Commands
AnyLLM has built-in "slash commands" that allow you to manage the session and interact with the application directly, bypassing the AI agent. Commands always start with a / character and must be entered at the very beginning of the line.
You can type / to see a dropdown list of all available commands.
/exit
Purpose: To properly exit the AnyLLM application.
This command ensures a clean shutdown of the application. Importantly, it saves the current session context (if it is enabled) before exiting. This means that the next time you run AnyLLM in this project, the agent will "remember" where you left off, including the conversation history, created files, and decisions made.
/clear
Purpose: To clear the terminal screen and, if necessary, reset the current session.
This command has two modes of operation:
- If session mode is disabled: The command simply clears the visible terminal area, giving you a clean slate.
- If session mode is enabled: The command not only clears the screen but also completely resets the current session context. This is equivalent to starting a brand new session in the project. All conversation history, created files, and decisions made within the session will be forgotten.
/summarize
Purpose: To create a semantic summary of the current conversation.
This is one of the most powerful commands for long sessions. It instructs the AI agent to "re-read" your entire conversation and extract the essence:
- Key Decisions: For example, "use SQLite for the database."
- Constraints: For example, "do not use external libraries."
- Important Questions and Answers.
Benefit: This concise summary is added to the agent's "memory." In subsequent interactions, it won't need to re-read dozens of old messages to recall an important constraint—it will always have it right in front of it. This significantly improves its accuracy in long and complex tasks and saves your tokens.
/init
Purpose: To analyze the entire project and generate a detailed report file named ANYLLM.md.
This command launches a specialized agent that acts as an expert code analyzer. The agent will recursively study your project's structure, analyze configuration files (package.json, composer.json, etc.), and determine the main technology stack, entry points, architecture, and launch instructions.
All the gathered information will be structured and saved into an ANYLLM.md file in the root of your project. This file can serve as an excellent starting point for new developers or for automating documentation generation.
Warning: This process can be time-consuming and token-intensive, as it requires an in-depth analysis of numerous files.