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

# Pipeline Traces

> Granular latency and logic execution tracking.

<Frame>
  <img src="https://mintcdn.com/arivu/piekGGEpY74vFGfF/images/arivu-traces.png?fit=max&auto=format&n=piekGGEpY74vFGfF&q=85&s=a259d53fba397ca328e546392b347e08" alt="Pipeline Traces" className="rounded-xl border border-gray-200 dark:border-gray-800" width="1919" height="914" data-path="images/arivu-traces.png" />
</Frame>

## Overview

Arivu operates via a complex LangGraph state machine. The Traces dashboard provides a visual map of exactly how much time each node took and the transformations applied to the data payload locally.

<CardGroup cols={2}>
  <Card title="Node-Level Telemetry" icon="microchip">
    Identify latency bottlenecks down to the exact pipeline boundary.
  </Card>

  <Card title="State Visualization" icon="chart-simple">
    Monitor how the state dictionary morphs through generation and verification.
  </Card>
</CardGroup>

## What You Can Inspect

<AccordionGroup>
  <Accordion title="Execution Loops" icon="rotate" defaultOpen>
    Visualize self-correction loops. When the `query_verifier` rejects SQL, the trace shows the round-trip back to `sql_generator` with the error context injected.
  </Accordion>

  <Accordion title="Token Usage" icon="coins">
    LLM token usage tracking per node. See exactly how many input/output tokens were consumed and calculate cost-per-query.
  </Accordion>

  <Accordion title="Latency Breakdown" icon="timer">
    A per-node millisecond breakdown: intake, generation, verification, execution, and response. Quickly spot if the LLM or the database is the bottleneck.
  </Accordion>
</AccordionGroup>

<Note>
  Pipeline trace data persists inside your local `arivu.db` (SQLite) or a Redis cluster if configured. Your telemetry never leaves your infrastructure.
</Note>