Skip to main content
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. 
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.
  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.