This is a simple web app to call LLM APIs.

It aims to create a clean and fast UI baseline, while remaining a single-HTML file with vanilla JS.

Simply paste the code into LLMs to safety-check/customize it.

Main Goals:

• Concise, single-file code
• Easy to customize with LLMs or check for safety
• Design:
  • Consistent layout and color (tailwind)
  • Readable, high-contrast text

• Streaming output
• Multi-line input bar

• Markdown rendering (markdown-it)
• Code highlighting (highlight.js)
• Math rendering (katex)

• Store chat history locally (IndexedDB)
• Search bar
• Full text search
• Optimized DB operations
• Resize / Hide sidebar
• Responsive UI / Phone support

• CDN library caching
• Load large libraries in the background
• Auto-scrolling (disabled by default)
• Auto-focus
• Scroll chaining
• Memorize scrolling position
• Command+Click for a new window

• Copy button
• Correct LaTeX Copying
• Markdown tables
• Support the checklist syntax
• Skip CJK "Enter" button in input box
• Fix CJK characters bolding in markdown

• Delete / Pin / Rename chat
• Stop generation
• Edit message
• Retry
• Branching
• Allow parallel generation

• Change API Keys
• Manage models (api keys, base url, system prompts, temperature, top_p, max_tokens)
• Manage appearance (font, color, padding, darkmode preset)
• Slash commands
• Shortcut keys
• Export data
• Delete all images

• List and jump to messages
• Metadata (pricing, token count, raw json)
• Image uploads
• Show thinking trace
• Code Execution (WebAssembly)
• SVG / Mermaid rendering
• Print page