Skip to content

AI agent to automatically check grammar and spelling on documentation files

License

Notifications You must be signed in to change notification settings

AgentOps-AI/Spellcaster

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

58 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Spellcaster Logo

AI spelling and grammar checker for documentation

🐦 Twitter   •   📢 Discord   •   🖇️ AgentOps

Python 3.10+ License: MIT

Spellcaster is an open-source CLI tool that uses AI to improve your codebase's documentation. It scans repositories for:

  • Grammar errors
  • Spelling mistakes
  • Issues in code examples

Spellcaster helps identify potential problems to help you write clear, concise, and professional documentation.

demo.mp4

To run Spellcaster:

pip install spellcaster
spellcaster -d /path/to/your/docs/directory/

Screenshot

🦉 Features

  • Grammar and Spelling Checks: Spellcaster reviews your documentation and comments to ensure they are clear, correct, and professional.
  • Comprehensive File Scanning: Spellcaster reads across multiple file formats (Markdown, plain text, etc.) to detect issues.

Installation

  1. Install package:

    pip install spellcaster
  2. Set up environment variables:

    Create a .env file in your project root and add the following variables:

    AGENTOPS_API_KEY=your_agentops_api_key
    ANTHROPIC_API_KEY=your_anthropic_api_key
    OPENAI_API_KEY=your_openai_api_key
    GROQ_API_KEY=your_groq_api_key
    

Spellcaster uses AgentOps for cost and latency tracking. You can sign up for an API key here.

Replace your_*_api_key with your actual API keys for each service.

Usage

To run Spellcaster:

  1. In your terminal, use the following command structure:

    spellcaster [options]

    Options:

    • -d, --directory: The directory to scan
    • -u, --url: The GitHub repository URL to clone and scan
    • -l, --llm_provider: The LLM provider to use (choices: claude, sonnet, 3.5, gpt4o, gpt4, gpt3.5; default: MODEL from config)
    • -p, --proper_nouns: A string of proper nouns to include in the prompt (default: "* Llama3.1-70B\n* Cerebras\n* Cohere\n* OpenAI\n* AgentOps\n* Spellcaster")
    • -f, --file_types: File types to scan (default: from FILE_TYPES in config)
    • -m, --max_files: Maximum number of files to scan (default: MAX_FILES from config)
  2. Examples:

    Scan a local directory with custom LLM provider:

    spellcaster --directory /path/to/your/docs --llm_provider gpt4

    Scan a GitHub repository:

    spellcaster --url https://github.com/username/repo
  3. Spellcaster will analyze the specified directory or repository and output any detected issues, along with suggestions for fixing them.

Configuration

You can customize Spellcaster's behavior by adjusting the config.py file:

  • FILE_TYPES: A list of file extensions to scan. By default, it includes .mdx and .md files. You can add or remove file extensions as needed.
  • MAX_FILES: The maximum number of files to scan. By default, it's set to 500. You can change this number to suit your needs.
  • MODEL: The default LLM provider to use. Model list available here.