LlamaIndex Agent Telemetry

Implementation evidence

LlamaIndex Agent Telemetry: from boundary to verified row

Use LlamaIndex Agent Telemetry at a controlled application boundary, keep the event contract small, and verify a known outcome before building aggregate views.

1
Choose the outcome
LlamaIndex agent monitoring
2
Define the contract
run_id, workflow, agent_name, model_alias, index_version, prompt_version, and release
3
Instrument the boundary
Wrap the public agent.run or workflow boundary and emit one terminal application outcome after the handler completes.
4
Verify the evidence
Exercise a known fixture, then inspect agent_run_completed for one correctly typed terminal row.

Before you start

Prerequisites and boundaries

LlamaIndex and telemetry-sh initialized in a trusted Python process
An application-owned run ID plus stable workflow, agent, index, and release versions
A reviewed policy for prompts, outputs, retrieved nodes, tool arguments, and trace content

Delivery setup

Install and initialize server-side

Initialize Telemetry for synchronous code or TelemetryAsync for asyncio code once per process with a server-side key. Keep ingestion credentials out of browser bundles, client-visible environment variables, source control, logs, and exception messages.

llamaindex-install

pip installation

bash

python -m pip install telemetry-sh

1Prepare one reusable server-side delivery client with bounded network behavior.
2Add the outcome event at the success, failure, retry, or timeout boundary.
3Send controlled fixtures and inspect the stored rows before enabling an alert.

Snippet

Start with one structured event

Add this shape where the workflow completes, fails, or retries. Then build the dashboard from real fields.

llamaindex

LlamaIndex Agent Telemetry event

python

from time import perf_counter

async def run_support_agent(agent, user_message: str, run_id: str):
    started_at = perf_counter()
    status = "success"
    error_type = None

    try:
        return await agent.run(user_msg=user_message)
    except Exception as error:
        status = "failed"
        error_type = classify_agent_error(error)
        raise
    finally:
        await telemetry.log("agent_run_completed", {
            "run_id": run_id,
            "workflow": "support_resolution",
            "agent_name": "llamaindex_support_agent",
            "status": status,
            "error_type": error_type,
            "duration_ms": round((perf_counter() - started_at) * 1000),
            "index_version": INDEX_VERSION,
            "prompt_version": PROMPT_VERSION,
            "release": APP_RELEASE,
        })

Event contract

run_id, workflow, agent_name, model_alias, index_version, prompt_version, and release

status, duration_ms, tool_call_count, retry_count, human_handoff, and error_type

estimated_cost_usd, accepted, evaluation_score, evaluator_version, and environment when approved

Implementation checkpoints

Checkpoint 1

Wrap the public agent.run or workflow boundary and emit one terminal application outcome after the handler completes.

Checkpoint 2

Use LlamaIndex workflow events and instrumentation for step-level inspection; send only the aggregate fields needed for a reviewed SQL question to Telemetry.

Checkpoint 3

Do not serialize prompts, agent messages, retrieved nodes, embeddings, tool inputs, tool outputs, or exception messages into the event.

Verification

Prove the event arrived

Run this after exercising known success and failure cases. Replace the fallback table name if your final event contract differs from the snippet.

llamaindex-verification

LlamaIndex Agent Telemetry verification query

sql

SELECT *
FROM agent_run_completed
ORDER BY timestamp_utc DESC
LIMIT 20;

Confirm one terminal row per logical outcome, with the expected status, identifiers, units, and UTC time.

Inspect the inferred schema and verify that retries do not change field types or generate a new logical event ID.

Search the stored fields for credentials, raw payloads, prompts, private content, and unbounded error messages.

Exercise a provider timeout, ingestion rejection, and process shutdown before treating the dashboard as complete.

Implementation references

Review the event contract, data-safety guidance, and upstream primary documentation before enabling a new production path.

Evaluate AI agents with SQL RAG evaluation SQL Agent task success SQL LlamaIndex agent documentation LlamaIndex multi-agent workflow example

Production boundary

Keep the outcome event small and recoverable

This pattern provides

A bounded, SQL-ready outcome beside the upstream workflow.
Stable fields for dashboards, alerts, and cross-event correlation.
A fixture-driven path for validating success, failure, retry, and timeout behavior.

This pattern does not provide

An OTLP exporter, automatic collection pipeline, or replacement for detailed traces and diagnostic logs.
Exactly-once delivery merely because the payload contains an event ID.
Permission to collect raw provider payloads, user content, credentials, or regulated data.

Delivery and idempotency Batching and shutdown Troubleshoot ingestion

Event schema starting points

Review the row grain, emit boundary, required types, privacy classes, example payload, and validation checklist before adapting a query or snippet to production.

AI operations10 fields