onModelInvocation

The onModelInvocation event gives you raw access to the streaming chunks returned by the LLM as they arrive over the network.

Usage

Because Fragola manages the stream processing internally, you rarely need to modify the stream here. However, this event is essential for:

• Piping the raw stream to a UI or WebSocket.
• Monitoring generation speed or logging raw deltas.

Payload Details

agent.onModelInvocation(async ({ stream, context }) => {
    // You can iterate over the stream if you want to inspect chunks
    // Note: Fragola uses `tee()` under the hood, so consuming this stream does not break the internal parser.
    for await (const chunk of stream) {
        process.stdout.write(chunk.choices[0]?.delta?.content || "");
    }
});