Message Types

Every message in LeafMesh's conversation system has a type that identifies its origin and purpose.

Type Reference

Type	Role	Source
`user_message`	`user`	Direct user input via `mesh_call()`
`agent_response`	`assistant`	Agent's LLM-generated or intelligence-function response
`mesh_call`	`system`	Inter-agent mesh communication
`tool_result`	`system`	Result from a tool execution

Message Structure

{
    "role": "user",                        # OpenAI-compatible role
    "content": "What is 2 + 2?",          # Message content
    "timestamp": "2026-02-26T10:30:00Z",  # ISO 8601 timestamp
    "agent_name": "solver",               # Which agent handled this
    "type": "user_message"                # Message type
}

How Types Are Used

Message assembly formats user_message and agent_response entries as user/assistant pairs in the LLM message array
Session storage keeps all message types in the conversation history
Debugging uses the type and agent_name fields to reconstruct the full interaction chain

Next Steps

Message System — How messages are constructed
Session Management — Conversation history

LeafMesh — Typed, attributed messages