跳到正文
VibeFather
免费注册

用这个模板做的 2 个案例

同一份提示词,不同的 [VARIABLES] 变量。点一个在上方预览。

关于这个模板

AI 聊天应用界面是一个完整的助手工作区,面向构建 AI 产品的开发者和小型团队。它打开时就是一个实用、完整的对话,包含可读的代码示例、对比表格和可用的输入框。深梅子色侧边栏为界面提供了强有力的框架,而温暖的米白色表面和珊瑚色控件让主对话区保持亲切感。深色主题将同一套配色带入更安静的梅子黑工作区。

该模板包含让聊天应用感觉完整的小交互。对话可以搜索、置顶、重命名和删除。回复在本地流式输出,可以停止,也可以重新生成。编辑较早的消息会用新的回答替换后续轮次。Markdown 表格和高亮代码块在窄屏幕上仍然可读,复制控件提供清晰的反馈。文件和图片标签会打开真实的本地预览,而不是装饰性占位符。

模型选择器会改变本地回复风格,助手资料保留可编辑的系统提示词,用于通用帮助、写作和开发。新对话提供有用的起点。键盘快捷键、移动端抽屉、可见的焦点状态和持久化草稿帮助工作区在各类设备上保持舒适。对话历史、附件、偏好设置和资料都存储在浏览器中。

使用默认的虚构产品工作室,或替换应用名称、工作区、成员、助手、受众和示例项目。示例对话和附件文本会跟随这些值,同时布局保持一致。一个隔离的响应适配器为以后连接你自己的模型服务提供了明确的位置。交付的演示完全在本地运行,没有凭据、外部模型请求或远程上传。每个视觉元素都用代码绘制,因此界面不需要图片服务或图库照片。

模板提示词

在线演示就是用这份提示词原样做出来的,没有额外补充。其他案例只是改了 [VARIABLES] 变量的值。 提示词保持英文原版,国内常用的 AI 编程工具都能直接理解;下面附中文参考译文。

# AI Chat App Interface

Build this website exactly as described below. Replace the [VARIABLES] with your own values; keep everything else.

## Brief
Build [BRAND], a complete browser-local AI chat workspace for [WORKSPACE]. Use the exact variable defaults on the initial demo. Lead with a seeded conversation containing a highlighted code answer and a Markdown table, a deep-plum history sidebar, and a vivid coral composer action. Implement the full interaction loop, light/dark themes, editable assistant profiles and persistent local attachments.

- Audience: [AUDIENCE] who want a focused place to work through [PROJECT].
- Goal: Let [USER_NAME] organize conversations, ask [ASSISTANT_NAME] for useful local example replies, revise a request and keep the workspace between visits. A single sendMessage adapter is the sole future integration point; the delivered site makes no external requests.
- Tone: Warm, concise, thoughtful and practical. Specific product-work examples, restrained humor, no grand AI claims, no marketing hero.

## Stack
- Framework: Vite + React + TypeScript
- Styling: Tailwind CSS 4 + CSS variables
- Libraries: react (MIT), react-dom (MIT), vite (MIT), @vitejs/plugin-react (MIT), typescript (Apache-2.0), tailwindcss (MIT), @tailwindcss/vite (MIT), react-markdown (MIT), remark-gfm (MIT), highlight.js (BSD-3-Clause), lucide-react (ISC), @fontsource/space-grotesk (OFL-1.1), @fontsource/dm-sans (OFL-1.1), @fontsource/ibm-plex-mono (OFL-1.1)
- Build: `npm run build` → static files in `dist/`. No backend, no server code, no external APIs.
- Setup:
  - Use Vite with base: './', build.sourcemap:false and the React and Tailwind Vite plugins; build static files directly into dist/.
  - Keep all copy, variable defaults, seeded conversation content, profile defaults and local response banks in src/content.ts; components reference this data instead of duplicating copy.
  - Use a typed React reducer or equivalent small state layer for conversations, drafts, profiles, attachments and active-stream identity. No state library is needed.
  - Implement the only response boundary in src/adapters/sendMessage.ts as an abortable async generator. Every send, retry, regenerate and edit-and-resend calls it; never call a real model API.
  - Use react-markdown with remark-gfm and a custom code renderer using highlight.js on code text only. Disable raw HTML and unsafe URL protocols; never highlight or execute arbitrary HTML.
  - Import only needed local font weights: Space Grotesk 500/600/700, DM Sans 400/500/600/700, IBM Plex Mono 400/500. Use CSS variables for both exact palettes and custom keyframes for tiny UI transitions.
  - Use index.html as the only page and client state for the workspace; no route library is required. Set the document title to [BRAND] and create an original local SVG favicon.
  - Create ASSETS.md documenting the locally bundled icon/font packages and original code drawings, with source URLs, authors and licenses; include no remote visual assets.
  - When variables are unchanged, use every default exactly, including the seeded project copy. If a value is changed later, rewrite all dependent visible prose, system prompts, snippets, filenames and attachment text coherently while keeping structure, palette and functionality.

## Variables
- [BRAND] — Assistant app name (default: "Plumline"). A fictional product name; derive the favicon label and storage namespace from this value.
- [WORKSPACE] — Workspace name (default: "Juniper Studio"). The fictional team or personal workspace shown in the sidebar.
- [USER_NAME] — Your display name (default: "Nora Vale"). A fictional workspace member; show their first name in the empty-state greeting.
- [ASSISTANT_NAME] — Assistant name (default: "Mira"). The fictional assistant shown beside replies; keep profile role labels separate.
- [AUDIENCE] — Primary audience (default: "small product teams"). Adapt sample conversations, starter prompts and profile descriptions to this audience.
- [PROJECT] — Example project (default: "a collaborative notes app"). The default chat works through a launch checklist for this project; adapt code data, table rows and files together if changed.
- [CREDIT] — keep a small "Template by VibeFather" link to https://vibefather.ai/t/ai-chat-app-interface in the footer (default: yes; delete this line to remove it)

The copy under Pages & Sections is written for the default values. If your values differ, rewrite every headline, paragraph, label and image for them — same length, tone and structure. Layout, colors, type and motion stay as described.
Keep all copy, image paths and links in one content file (for example `src/content.ts`), keyed by section, so the content can be swapped later without touching components.

Pro 模板

解锁完整提示词和源码

  • 完整提示词:设计、版块、动效、素材
  • 每个在线演示的源码
  • 商用授权

在线演示和 DESIGN.md 仍然免费。

$9/月起 · 年付和终身正在上线特惠

DESIGN.md

单独的设计规范:配色、字体、间距和动效。可以把这套视觉用到你已有的项目上。

---
name: "AI Chat App Interface"
colors:
  warm_canvas: "#FAF7F2"
  deep_plum: "#3B1E54"
  vivid_coral: "#FF6B4A"
  ink: "#2E2336"
  muted_plum: "#726578"
  soft_blush: "#F3E4DE"
  paper_surface: "#FFFFFF"
  warm_border: "#D9CADC"
  plum_code: "#24122F"
  dark_canvas: "#170F20"
  dark_sidebar: "#21152B"
  dark_surface: "#2C1B37"
  dark_border: "#3D294A"
  dark_secondary: "#C6B5CE"
typography:
  wordmark_conversation_and_dialog_headings: { family: "Space Grotesk", weights: [500, 600, 700] }
  interface_labels_messages_and_body_copy: { family: "DM Sans", weights: [400, 500, 600, 700] }
  code_keyboard_shortcuts_and_file_metadata: { family: "IBM Plex Mono", weights: [400, 500] }
spacing: "Use a 4px base scale. Desktop: 264px sidebar, 72px main header, 790px reading width, 40px outer main gutters, 20px message gaps, 10px paragraph gaps, 16px field padding. Mobile: 12–16px gutters, 44px minimum touch targets, 12px control gaps. Keep the initial code and table answer compact enough to appear above the composer at 1440×900."
radius: "8px small buttons; 10px history active row; 12px code blocks, suggestion cards and popovers; 14px user bubble; 16px composer and dialogs. Avoid excessive pill-shaped containers; reserve pills for small selectors/status."
---

# AI Chat App Interface — Design System

Apply this design system to an existing project. It describes the look and feel only; for the full site, use prompt.md.

## Colors
- **Warm canvas** `#FAF7F2` — Required light-mode page background, transcript canvas and warm-white text on plum.
- **Deep plum** `#3B1E54` — Required saturated sidebar, primary headings, button text on coral, geometric details; occupies a substantial part of the first screen.
- **Vivid coral** `#FF6B4A` — Required New chat and Send controls, original mark, focus highlights and sidebar active edges; keep saturated and pair with plum text, never low-contrast white small text.
- **Ink** `#2E2336` — Primary light-mode body text; high contrast on warm surfaces.
- **Muted plum** `#726578` — Light-mode secondary text, metadata, helper text and icons; keep readable at small sizes.
- **Soft blush** `#F3E4DE` — User message bubbles and attachment chips.
- **Paper surface** `#FFFFFF` — Light-mode composer and menu surfaces.
- **Warm border** `#D9CADC` — Light-mode separators and field boundaries.
- **Plum code** `#24122F` — Code blocks in both themes, with ivory, lilac, peach and coral syntax colors.
- **Dark canvas** `#170F20` — Dark-mode page and transcript background.
- **Dark sidebar** `#21152B` — Dark-mode sidebar.
- **Dark surface** `#2C1B37` — Dark-mode bubbles, composer and menus.
- **Dark border** `#3D294A` — Dark-mode separators and controls.
- **Dark secondary** `#C6B5CE` — Dark-mode secondary copy; primary text remains #FAF7F2.

## Typography
- **Wordmark, conversation and dialog headings** — Space Grotesk (OFL), weights 500/600/700. Self-host the font files.
- **Interface labels, messages and body copy** — DM Sans (OFL), weights 400/500/600/700. Self-host the font files.
- **Code, keyboard shortcuts and file metadata** — IBM Plex Mono (OFL), weights 400/500. Self-host the font files.

## Spacing & shape
- Spacing: Use a 4px base scale. Desktop: 264px sidebar, 72px main header, 790px reading width, 40px outer main gutters, 20px message gaps, 10px paragraph gaps, 16px field padding. Mobile: 12–16px gutters, 44px minimum touch targets, 12px control gaps. Keep the initial code and table answer compact enough to appear above the composer at 1440×900.
- Radius: 8px small buttons; 10px history active row; 12px code blocks, suggestion cards and popovers; 14px user bubble; 16px composer and dialogs. Avoid excessive pill-shaped containers; reserve pills for small selectors/status.

## Motion
- Use only restrained 120–180ms opacity/color/4–8px movement transitions for controls and overlays.
- Stream newly requested replies locally with deterministic 30–75ms word-group pacing after a 450ms thinking state; completed seeded replies are static on load.
- Autoscroll only near the transcript bottom and preserve the reading position otherwise.
- Respect prefers-reduced-motion: no transforms, blinking cursor, bounce or continuous decoration; response text updates in 150ms batches.

## Imagery
All visuals are original CSS/SVG and real interactive UI. The visual anchor is the polished seeded transcript with a deep-plum syntax-highlighted code block, readable comparison table, warm message bubble and coral controls. Create a simple original two-stroke mark and abstract geometric member avatar in code. No photos, generated raster art, real logos, image services or externally fetched thumbnails. User-selected images are local attachment previews only.

## Tone
Warm, concise, thoughtful and practical. Specific product-work examples, restrained humor, no grand AI claims, no marketing hero.

---
— Prompt by VibeFather · https://vibefather.ai/t/ai-chat-app-interface · CC BY-NC 4.0 (non-commercial; commercial use needs a VibeFather Pro license) · Request removal: https://vibefather.ai/takedown

改成你自己的

  • 独立开发者:将 BRAND 设置为你的助手名称,WORKSPACE 设置为你自己的工作室,并围绕你的产品发布清单重写预置对话。
  • 客户支持团队:保留历史侧边栏和可编辑资料,然后为退款和入门问题重写系统提示词和起始提示词。
  • 写作教练:将 ASSISTANT_NAME 和资料指向起草和编辑帮助,并将代码示例和表格替换为大纲和修改示例。
  • 学习社区:将 AUDIENCE 设置为学生,调整起始提示词以适应学习主题,并使用本地附件预览来共享笔记和图片。

AI 聊天应用界面:常见问题

AI 聊天应用界面模板开箱即用包含哪些功能?

AI 聊天应用界面模板打开时就是一个预置对话,包含高亮代码示例和 Markdown 表格。对话可以搜索、置顶、重命名和删除,回复在本地流式输出,可以停止或重新生成。它还增加了模型选择器、用于通用帮助、写作和开发的可编辑助手资料、键盘快捷键、移动端抽屉、持久化草稿以及本地文件和图片预览。

AI 聊天应用界面会连接 Claude 或 GPT 这样的真实 AI 模型吗?

不连接。AI 聊天应用界面模板完全在浏览器中运行,没有凭据、外部模型请求或远程上传。回复来自本地响应库,模型选择器只改变本地回复风格。提示词保留了一个隔离的响应适配器,作为以后连接你自己的模型服务的唯一位置,因此每次发送、重试和重新生成都保持在同一个边界内。

怎么用 AI 做一个AI 聊天应用界面?

复制本页的完整提示词(Pro),或者运行 npx vibefather add ai-chat-app-interface,把 [BRAND]、[PROJECT]、[AUDIENCE]、[USER_NAME] 换成你自己的内容,再粘贴到 Claude Code、Codex、Cursor 等 AI 编程工具里。在线演示就是用这份提示词原样做出来的。