LLM Editor

A CLI tool to edit files using LLMs.

Installation

pip install llm-editor

or, if you are installing from git directly

git@github.com:gptabhinav/llm-editor.git
pip install .

Configuration

You can configure the tool using a YAML file.

YAML Configuration

Run the following command to generate a default configuration file at ~/.llm-editor/config.yaml:

edit --init-config

Then edit the file to add your API key:

llm:
  provider: openai
  api_key: "your-api-key"
  model: "gpt-4o"

app:
  backup_enabled: true
  backup_suffix: ".backup"

Usage

1. Using Prompt Tags (Scripting Mode)

Add your instructions directly to the file using <tag> markers.

Example input.txt:

<tag> start_prompt
Convert the following Python code to JavaScript.
<tag> end_prompt

def greet(name):
    print(f"Hello, {name}!")

greet("World")

Run the tool:

edit input.txt

2. Interactive Mode

If the input file does not contain prompt tags, the tool will open your default text editor (configured via $EDITOR). You can type your instructions there, save, and close the editor to proceed.

3. Chat Mode

You can start an interactive chat session about your files using the --chat flag. This is useful for asking questions, getting explanations, or discussing potential changes without modifying the files directly.

Single File:

edit file.py --chat

Multiple Files: You can pass multiple files to load them all into the chat context.

edit file1.py file2.py --chat

Directories & Patterns: You can also pass directories or glob patterns to load multiple files at once.

# Load all files in a directory (recursive)
edit src/ --chat

# Load all Python files using glob pattern
# (Use quotes to ensure the tool handles the pattern matching)
edit "src/**/*.py" --chat

Options

• --outfile <path>: Write output to a specific file (input file is preserved).
• --inplace: Overwrite the input file directly (skips backup).
• --init-config: Initialize the configuration file.
• --chat: Start an interactive chat session about the file(s).

Usage Tracking & Costs

The tool automatically tracks token usage and estimates costs for each operation.

Logs are stored in ~/.llm-editor/logs/usage.log and contain:

• Timestamp
• Model used
• Token counts (Input/Output/Total)
• Estimated cost (USD)

Example log entry:

--- Usage Report (2023-10-27T10:00:00.123456) ---
Model: gpt-4o
Input Tokens:  150
Output Tokens: 50
Total Tokens:  200
Estimated Cost: $0.001250
------------------------------