LLM Mode
Use natural language to interact with your terminal.
Overview
LLM Mode lets you describe what you want to do in plain English. The AI interprets your request, runs the appropriate commands, and reports back.
Setup
Add your OpenAI API key to .env:
OPENAI_API_KEY=sk-your-api-key-here
OPENAI_MODEL=gpt-4.1-nano
Usage
Enable LLM Mode
In Telegram, send:
/llm
Send Natural Language Commands
show me the git status
list all JavaScript files in src
what changed in the last commit?
find files containing "TODO"
create a new branch called feature-auth
run the tests and summarize results
Disable LLM Mode
/llm
Clear Conversation History
/llm_clear
Examples
Git Operations
You: "show me recent commits"
AI: Running: git log --oneline -10
abc1234 Fix authentication bug
def5678 Add user profile page
...
You: "what changed in the last commit?"
AI: Running: git show --stat HEAD
2 files changed, 45 insertions(+), 12 deletions(-)
File Operations
You: "find all config files"
AI: Running: find . -name "*.config.*" -o -name "config.*"
./config/default.json
./jest.config.js
...
You: "show me what's in package.json"
AI: Running: cat package.json
[displays file contents]
Code Analysis
You: "find all TODO comments in the codebase"
AI: Running: grep -r "TODO" --include="*.js" --include="*.ts"
src/utils.js:12: // TODO: Add error handling
src/api.js:45: // TODO: Implement caching
How It Works
- Your message is sent to OpenAI’s API
- The AI decides which command(s) to run
- Commands are executed in your workspace
- Output is analyzed and summarized
- AI can run follow-up commands if needed
Supported Models
Configure the model in .env:
OPENAI_MODEL=gpt-4.1-nano # Fast, cheap (default)
OPENAI_MODEL=gpt-4o-mini # Balanced
OPENAI_MODEL=gpt-4o # Most capable
Limitations
- Commands timeout after 2 minutes
- Output is truncated if too long
- Maximum 10 command iterations per request
- Requires active internet connection