An experimental client for the web from Unternet.
- Run
npm install - Copy
.env.exampleto.envand fill in the required environment variables - Start the native client with
npm run dev
Operator has support for running LLM inference locally using Ollama. By default, this will be used if no Vite OpenAI API key has been provided as an environment variable.
To set this up on Linux use this to download and install Ollama:
curl -fsSL https://ollama.com/install.sh | shOr download the binary from their webisite.
Once installed use this to download the qwen2.5-coder:3b model which is the default.
ollama run qwen2.5-coder:3bWe have tested it with the following and mistral seems to work best. Please let us know if you have tried others that work well / don't work so we can update this table.
| Model | Search | Error |
|---|---|---|
| deepseek-r1:latest | use @web to search for weather 02155 | Error handling command input: AI_APICallError: Invalid JSON response |
| mistral | use @Web to get me top 5 news stories about FB stock | ✅ |
| qwen2.5-coder:3b | use @Web search for top news in india | 🟠 executes the search but does not add a response |
Windows builds are currently unsigned. Signing can be enabled by adding windows-certs.pfx and setting the WINDOWS_CERTS_PASSWORD secret in GitHub Actions and updating the certificateFile and certificatePassword fields in package.json under the "win" section.
You can optionally add the Unternet API to enable features like web search. In order to do this, add your Unternet API key (don't have one? email us!). For now, this will only work when you build as the API is expected to be local.