PyGPTPrompt offers the following features for interacting with various components:
/robots <url>: Fetch the robots.txt file of a specified URL./browse <url>: Retrieve HTML content from a specified URL, convert it to Markdown, and cache it locally. If the content already exists in the cache, the cached version is returned.
/rss <url>: Fetch and display full-text articles from an RSS feed.
/ls <dir>: List files in a local directory without revealing sensitive details. Access to filesystem commands is restricted by configuration to enhance privacy.
/read <file> [start_line] [end_line]: Read the content of a local file. This command retrieves the content of the specified file, optionally specifying the range of lines to read.
/<command>: Execute a shell command (restricted by configuration).
Examples:
/cat <file>: Read the content of a local file./git status: Display the status of a git repository.
Please note that the available commands may be subject to configuration
restrictions. For detailed information on each command, you can refer to the
/help command.
PyGPTPrompt uses a config.json file to manage various settings, including the
OpenAI GPT model, security configurations, and more. Here's a breakdown of each
section:
Configures the chat model used by PyGPTPrompt.
model: The name of the chat model. "gpt-3.5-turbo" is currently recommended.max_tokens: The maximum number of tokens for the model to generate in a single response.temperature: Controls the randomness of the model's output. Higher values (closer to 1) make the output more random, while lower values make it more deterministic.base_limit_percentage: Represents the percentage of the context window size reserved for resulting output. Content exceeding this percentage is written to a file and its path is returned.system_message: The initial message that sets the behavior of the assistant.
Configures various paths used by PyGPTPrompt.
session: The path to the directory where session data is stored.environment: The path to the environment file.storage: The path to the directory where the application stores its data.
Configures access controls for shell commands and file paths.
allowed_commands: A list of allowed shell commands.disallowed_commands: A list of explicitly disallowed shell commands.disallowed_strings: A list of strings that, if present in a command, will cause it to be disallowed.disallowed_chars: A list of characters that, if present in a command, will cause it to be disallowed.
allowed_paths: A list of allowed file paths. Subdirectories and files within these paths are implicitly allowed unless explicitly disallowed.disallowed_paths: A list of explicitly disallowed file paths, taking precedence over allowed paths.
Configures the style of the output.
italic: The string to use for italic text.bold: The string to use for bold text.
Handle the configuration file with care to ensure security. Pay attention to the
access section, ensuring that disallowed_paths adequately protect sensitive
files and directories. Review the allowed_commands to exclude any potentially
risky commands. Customization options are available, but it's crucial to make
informed choices to maintain a secure environment.
Here's an example of a config.json file:
{
"chat_completions": {
"model": "gpt-3.5-turbo",
"max_tokens": 1024,
"temperature": 0.5,
"base_limit_percentage": 0.1,
"system_message": {
"role": "system",
"content": "I am ChatGPT, an AI programming assistant utilizing the `pygptprompt` interactive CLI based on `prompt-toolkit`..."
}
},
"path": {
"session": "/path/to/sessions",
"environment": "/path/to/env",
"storage": "/path/to/storage"
},
"access": {
"shell": {
"allowed_commands": ["ls", "cat", "cd", "tree", "git"],
"disallowed_commands": ["rm", "sudo", "su"],
"disallowed_strings": ["-rf /", ":(){ :|:& };:"],
"disallowed_chars": ["&", "|", ";", "*", ">"]
},
"file": {
"allowed_paths": ["/path/to/source/code", "/path/to/files"],
"disallowed_paths": [
"/path/to/env",
"/path/to/private/keys",
"/path/to/config.json"
]
}
},
"style": {
"italic": "italic",
"bold": "bold"
}
}