ChinWag

A minimal React UI for Ollama - chat with local LLMs using a clean, modern interface. Built with Vite, Tailwind CSS, and shadcn/ui components.

Features:

• Chat interface with streaming responses
• Multiple chat sessions with search and filtering
• Model selection and parameter tuning
• Automatic conversation naming and tagging
• System prompt customization
• Markdown support with code highlighting
• Bookmarks and favorites support

Getting Started

Prerequisites

• Install Ollama
• Install Node.js (latest LTS version)

  Setup

  Clone the repository

  git clone https://github.com/ChillScientist/ChinWag.git
  

  Navigate to project directory

  cd chinwag
  

  Install dependencies

  npm install
  

  Start the development server

  npm run dev
  

  Configuration

  Ensure Ollama is running locally on port 11434 (default port) Pull at least one model in Ollama before starting:

  ollama pull llama2
  

  Features

  Basic Features

• From a listbox, the user can select an installed Ollama model
• In a text box, the user can type their prompt and press Enter key or press Send button
• The text box supports multiple lines of text. Pressing Shift+Enter adds a new line
• The text is sent to Ollama using the model selected and the user message entered
• The assistant message response is displayed in a text box
• The new user prompt text box is selected so the user can starting typing after the assistant message is generated

  System Prompt

• The user can enter a system prompt into a text box

  Streaming

• Support Ollama stream: true
• Text is displayed as it streams from Ollama

  Delete Messages

• The user can delete any user or assistant message by pressing a Delete icon/button next to the message
• All other messages are retained

  Edit Messages

• The user can edit any user or assistant message

  Regenerate Messages

• The user can press a Regen(erate) button to regenerate the most recent assistant message response
• The user can press Ctrl+Space to regenerate the most recent assistant message response

  Scroll to Bottom

• The window scrolls to show new assistant text as it streams in

  Chat Session History

• The user can create multiple chat sessions
• The user can press a New Chat button create a new session
• The user can rename the chat session. Defaults to "New Chat". Editing the name is "inline" (not modal dialog box).
• The user can delete a chat session
• The user can switch between chat sessions

  Chat Session Persistence

• Chat session history is retained across instances of running the application.

  Resizable Session SideBar

• The user can collapse/expand the Session Bar
• The user can resize the expanded Session Bar size

  Settings SideBar

• The user can select the model, set the prompt, and other supported settings from a sidebar
• The settings are per-session
• The settings are persisted with the session data

  Stop Assistant Message generation

• The user can stop an assistant message that's being generated before it finishes

  Tags and Notes

• The user can add tags to a session
• The user can add notes to a session
• The user can have the LLM generate tags or notes based on the conversation

  Markdown support

• Support Markdown in the Assistant message (most modern LLM's default to outputting formatted text in Markdown)

  Search

• User can filter the list of sessions in the Sessions Sidebar based on search criteria
• Case insensitive, partial word matching
• Search is real-time
• Search is weighted (name, tags, notes, message content)
• Supports operators (system:, name:, tag:, nope:, in:)

  Bookmarks and Favorites

• The user can bookmark a session for quick access to in-progress or reference conversations
• The user can favorite a session to mark it as particularly valuable or frequently accessed
• The user can filter sessions by bookmarked or favorited status
• The user can use search operators (type:bookmarked, type:favorite) to find marked sessions