DeltaKit Documentation

fromAgnoAgents converts Agno agent session messages into DeltaKit's Message<ContentPart>[] format.

import { fromAgnoAgents } from "@deltakit/core";

Signature

function fromAgnoAgents(
  messages: Record<string, unknown>[]
): Message<ContentPart>[]

Parameter	Type	Description
`messages`	`Record<string, unknown>[]`	Raw messages from Agno's `aget_session_messages()`

Returns an array of Message<ContentPart> in chronological order.

Usage

// Load history from your Agno API
const history = await fetch("/api/chat-agno/history").then((r) => r.json());

// Convert to DeltaKit messages
const messages = fromAgnoAgents(history);
// => Message<ContentPart>[]

For using converted messages with the React hook, see Initial Messages.

Supported Message Types

The converter handles all Agno message roles:

User Messages

Role	Mapping
`user`	`Message` with `TextPart` containing the message content

Assistant Messages

Field	Mapping
`reasoning_content`	`ReasoningPart` with the thinking text
`tool_calls`	`ToolCallPart[]` for each tool invocation
`content`	`TextPart` with the response text

Tool Messages

Role	Mapping
`tool`	Tool result attached to matching `ToolCallPart` by tool name

Example Message Flow

// Raw Agno messages from aget_session_messages()
const agnoMessages = [
  { role: "user", content: "Calculate 2+2" },
  { 
    role: "assistant", 
    reasoning_content: "The user wants me to calculate...",
    tool_calls: [{ 
      id: "calc-1",
      function: { name: "calculate", arguments: '{"expr":"2+2"}' }
    }]
  },
  { role: "tool", tool_name: "calculate", content: "4" },
  { role: "assistant", content: "2 + 2 = 4" }
];

// Converted to DeltaKit format
const messages = fromAgnoAgents(agnoMessages);
// => [
//   { id: "agno_0", role: "user", parts: [{ type: "text", text: "Calculate 2+2" }] },
//   { 
//     id: "agno_1", 
//     role: "assistant", 
//     parts: [
//       { type: "reasoning", text: "The user wants me to calculate..." },
//       { type: "tool_call", tool_name: "calculate", argument: '{"expr":"2+2"}', callId: "calc-1" }
//     ]
//   },
//   // Tool result gets attached to the previous tool_call
//   { 
//     id: "agno_3", 
//     role: "assistant", 
//     parts: [{ type: "text", text: "2 + 2 = 4" }] 
//   }
// ]

Server Example (FastAPI + Agno)

from fastapi import APIRouter
from agno.agent import Agent
from agno.models.openrouter import OpenRouter

router = APIRouter()

@router.get("/api/chat-agno/history")
async def get_history():
    agent = Agent(
        model=OpenRouter(id="moonshotai/kimi-k2.5"),
        session_id="my-session"
    )
    # Get raw Agno messages
    messages = await agent.aget_session_messages(session_id="my-session")
    # Return as JSON - will be deserialized into Record<string, unknown>[]
    return messages

Notes

Messages are returned in chronological order.
Consecutive assistant messages (reasoning, tool calls, text) are grouped into a single Message.
Tool results (role: "tool") are matched to their corresponding ToolCallPart by tool_name.
A new user message always starts a new message group.
Messages without recognized roles are silently skipped.

fromAgnoAgents

On this page