Question about Token Usage Monitoring in Neuron AI Framework

[ololoDevs]

I'm currently using the Neuron AI framework to build a chatbot with Claude API, and I've noticed unexpectedly high token consumption in my requests (see attached screenshot - some requests consume 20k+ input tokens).

Looking at the Anthropic Console logs, I can only see basic metrics (timestamp, model, token counts), but I can't see the actual content being sent to the API - what prompts, system messages, or context are being included in each request.

My questions:

Does Neuron AI have built-in logging or debugging capabilities to see the full prompt structure being sent to Claude API?
Is there a way to inspect what exactly is included in each API call (system prompts, conversation history, context, etc.)?
Are there any configuration options to control or optimize how much context/history is being sent with each request?
Does the framework have any token usage monitoring or debugging tools I might have missed?

I'd like to understand and optimize the token consumption, but I need visibility into what's actually being sent to the API.
Thank you for your help!

[ilvalerione]

You can use the NeuronAI\Observability\LogObserver. You can create an instance of this class passing a logger compatible with PSR LoggerInterface.

$agent = MyAgent::make()->observer(
    new LogObserver($logger)
);