> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nuabase.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Observability & Storage

> Trace, debug, and audit every LLM interaction

LLM applications can be difficult to debug. Inputs are large, prompts are complex, and outputs are non-deterministic. Nuabase acts as a system of record for all your AI interactions.

## Request Storage

Nuabase stores the full context of every request:

1. **Input Data**: The raw JSON data you passed in.
2. **Prompt**: The exact prompt template used.
3. **Output**: The raw LLM response and the parsed JSON result.
4. **Metadata**: Token usage, latency, model version, and cost.

This means you don't need to log huge blobs of text in your own application database.

## Retrieving Request Details

Every Nuabase response includes an `llmRequestId`. You can use this ID to fetch the full details of that execution later.

```typescript theme={null}
const result = await myFn(data);

if (result.isSuccess) {
  console.log("Request ID:", result.llmRequestId);
  
  // Later, or in a different admin tool:
  const details = await nua.getRequest(result.llmRequestId);
  console.log("Original Input:", details.input);
  console.log("Prompt Used:", details.prompt);
}
```

## Usage in Development

During development, this is invaluable. You can:

1. Run a function from your local machine.
2. Go to the [Nuabase Console](https://console.nuabase.com).
3. See exactly what was sent to the LLM and why it responded the way it did.

## Compliance & Auditing

For enterprise use cases, having a complete audit trail of every AI decision is often a requirement. Nuabase provides this automatically without any extra engineering effort.