v0.3.6

Asset SHA256: 011dca4bab80d88b7fe492211bb3e3b30d53f09a61709d77bef0f12367e7f8ec

English

This release adds a low-key creator support entry to the floating panel.

• Adds a small coffee support button in the floating panel header.
• Shows an optional Alipay support QR code and the maintainer WeChat contact.
• Clarifies that support is voluntary and does not change extension features, grant a commercial license, create paid support, or gate community group access.
• Tightens the support popover layout, copy, and paragraph spacing so the panel reads naturally without oversized gaps.
• Bumps the unpacked extension manifest to 0.3.6 Coffee Support, so Chrome shows the refreshed local version.

Privacy and network behavior are otherwise unchanged: the extension remains local-first and BYOK; selected images are sent only to the user-configured endpoint. The extension does not process payments or verify support activity.

Verification for this release:

• npm run check:api-options
• npm run check:storage
• npm run check:json-repair
• npm run check:prompt-goal
• npm run check:url
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run release:package
• npm run release:check
• git diff --check

中文

这个版本在浮窗里加入一个低调的创作者支持入口。

• 在浮窗 header 里新增一个小的“请我喝杯咖啡”入口。
• 展示可选的支付宝赞赏二维码和创作者绿泡泡联系方式。
• 明确赞赏完全自愿，不影响插件功能，也不代表商业授权、付费客服或进群门槛。
• 收紧支持弹窗的布局、文案和段落间距，避免出现过大的空白。
• 将本地加载插件的 manifest 更新到 0.3.6 Coffee Support，Chrome 刷新后会显示新版本。

隐私和网络行为除此以外没有变化：插件仍是 local-first 和 BYOK；被选择的图片只发送到用户自己配置的端点。插件不处理付款，也不验证赞赏行为。

本版本验证：

• npm run check:api-options
• npm run check:storage
• npm run check:json-repair
• npm run check:prompt-goal
• npm run check:url
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run release:package
• npm run release:check
• git diff --check

v0.3.5

Asset SHA256: a4fedbb122500908339c401c413c2d1ded23f94cd6574c91568a18e444cdc8e8

English

This release improves long-running image analysis reliability for slower reasoning models.

• Raises the default API request timeout from 180 seconds to 600 seconds.
• Adds an API timeout setting in the options page, clamped between 60 and 1800 seconds.
• Adds a retry button on failed analysis results when the original source image or image URL is still available.
• Bumps the unpacked extension manifest to 0.3.5 Timeout Retry, so Chrome shows the correct refreshed local version.

Privacy and network behavior are unchanged: the extension remains local-first and BYOK; selected images are sent only to the user-configured endpoint.

Verification for this release:

• npm run check:api-options
• npm run check:storage
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run release:package
• npm run release:check
• git diff --check

中文

这个版本提升较慢深度思考模型识图时的稳定性。

• 将默认 API 请求超时时间从 180 秒提高到 600 秒。
• 在设置页新增 API 超时秒数，限制范围为 60 到 1800 秒。
• 识别失败且原始图片或图片 URL 仍可用时，失败面板新增“重新识别”按钮。
• 将本地加载插件的 manifest 更新到 0.3.5 Timeout Retry，Chrome 刷新后会显示正确版本。

隐私和网络行为没有变化：插件仍是 local-first 和 BYOK；被选择的图片只发送到用户自己配置的端点。

本版本验证：

• npm run check:api-options
• npm run check:storage
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run release:package
• npm run release:check
• git diff --check

v0.3.4

Asset SHA256: 064b6b3233f6df39cdcfc24902273028251aca2bf99aa4befaef72d8328f83a1

English

This release improves image-to-prompt-to-image reconstruction fidelity for generator handoff.

• Portrait, landscape, and square source images now preserve their source orientation, aspect ratio, and crop more consistently across English prompt copy, JSON prompt copy, and Open in generator.
• Contradictory frame wording from compatible vision models is corrected before the prompt reaches image generators.
• JSON Prompt copy now uses a structured generator-facing reconstruction prompt: adaptive modules, whole-image atmosphere, subject, composition, style, color, visible text, constraints, and auxiliary description. Internal schema fields stay out of the copied prompt.
• Source-style guidance now better separates visible subject matter from rendering style, so subject labels, costumes, eras, or genre guesses are less likely to override the actual medium, abstraction level, detail budget, and ornament density of the source image.
• Negative prompts keep style and detail blockers conditional, reducing over-heavy universal blocker lists while still protecting low-detail, soft, flat, or layout-critical sources from common drift.
• Popup and options HTML were simplified to avoid Chrome extension load warnings around module preloading.
• The injected page script now handles extension reloads more quietly and is checked against real page injection, reducing Chrome extension error-list noise.

Privacy and network behavior are unchanged: the extension remains local-first and BYOK; selected images are sent only to the user-configured endpoint. This release adds no telemetry, analytics, payment lock, or forced cloud dependency.

Verification for this release:

• npm run check:storage
• npm run check:json-repair
• npm run check:prompt-goal
• npm run check:api-options
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run smoke:history
• npm run test:real-extension
• npm run release:package
• npm run release:check
• User-installed refreshed-extension acceptance: passed

中文

这个版本优化“图片 -> 提示词 -> 图片”的复刻链路，重点是把真正交给图像生成器的提示词变得更稳定。

• 竖版、横版和方图会更稳定地保留原图方向、比例和裁切，并贯穿英文复制、JSON Prompt 复制和“打开生成器”链路。
• 兼容视觉模型偶尔返回的矛盾画幅描述，会在交给生成器前被修正。
• JSON Prompt 复制现在是面向生成器的结构化复原提示词：包含动态模块、整图氛围、主体、构图、风格、色彩、可见文字、约束和辅助描述。内部 schema 字段不会进入复制内容。
• 源图风格约束会更清楚地区分“可见主体”和“渲染风格”，避免主体标签、服饰、时代或类型推断压过原图真实的媒介、抽象程度、细节密度和装饰密度。
• 反向词里的风格/细节 blocker 保持条件式，避免所有提示词都变成过重的通用禁止词，同时继续保护低细节、柔和、扁平或版式关键的图片不发生常见漂移。
• 简化 popup 和 options HTML，避免 Chrome 插件页围绕 module preload 出现加载警告。
• 网页注入脚本会更安静地处理扩展重新加载，并增加真实页面注入检查，减少 Chrome 插件错误列表噪音。

隐私和网络行为没有变化：插件仍是 local-first 和 BYOK；被选择的图片只发送到用户自己配置的端点。本版本没有新增遥测、分析、付费锁或强制云依赖。

本版本验证：

• npm run check:storage
• npm run check:json-repair
• npm run check:prompt-goal
• npm run check:api-options
• npm run check:content-script
• npm run check:extension-pages
• npm run typecheck
• npm run build
• npm run smoke:history
• npm run test:real-extension
• npm run release:package
• npm run release:check
• 用户安装版刷新验收：已通过

v0.3.3

English

What changed

• Makes both the English prompt and JSON prompt safer to paste into Image2 and other image generators.
• Changes the JSON result and history JSON copy so the first field is a generator-facing prompt, not internal metadata.
• Cleans stray structural labels and reference-upload wording from generator-facing prompt text while preserving true visible text from the source image.
• Keeps the internal structured analysis available for saved records and advanced workflows.
• Improves recovery for malformed JSON-like prompt fragments so comma-separated visual details are preserved.

Upgrade guidance

Install or reload zhijuan-prompt-card-0.3.3.zip from this release. If the extension was already installed, refresh it in chrome://extensions before testing the JSON prompt output.

Verification

• Local regression, typecheck, build, release package, and release scan passed.
• Browser smoke and Bilibili E2E passed, including JSON copy checks for prompt first and no internal schema metadata.
• User-installed refreshed-extension acceptance was confirmed before merge.

Thanks to issue reporters and Image2/model-compatibility testers.

中文

更新内容

• 优化英文 Prompt 和 JSON Prompt，复制到 Image2 等外部图像生成器时更干净、更适合直接生成。
• JSON 结果和历史 JSON 复制现在会以面向生成器的 prompt 字段开头，而不是内部元数据。
• 清理误进入生成提示词的结构化标签和参考图上传话术，同时保留原图中真实可见的文字。
• 内部结构化分析仍保留给保存记录和高级工作流使用。
• 提升对异常 JSON-like 提示词片段的恢复能力，减少逗号分隔的视觉细节丢失。

升级方式

从本 release 安装或重新加载 zhijuan-prompt-card-0.3.3.zip。如果已经安装过扩展，请先在 chrome://extensions 中刷新扩展，再测试 JSON Prompt 输出。

验证

• 本地回归、typecheck、build、release package 和 release scan 已通过。
• 浏览器 smoke 与 Bilibili E2E 已通过，包含 JSON 复制检查：prompt 字段在前，且不包含内部 schema 元数据。
• 合并前已完成用户侧刷新安装版本验收。

感谢问题反馈者，以及 Image2 / 模型兼容性测试反馈。

v0.3.2

English

• Makes the English result the primary copy-ready recreation prompt, with length scaling to the image instead of a fixed short cap.
• Improves structured JSON output with a copy-ready generation prompt, matching negative prompt, and spatial dynamics summary so copied JSON keeps source-defining layout, motion, material, text, and drift-control details.
• Improves parsing resilience for compatible model responses that are close to JSON but not strictly formatted, reducing avoidable recognition failures.
• Preserves more source image detail before analysis, especially for dense UI, poster, and text-heavy images.
• Removes redundant hidden language output and duplicate prompt surfaces from new model responses while keeping Chinese/English prompts, structured JSON, negative prompt, and style tags.

中文

• 英文结果作为主要可复制复刻提示词，长度按图片复原需要扩展，不再受固定短词数限制。
• 改进结构化 JSON 输出，复制 JSON 到生成器时也更容易保留原图的版式、动势、材质、文字和漂移控制信息。
• 提升兼容模型返回近似 JSON 时的解析韧性，减少不必要的识别失败。
• 分析前更好保留源图细节，尤其面向密集 UI、海报和文字较多的图片。
• 新模型输出移除冗余隐藏语言输出和重复提示词入口，同时保留中英文 Prompt、结构化 JSON、反向词和风格标签。

Verification

• User refreshed-extension acceptance: passed before release.
• npm run check:api-options: passed.
• npm run check:json-repair: passed.
• npm run check:prompt-goal: passed, 58 contract rules, 12 simulated fidelity cases, 1 JSON generator-readiness case.
• npm run typecheck: passed.
• npm run check:storage: passed.
• npm run check:url: passed.
• npm run build: passed.
• npm run smoke:history: passed.
• npm run e2e:bilibili: passed.
• npm run release:package: passed.
• npm run release:check: passed.
• Package sha256: 6e20c9fa64b17b1c0f5c1efbe4cb0ec2e909e33b376f8238df8532f9836c36f8.

Upgrade

Download zhijuan-prompt-card-0.3.2.zip, unzip it, then refresh or reload the unpacked extension in chrome://extensions.

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

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

v0.3.0 Prompt Fidelity / 提示词复刻优化

English

• Improved image-to-prompt reconstruction fidelity for stronger image-to-prompt-to-image consistency.
• Better preserves recognizable subjects, composition, visible text, layout, visual style, material feel, and scene structure when supported by the image.
• Improves poster, UI, screenshot, and design-reference handling.
• Produces clearer reusable prompts with better drift control for follow-up image generation.

中文

• 优化图片反推提示词效果，提高“图片 -> 提示词 -> 图片”的复刻一致性。
• 更好保留图片中的可识别主体、构图、可见文字、版式、视觉风格、材质质感和场景结构。
• 改进海报、UI、截图和设计参考图的处理。
• 输出更清晰、更可复用的提示词，减少后续图像生成漂移。

Verification

• Prompt goal check, typecheck, storage check, URL check, build, visual history smoke, Bilibili E2E, release package, release scan, and diff whitespace check passed.
• Asset: zhijuan-prompt-card-0.3.0.zip

Zhijuan Prompt Card v0.2.0

Zhijuan Prompt Card v0.2.0

Public release package for Chrome/Chromium Load unpacked installation.

• Local-first image-to-prompt Chrome extension.
• BYOK vision endpoint configuration.
• No login, no credits, no bundled cloud, no telemetry, no analytics, no paywall.
• Includes LICENSE, NOTICE, PRIVACY.md, and THIRD_PARTY_NOTICES.md.
• Packaged asset: zhijuan-prompt-card-0.2.0.zip
• Clean-history release commit: dea7aa5
• SHA256: a6e3ba2f572c18842e944442783515942b891430f7c918f679284e960af0c9e0

Verification run:

• npm ci
• npm run typecheck
• npm run check:storage
• npm run check:url
• npm run build
• npm run release:package
• npm run release:check
• npm run smoke:history