Processing

Agent can respond with multiple messages, each separated by newlines. This allows for a structured approach a more natural conversation flow.

Overview

For processing to be effective we need to combine the following skills:

Parsing: Interpret user intent and translate it into a defined skill.
Definition: Define skills, parameters and handlers.

Example

User request

I want to check if vitalik.eth is available

Parsing

The agent will parse the response into an array of skills and messages. This happens in the processMultilineResponse function.

gpt.ts

["Let me check that domain for you", "/check vitalik.eth"];

Execution

The agent will execute each skill in the array.

index.ts

  for (const message of messages) {
    // Check if the message is a command (starts with "/")
    if (message.startsWith("/")) {
      // Execute the skill associated with the command
      const response = await context.executeSkill(message);
      if (response && typeof response.message === "string") {
        // Parse the response message
        let msg = parseMarkdown(response.message);
        // Add the parsed message to chat memory as a system message
        chatMemory.addEntry(memoryKey, {
          role: "system",
          content: msg,
        });
        // Send the response message
        await context.send(response.message);
      }
    } else {
      // If the message is not a command, send it as is
      await context.send(message);
    }
  }