Ask Ollama

Ask Ollama is a local chat app for Ollama. It has a React + Material UI interface, a Vite web build, and an Electron desktop shell.

Features

• Chat with any installed Ollama model.
• Stream answers as Ollama writes them.
• Render answers as Markdown, including code blocks, tables, links, and lists.
• Keep multiple saved chats in the left sidebar.
• Resize the left sidebar.
• Group chats by Pinned, Today, Yesterday, and Older.
• Create, switch, and delete chats.
• Undo deleted chats from toast.
• Save chats in browser local storage.
• Remember the selected model per chat.
• Show model badges for text, vision, code, and large-context guesses.
• Refresh installed models from Ollama.
• Use a system prompt to guide future answers.
• Save the system prompt per chat.
• Open a command palette for common actions.
• Search inside the current chat.
• Jump to next and previous search match.
• Highlight matching search text.
• Export a chat as Markdown.
• Export image previews in Markdown.
• Clear the current chat.
• Use starter chips from the empty chat view.
• Edit a user message into a new branch chat.
• Regenerate the latest answer.
• Copy assistant answers.
• Copy individual code blocks.
• Read assistant answers out loud with browser text to speech.
• Dictate prompts with browser speech to text when supported.
• Choose voice, speech speed, and pitch.
• Auto-read new answers.
• Choose mic language.
• Keep mic listening when browser supports it.
• Cancel an active request.
• Stop generation from the active answer.
• See answer time and word count.
• Auto-scroll while new answer text appears.
• Stop auto-scroll when you scroll up.
• Resume auto-scroll when you return near the bottom.
• Attach text files.
• Attach PDFs and extract their text.
• Attach images and send them to vision models.
• Choose full file or first-part file mode for long files.
• Paste images from the clipboard.
• Drag and drop files anywhere in the app.
• Show image thumbnails in the attachment list.
• Show attached image previews in chat messages.
• Click image attachments in messages to preview them.
• Show attachment type and size.
• Warn when a file is large and may exceed model memory.
• Warn when a PDF has no extractable text.
• Warn when an image is attached but the selected model does not look like a vision model.
• Show a clearer error when Ollama is not running.
• Rename chats.
• Pin chats to the top.
• Confirm chat deletion.
• Use prompt templates.
• Use keyboard shortcuts: Cmd/Ctrl+K search, Cmd/Ctrl+N new chat, Esc cancel.
• Open settings for theme, font size, default model, backup, and import.
• See model details from Ollama metadata.
• See a context-size estimate before sending.
• Switch between dark and light theme.
• Back up and restore all chats as JSON.
• Use src/images/ollama.png as the app logo, favicon, window icon, and packaged app icon.

Requirements

• Node.js and npm.
• Ollama running on the machine.
• At least one Ollama model installed.
• A vision model, such as llava, if you want image understanding.

Install

npm install

Run Web App

npm run dev

Vite proxies Ollama calls from /api/ollama to http://localhost:11434.

Build Web App

npm run build

Run Desktop App

npm run desktop

The desktop app talks directly to http://localhost:11434.

Package Apps

Build Windows portable app:

npm run dist:win

Build Mac app:

npm run dist:mac

Outputs go to release/. Ollama must still be running on the target machine.

Backend Configuration

Default web API base:

/api/ollama

Override it with:

VITE_OLLAMA_API_BASE_URL=https://example.com/api/ollama

The app uses:

• GET /api/tags to load models.
• POST /api/generate to stream answers.

Project Structure

• src/components/ - React and Material UI components.
• src/lib/ - Ollama API, file helpers, PDF text, model capability helpers.
• src/images/ollama.png - logo and icon image.
• electron/ - desktop app shell and preload bridge.
• vite.config.js - Vite config and Ollama proxy.
• package.json - scripts and Electron packaging config.

Notes

• Attached books and text files are sent as-is, not summarized first.
• Very large files may exceed the selected model context.
• Image attachments need a vision-capable Ollama model.