Using CLAi
Successfully leveraging CLAi begins with proper setup and understanding its core functionalities. Ensure you have entered a valid API key under the settings tab and configured an SSH server connection.
Interaction Panel
The Interaction Panel serves as the primary interface for communicating with CLAi. Located at the bottom of this panel, a text box awaits your prompts or goals, translating your requests into actionable tasks for CLAi.
System Output Panel
Adjacent to the Interaction Panel, the System Output Panel mimics a terminal's functionality, showcasing the commands executed by CLAi and their respective outputs. For commands that CLAi cannot execute directly due to security concerns, an input box at the bottom of this panel allows for manual command entry.
Starting with CLAi
-
Connect: Begin by selecting your SSH server from the dropdown and clicking the connect button. Upon successful connection, the System Output Panel will display the server's initial response, such as a welcome message from the operating system.
-
Interact: CLAi introduces itself in the Interaction Panel, prompting you to input a command or goal. This could range from simple queries like listing directory contents to more complex tasks such as configuring databases or web servers.
Prompting CLAi
CLAi is adept at executing both straightforward commands and complex tasks involving multiple steps. It will intelligently navigate through tasks, seeking further input if the path to completion becomes unclear.
Sending Secure Commands
For sensitive operations, prefix commands with secure:
in the input box at the bottom of the System Output Panel. This ensures that commands, particularly those involving sensitive data, remain confidential and are excluded from both the system output and conversation history.
Usage Tips
-
Be Direct: Initiate interactions with CLAi using clear and executable commands. This approach minimizes confusion and enhances task execution efficiency.
-
Model Selection: While various models are available, the
GPT-4-1106-preview
model is recommended for its effectiveness. Should CLAi's performance not meet expectations, consider experimenting with different models.
Conversation Length Management
Utilize the Max Conversation Tokens setting to cap the token count of conversations, aiding in managing both the flow of interaction and associated costs. Refer to the Configuring CLAi section for detailed guidance.
Clearing Conversations
Post-goal achievement, use the Clear Conversation button to initiate a fresh dialogue. This practice not only helps in managing conversation lengths and costs but also ensures CLAi remains focused on new tasks.
Limitations
Despite its advanced capabilities, CLAi encounters limitations with:
-
Command Execution: Security precautions and model constraints may restrict some command executions. Utilize the secure command feature as a workaround.
-
Interactive Tools: Tools requiring interactive input, such as
vim
ornano
, present usability challenges due to their interactive nature. -
File Editing: CLAi's effectiveness in editing files, particularly large ones, is sometimes limited by model constraints and the system's token limit, necessitating careful consideration of the Max Response Tokens setting.