feat: add tab-completion for magic commands using readline and added working groq model

[DevaRajan8]

Fixes #1697

Problem

Open Interpreter has 12+ magic commands (%help, %reset, %verbose,
etc.) but users must remember exact spellings with no discoverability.
Although readline was already imported for up-arrow history,
tab-completion was never wired up for magic commands.

Solution

Added a MagicCommandCompleter class and registered it with readline:

• Typing % + Tab shows all available magic commands
• Typing %re + Tab completes to %reset
• Non-% input is unaffected (no completion)
• Up-arrow history is unaffected
• Completer is re-registered before each input call to prevent
  other libraries (e.g. litellm) from resetting it

Changes

Only one file changed: interpreter/terminal_interface/terminal_interface.py

1. Added MagicCommandCompleter class at the top of the file
2. Registered the completer in the existing readline try block
3. Re-registered completer before each input() call to ensure
   it is never overridden by imported dependencies

Code Added

class MagicCommandCompleter:
    """Readline completer for Open Interpreter magic commands."""
    
    MAGIC_COMMANDS = [
        "%help", "%verbose", "%auto_run", "%reset",
        "%save_message", "%load_message", "%undo",
        "%tokens", "%info", "%jupyter", "%markdown", "%%",
    ]
    
    def __init__(self):
        self.matches = []
    
    def __call__(self, text, state):
        if state == 0:
            if text.startswith("%"):
                self.matches = [cmd for cmd in self.MAGIC_COMMANDS 
                                if cmd.startswith(text)]
            else:
                self.matches = []
        return self.matches[state] if state < len(self.matches) else None

Demo

Testing

Tested on Python 3.13.3, Ubuntu Linux, open-interpreter 0.4.3:

• % + Tab → shows all 12 magic commands
• %re + Tab → completes to %reset
• %sa + Tab → completes to %save_message
• %info entered directly → works as before
• Up-arrow history → unaffected
• Normal (non-%) input → no completion triggered
• Windows (no readline) → falls back gracefully

Groq Model Update:

• Changed to a Working Groq Model (LLama 3.1 to LLama 3.3 versatile 70b model)

now uses llama-3.3-70b-versatile
Model loads without decommission errors
Basic functionality verified

[endolith]

Probably should split this into two PRs since readline and groq are completely unrelated issues

[DevaRajan8]

Probably should split this into two PRs since readline and groq are completely unrelated issues

ok , will do that