Skip to content

v0.3.1 Prompt Fidelity and API Compatibility / 提示词保真与 API 兼容

Choose a tag to compare

View all tags
@papperrollinggery papperrollinggery released this 16 Jun 15:47
· 16 commits to main since this release
v0.3.1
953bd40

English

Compared with v0.3.0, v0.3.1 is a blocker fix and fidelity polish release:

  • Keeps the v0.3.0 image-to-prompt reconstruction improvements while reducing over-cleaned, overly generic, or AI-polished outputs.
  • Improves source-image fidelity for soft focus, haze, film blur, low contrast, depth-of-field, lighting texture, and intentional low-fidelity looks.
  • Improves visible-person, material, surface, and layout consistency so key source details are less likely to be dropped.
  • Adds in-extension update notices and upgrade guidance, including GitHub release checks, update-note links, and unpacked-install reload instructions.
  • Improves compatibility with GPT-5-style OpenAI-compatible endpoints and the recommended BridgeDeck local workflow.
  • Adds regression coverage for model request option compatibility.

Thanks to FermiMAO for reporting the GPT-5.5 compatibility issue in #1. Thanks also to everyone who opens issues, sends PRs, and helps verify model compatibility.

中文

相比 v0.3.0，v0.3.1 是一次 release blocker 修复和保真度补强版本：

  • 保留 v0.3.0 的图片反推提示词复刻能力，同时减少过度清理、过度泛化和 AI 精修感输出。
  • 提升对虚焦、雾化、胶片模糊、低对比、景深、光线质感和刻意低清风格的原图保真。
  • 改进真人、材质、表面关系和版式细节的一致性，降低关键可见信息丢失概率。
  • 增加插件内更新提示和升级指引，包括 GitHub Release 检查、更新说明入口和本地加载插件的重新加载说明。
  • 改进 GPT-5 风格 OpenAI-compatible endpoint 和推荐 BridgeDeck 本地流程的兼容性。
  • 增加模型请求选项兼容性的回归检查。

感谢 FermiMAO 在 #1 中报告 GPT-5.5 兼容问题。也感谢所有提交 issue、PR 和帮助验证模型兼容性的用户。

Verification

  • check:api-options, typecheck, build, check:storage, check:url, check:prompt-goal, release:package, release:check, and diff whitespace check passed.
  • BridgeDeck live smoke passed with gpt-5.5, image_url input, and the v0.3.1 release package configuration.
  • Asset: zhijuan-prompt-card-0.3.1.zip
Assets 3
Loading