https://github.com/gautierpicon/raccoonai
Use Raccoon.ai with Ollama to benefit from a user interface that lets you exploit your artificial intelligence locally.
https://github.com/gautierpicon/raccoonai
ai artificial-intelligence ollama ollama-client
Last synced: 7 months ago
JSON representation
Use Raccoon.ai with Ollama to benefit from a user interface that lets you exploit your artificial intelligence locally.
- Host: GitHub
- URL: https://github.com/gautierpicon/raccoonai
- Owner: Gautierpicon
- Created: 2025-03-16T13:40:58.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-03-23T23:03:54.000Z (7 months ago)
- Last Synced: 2025-03-24T00:18:51.789Z (7 months ago)
- Topics: ai, artificial-intelligence, ollama, ollama-client
- Language: JavaScript
- Homepage:
- Size: 553 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Raccoon.ai [](https://www.gnu.org/licenses/gpl-3.0)
## Description
Raccoon.ai is a web user interface for Ollama that allows you to leverage locally-run artificial intelligence models. This application offers an intuitive user experience for interacting with Large Language Models (LLMs) running on your own machine via Ollama.
### Why 'Raccoon'?
Because raccoons are cuts aren't they?
## features (non-exhaustive list)
### All Ollama features
- 🚀 Local execution
- 🔒 All data remains local, preventing sensitive information from leaking out
- 🤖 Supports various models such as LLaMA, Mistral, Gemma etc.
- 💻 Multi-platform compatibility: Works on Windows, Mac and Linux
- 🎛️ Resource optimisation: Efficient use of CPU and GPU for improved performance
### Currently present
- 🔄 Streaming support for real-time responses
- 📋 Persistent conversation history
- 🧠 Model selection from those available in your Ollama installation
### Future Plans
#### v1.0.0-beta
- 📝 Support for Markdown rendering from different AI models
- 🌗 Light, Dark, and System color modes
#### v1.0.0
- 💾 Saving past conversations and navigating through them via a scrollable menu
- 📋 Possibility of copying and pasting ia answers and having them regenerated if the answer was not conclusive.
#### v2.0.0
- 📥 Choosing how to use Raccoon.ai
- 📍 Solution 1 (current)
Download the project locally using GitHub (see "installation" section)
- 🛜 Solution 2 (future) (simpler solution but needs internet)
Using the hosted website (no data leaves your computer)
- 🔄 Ability to import and export conversations to, for example, store them on a USB drive and use them on different devices. Ideal for data security.
- 📂 Import files in conversations like images or pdf (only for multimodal models like [Gemma 3](https://ollama.com/library/gemma3))
#### v3.0.0
- 📱 Optimization of the interface for mobile use
## Prerequisites
- [Node.js](https://nodejs.org/) (v18 or higher)
- [Ollama](https://ollama.ai/) installed
## Installation
1. Start the Ollama server
```bash
ollama serve
```
2. Clone the repository
```bash
git clone https://github.com/Gautierpicon/RaccoonAI.git
cd RaccoonAI
```
3. Install dependencies
```bash
npm install
```
4. Start the development server
```bash
npm run dev
```
5. Open your browser at [http://localhost:3000](http://localhost:3000)
## API Architecture
### Communication with Ollama
Raccoon.ai communicates with Ollama via two main API endpoints:
1. **`/api/ollama/models`** - Retrieves the list of available models
- Method: `GET`
- Ollama endpoint used: `http://localhost:11434/api/tags`
- Response: List of models available in your Ollama installation
2. **`/api/ollama`** - Sends requests to models and receives responses
- Method: `POST`
- Ollama endpoint used: `http://localhost:11434/api/chat`
- Request body:
```json
{
"prompt": "Your message",
"model": "model-name",
"conversation": [
{ "role": "user", "content": "Previous message" },
{ "role": "assistant", "content": "Previous response" }
]
}
```
- Streaming: Uses Server-Sent Events (SSE) to stream the response in real-time
### Streaming Implementation
The project uses Next.js and browser streaming APIs to provide a real-time experience:
1. The API creates a `TransformStream` on the server side
2. The Ollama response is read progressively in chunks
3. Each fragment is encoded and sent to the client via SSE (Server-Sent Events)
4. The client decodes the events and updates the user interface in real-time
```javascript
// Excerpt from src/app/api/ollama/route.js
const stream = new TransformStream();
const writer = stream.writable.getWriter();
// Background processing
fetchOllamaResponseStreaming(prompt, model, conversation, writer, encoder);
// Returns a response stream
return new Response(stream.readable, {
headers: { 'Content-Type': 'text/event-stream' }
});
```
## Technologies Used
- **Next.js 15** - React framework
- **React 19** - UI library
- **Tailwind CSS** - Utility-first CSS framework
- **Framer Motion** - Animations
- **Server-Sent Events** - For response streaming
## Troubleshooting
### Ollama Not Detected
Make sure Ollama is running on `http://localhost:11434`. You can verify by visiting this URL in your browser.
### Models Not Available
If no models appear in the selector, download one via [Ollama](https://ollama.com/search):
Example:
```bash
ollama run llama3.2:1b
```
### Streaming Issues
If responses don't appear in real-time:
1. Check the console logs for errors
2. Make sure your browser supports Fetch API and Streams
3. Verify that the firewall isn't blocking communication with Ollama
## Contribution
Contributions are welcome! Feel free to open an [issue](https://github.com/Gautierpicon/RaccoonAI/issues) or a [pull request](https://github.com/Gautierpicon/RaccoonAI/pulls) on the GitHub repository.