversized WebP output within the format's dimension limit and release the completed Agg canvas before encoding.
  • Reorganize both READMEs into previews, setup, parameters, fonts, output guidance, and troubleshooting. Keep all five requested font previews and both multilingual test links.
  • Preserve plugin identity, all 17 fonts and their licenses, font-size adjustments and ordering, shape-selection rules, themes, startup code, dependencies, rendering concurrency control, and one-image output.

The recommended settings for very large text-heavy maps are PNG, render_scale=1.0, and max_blob_mb=4.5. This does not remove pixel, format, or file-size limits.

Risk level

  • Low risk
  • Medium risk
  • High risk

Conservative classification: the renderer processes arbitrary user-supplied text, which can include sensitive information, and creates local temporary image files. It does not expose general-purpose filesystem access, execute user-provided code, or introduce any new sensitive capability in this release.

Required checks

  • Package and documentation were checked against the current submission requirements.
  • Maintainer has reviewed and accepts the Plugin Developer Agreement.
  • Local tests and the limitations of Dify CE/Cloud validation are documented below.
  • Contents are limited to runtime source, metadata, fonts, and required review/license documentation.
  • No credentials, environment files, Git data, virtual environments, caches, logs, IDE files, or private test material are included.
  • No executable programs are bundled. Required binary font/image assets are explained below.
  • The guides cover installation, usage, dependencies, network requirements, credentials, and source availability.
  • The nonempty PRIVACY.md is referenced by manifest.yaml.
  • The primary README is entirely English; the Chinese guide is readme/README_zh_Hans.md.

Security and privacy notes

Rendering runs inside the Dify plugin runtime, without an external rendering service, arbitrary URL fetching, SQL, shell execution, browser automation, or a user-supplied network destination. Markdown is treated as text to draw, not as executable code.

Input may contain personal or sensitive information chosen by the user. The plugin uses that input for the requested image, does not maintain its own content database, and removes temporary output files after execution. Dify's own storage and logging remain deployment responsibilities. New resolution diagnostics do not log the Markdown body.

Installation requires the declared Python dependencies. Externally hosted README images are documentation only and are not needed by the renderer.

Local validation

Official packaging tool: Dify Plugin CLI 0.6.10. Local workspace commands:

./work/implementation_v020/tools/dify-plugin-windows-amd64-0.6.10.exe version
./work/implementation_v020/tools/dify-plugin-windows-amd64-0.6.10.exe plugin package ./work/marketplace_0.0.23_20260908/source --max-size 50 -o ./work/marketplace_0.0.23_20260908/release/zhangyu-si_wei_dao_tu_sheng_cheng_qi_0.0.23.difypkg
./work/font_modes_v0.0.24/dev_venv/Scripts/python.exe ./work/marketplace_0.0.23_20260908/verification/audit_market_package.py

Results:

  • Official packaging succeeded; CRC and exact package-to-source comparison passed.
  • 52 members, 31,326,343 compressed bytes, 47,748,476 unpacked bytes; unpacked contents are below 50 MB.
  • SHA-256: 5e67dfc09cd89d09806b110692ba6e7d2153a5b03eeb8f6e4a47dd4ed5b8add9
  • The English README has zero Han, Hiragana, Katakana, or Hangul characters. Both guides were opened in a real browser and checked for image proportions, table layout, navigation, language switching, and reload behavior.
  • The final documentation repack changes only the two READMEs relative to the preceding tested 0.0.23 repair package. All 50 other members, including every runtime file, dependency declaration, and font, are byte-identical.
  • On that identical runtime, a fresh Linux Python 3.12 environment installed the declared dependencies, including dify_plugin 0.10.2, and passed the official Marketplace Toolkit test-plugin-install.py startup check at toolkit commit 57a21d1304b1108df3e6b90a15a4f5dd9f0915f9.
  • Actual package invocation with the original main-entrypoint warmup produced a 1,933 x 35,612 PNG for a 211-node case. Native-scale top, middle, bottom, and root crops were inspected; geometry checks found no overlapping nodes or out-of-node text in this PNG case.
  • A separate multilingual sample with Yeseva One and the light theme rendered 44 nodes, including rounded rectangles, circles, horizontal ellipses, and a vertical ellipse.
  • The existing Chinese optical-size sample produced byte-identical PNG, WebP, and GIF output compared with 0.0.22.

These local rendering and startup results are carried forward because the runtime is identical. This release has not been installed and tested end-to-end on a live Dify Community Edition or Dify Cloud deployment during this release-preparation run; no such result is claimed.

Reviewer notes

This PR should add exactly one file:

zhangyuqz/si_wei_dao_tu_sheng_cheng_qi/zhangyu-si_wei_dao_tu_sheng_cheng_qi_0.0.23.difypkg

Previously published packages remain untouched. The manifest author stays zhangyu; the repository account/directory stays zhangyuqz. The filename has no browser-added duplicate suffix.

Bundled binary data consists of the plugin icon and 17 open-source font files required for offline rendering. Font licenses and notices are included; no executables or bundled Python environments are shipped. The SDK requirement remains dify_plugin>=0.9.0 and was actually installed for the Linux check.

No breaking parameter or identity change is intended. Larger retained images can use more time and memory. WebP retains its side limit; GIF retains its existing frame budget. The already-existing very-large-GIF low-resolution/text-fit limitations were not fixed by this PNG-focused release. Use PNG for large text-heavy maps.

Private regression inputs and their generated images are not part of this package or this PR.

@github-actions github-actions Bot added the risk: high High-risk Marketplace submission label Sep 8, 2026
@zhangyuqz zhangyuqz changed the title Version update: zhangyu/si_wei_dao_tu_sheng_cheng_qi 0.0.23 Update si_wei_dao_tu_sheng_cheng_qi to v0.0.23 Sep 8, 2026
@crazywoola
crazywoola merged commit 430974c into langgenius:main Sep 11, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

risk: high High-risk Marketplace submission

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants