Text to Shell $t2sh

Turn plain language into powerful commands.

Powered by Ollama. Safe, fast, and intelligent command generation for Linux, macOS, and Windows with built-in safety checks and confirmation prompts.

🔒 100% Local

🛡️ Safety First

🌐 Cross Platform

🌍 Multilingual

user@ollash:~

Connected

llama3:latest

🚀 Advanced Features

Natural command execution with semantic search and safety-first automation

🔍 Interro-Powered Semantic Search

Find semantic occurences by searching your codebase using natural language.Powered by interro, it understands meaning, not just filenames or keywords.

📦 Local GGUF Model Support

Load and run open-source LLMs locally via GGUF format. Full support for Hugging Face-hosted models—no cloud dependency, full privacy.

🕓 History-Aware Shell Commands

Generate shell commands based on your previous command history and intent. Ollash learns from what you've done to help you do it faster.

🛡️ Reliable Multi-Layer Safety

Built-in command validation, risk assessment, and secure confirmation prompts protect you from harmful or destructive operations before they run.

🤖 Top Models From 30+ Supported Models

llama3

Meta's flagship model with superior reasoning

8B params RECOMMENDED

codellama

Code generation and programming assistant

7B params RECOMMENDED

Qwen3

Advanced multilingual language model

7B params RECOMMENDED

Qwen2.5-Coder-3B-Instruct-NL2SH-gguf

Qwen2.5-Coder (3B) model fine-tuned for natural language to shell command translation
by researchers at MIT CSAIL — read paper

3B params RESEARCH

llama-3.2-3b-instruct-nl2sh-gguf

LLaMA 3.2 (3B) model fine-tuned for natural language to shell command translation
by researchers at MIT CSAIL — read paper

3B params RESEARCH

Qwen2.5-Coder-7B-Instruct-NL2SH-gguf

Qwen2.5-Coder (7B) model fine-tuned for natural language to shell command translation
by researchers at MIT CSAIL — read paper

7B params RESEARCH

For adding models reach out to Team Chocos