Streaming

Basic Streaming

for await (const event of agent.stream({ task: 'Write a summary' })) {
  switch (event.type) {
    case 'run_start':
      console.log(`Run started: ${event.runId}`);
      break;
    case 'llm_token':
      process.stdout.write(event.token);
      break;
    case 'tool_start':
      console.log(`\nCalling: ${event.toolName}`);
      break;
    case 'tool_end':
      console.log(`Result: ${event.result}`);
      break;
    case 'done':
      console.log(`\nCompleted: ${event.result.status}`);
      break;
    case 'error':
      console.error(`Error: ${event.error}`);
      break;
  }
}

async for event in agent.stream(task="Write a summary"):
    match event.type:
        case "run_start":
            print(f"Run started: {event.run_id}")
        case "llm_token":
            print(event.token, end="")
        case "tool_start":
            print(f"\nCalling: {event.tool_name}")
        case "tool_end":
            print(f"Result: {event.result}")
        case "done":
            print(f"\nCompleted: {event.result.status}")
        case "error":
            print(f"Error: {event.error}")

AgentStreamEvent

The stream yields a union of event types:

Event	Fields	When
`run_start`	`runId`	Run begins
`step_start`	`stepIndex`	New step begins
`llm_token`	`token`	LLM produces a token
`llm_stream`	`event: ChatStreamEvent`	Raw LLM stream event
`tool_start`	`toolName`, `input`	Tool execution begins
`tool_end`	`toolName`, `result`, `duration`	Tool execution completes
`step_end`	`stepIndex`, `step: StepRecord`	Step completes
`done`	`result: RunResult`	Run completes
`error`	`error: string`	Error occurred

LLM Streaming Requirement

Streaming requires the LLM to implement the StreamingLLM interface. All built-in providers (Anthropic, OpenAI, Gemini, Ollama) support streaming.

import { isStreamingLLM } from '@ahzan-agentforge/core';

const llm = createLLM({ provider: 'anthropic', model: 'claude-sonnet-4-20250514' });
console.log(isStreamingLLM(llm)); // true

If the LLM doesn't support streaming, agent.stream() will throw an error.

Collecting Stream Results

To both stream and collect the final result:

let finalResult;

for await (const event of agent.stream({ task: 'Do something' })) {
  if (event.type === 'llm_token') {
    process.stdout.write(event.token);
  }
  if (event.type === 'done') {
    finalResult = event.result;
  }
}

console.log(finalResult.status); // "completed"

Next Steps

Hooks — lifecycle callbacks
LLM Streaming — LLM-level streaming details

Basic Streaming

AgentStreamEvent

LLM Streaming Requirement

Collecting Stream Results

Next Steps

On this page