Minimalist local LLM chat interface using Ollama
npm install yak-llm
A minimalist local LLM terminal chat interface using Ollama, native streaming, and JSONL memory.
Yak at your local model. Fully offline. Edit or delete your chat history(s).
---
- 🧠 Runs on any Ollama-compatible model (gemma3, llama3, mistral, etc)
- 🧵 Offline, Real-time streaming replies in the terminal
- 📜 Memory stored in lightweight jsonl files on your machine
---
- Node.js 18+
- Ollama installed and running
---
bash
npm install -g yak-llm
`Start
bash
yak start
`Start chatting
🦧> What's a good name for a chat app?
🤖: How about "Yak"? It's short, and yaks are fluffy like llamas!
If you have Ollama downloaded and running, Yak will automatically download a model if none is found.
Yak will also create a new
chat session for your first session.Stop chatting with command or ctrl+c
`
🦧> /bye
Usage
🚀 Main Commands
──────────────────────────────────────────────────
yak start Start chat session with your model
yak help Show this help message
yak list List all chat sessions
yak models List available models
💬 Chat Management
──────────────────────────────────────────────────
yak new Create new chat session
yak switch Switch to chat session
yak delete Delete chat session
yak --reset Clear current chat history
🤖 Model Management
──────────────────────────────────────────────────
yak model Switch to different model
yak models List downloaded models
💬 In-Chat Commands
──────────────────────────────────────────────────
/bye or /quit Exit chat session
http or www Chat will detect given URLs and crawl them
Models
You can see your downloaded Ollama models by running:
bash
yak models
`You can change the model with the CLI command
bash
yak model
`
or in the file manually.
`json
{
"model": "gemma3:1b",
}
`If you want more models, download them from Ollama! 🦙
History
Don't like your chat? All messages are logged to
.You can remove lines from file, or delete the entire log:
`bash
yak delete
``