Skip to content

Message types

Chang Jia Geng edited this page Feb 20, 2026 · 1 revision

Current as of 20 Feb 2026

Legend

  1. 🤖 Greeting message
  • This is a greeting from the AI that you receive when you first log in. Prior to this, the AI had been given a system prompt.
  1. User message
  • User queries are flushed against the right-hand side and are orange colored. Here you can prompt the agent to do things, re-run, or confirm instructions. Abuse it like you would your bioinformaticians.
  1. 🤖 Tool call
  • The agent sends an instruction to a tool. The contents of the instructions (if any) are inside the code block. Tool calls have the
  1. 🛠️ Tool result
  • The tool responds with the result. The type of result depends on the tool called. It can be a code block, a document, or a graph with links to download.
  1. 🤖 AI answer
  • This is the final message from the agent. When this message appears, the user is able to continue prompting.
  1. System message
  • This is injected into the conversation when the user operates the command buttons e.g., Usage cost, Erase memory, Repair AI. System messages have a transparent background color and are not part of the conversation history. The AI is not aware of these messages.
  1. ⁉️ Exceptions
  • This occurs when the agent responds with a message that is unexpected. Typically the error is "context length exceeded", or "unparsed tool message". Refer to troubleshooting.

Messages types 1-4

message-types-1

Messages types 3-5

message-types-2

Messages types 6 & 7

Message types 6 & 7

Clone this wiki locally