Chit is a browser-only chat interface for interacting with large language models through any OpenAI-compatible API. It requires no server — all data is stored in the browser via IndexedDB and regular application storage.

This personal project is a constant WIP, and I'm not (yet) accepting PRs for new features or fixes.

• 🤖 OpenAI-Compatible API — Connect to any OpenAI-compatible backend (Ollama, LM Studio, vLLM, etc.) with configurable base URL and API key
• 💬 Multiple Chat Sessions — Create, switch between, and manage multiple persistent chat conversations
• 💾 Persistent Storage — Chat history, settings, and media are saved to IndexedDB and survive browser reloads
• 🎨 Customizable System Prompts — Set per-chat system prompts with user and system variable templating
• 📚 Preset System Prompts — Built-in prompt templates for common use cases (analysis, code review, ELI5, etc.)
• 🎛️ Advanced Model Parameters — Fine-tune temperature, top_p, presence penalty, repeat penalty, top_k, and seed per chat
• 💭 Thinking Mode — Models that support reasoning output display collapsible thinking blocks separate from the final response
• 🖼️ Image Upload and Paste — Attach images via file picker or clipboard paste; images are resized and stored in IndexedDB
• 🎵 Audio Upload — Attach .wav and .mp3 audio files for multimodal inference, displayed as playable cards in the timeline
• 📄 Text File Attachments — Attach code and text files (.py, .js, .json, .txt, .md, .csv, .xml, .yaml, etc.) that are inline with messages
• 🔧 Tool System — Built-in tools (calculator, time, ask-questions, echo, say) with support for MCP (Model Context Protocol) server integration
• 🔊 Text-to-Speech — Read responses aloud with configurable TTS engines, auto-speak options, and quote detection for selective reading
• 🎧 Audio Feedback — Distinct sound cues for typing, response completion, and questions
• 📦 Backpack References — Attach external context references (URLs, tool outputs) to enrich the conversation context window
• ⏪ Chat Rewind — Rewind to any previous message to branch the conversation from that point
• 📋 Chat Duplicate and Chop — Duplicate messages to create branches, or chop the latest message to re-roll responses
• 📐 KaTeX Math Rendering — Inline and block math formulas rendered with KaTeX
• ✏️ Markdown Editor — Rich markdown preview with SVG copy, bbox annotation, and code block copy buttons
• 🎭 App Modes — Switch between standard chat mode and roleplay mode for different interaction styles

• Any OpenAI-compatible API server (Ollama, LM Studio, vLLM, etc.)
• A compatible AI model loaded in your backend

1. Clone the repository:

   git clone https://github.com/network47/chit.git
   cd chit

2. Install dependencies:

   npm install

3. Start the development server:

   npm run dev

4. Open your browser and navigate to http://localhost:5173

To build the application for production:

npm run build

This creates a production build in the docs folder.

To preview the production build locally:

npm run preview

The project is configured to deploy to GitHub Pages. On every push to master, a GitHub Actions workflow builds the project and deploys the docs folder to GitHub Pages.

You can also trigger a manual deployment from the Actions tab.

Chit is built with:

• Svelte 5 — runtime and compiler
• Vite — build tool and dev server
• SCSS — styling with nested syntax and variables
• Markdown It — markdown rendering
• KaTeX — math formula rendering
• MCP SDK — Model Context Protocol integration
• Zod — runtime type validation and schema definition

The project structure follows Svelte conventions:

• /src — Source code
  • /app — Main application components
    • /Chat — Chat timeline, input bar, and message rendering
    • /ChatSidebar — Sidebar with session list, config panels, and mode controls
    • /UI — Shared UI primitives (modals, dialogs, buttons, pills, context menus)
    • /components — Reusable standalone components (async media, markdown editor)
  • /lib — Core logic and utilities
    • /chatSession — Chat state management, storage, attachments, and streaming
    • /llm — LLM driver interface, OpenAI-compatible driver, and response detectors
    • /tools — Built-in tool definitions and index
    • /mcp — MCP (Model Context Protocol) manager and server connections
    • /voice — TTS engine interface, OpenAI and Web Speech implementations
    • /backpack — External context reference management
    • /inputCommands — Slash command parser and handlers
    • /modes — App mode logic (chat, roleplay)
    • /text — Text normalization and token estimation
    • /templating — System and user variable templating
  • /preset-prompts — Predefined system prompt presets
  • /vendor — Vendored third-party scripts (EXIF.js, FileSaver, highlight.js)
• /public — Static assets (favicon, manifest, syntax highlighting CSS)

2026, Network47.org