Pydantic AI Agent Telemetry

Implementation evidence

Pydantic AI Agent Telemetry: from boundary to verified row

Use Pydantic AI 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
Pydantic AI reliability
2
Define the contract
run_id, workflow, agent_name, provider, model_alias, and prompt_version
3
Instrument the boundary
Wrap Agent.run at the application boundary and emit one terminal outcome; use Pydantic AI instrumentation for detailed model and tool traces instead of duplicating them.
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

pydantic-ai and telemetry-sh initialized in a trusted Python process
A stable workflow name and application-owned run identifier
An allowlist that excludes prompts, outputs, dependency objects, tool arguments, and model messages

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.

pydanticai-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.

pydanticai

Pydantic AI Agent Telemetry event

python

from time import perf_counter

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

    try:
        result = await support_agent.run(prompt)
        return result.output
    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": "support_agent",
            "status": status,
            "error_type": error_type,
            "duration_ms": round((perf_counter() - started_at) * 1000),
            "release": APP_RELEASE,
        })

Event contract

run_id, workflow, agent_name, provider, model_alias, and prompt_version

status, duration_ms, retry_count, tool_call_count, and usage totals when approved

output_validated, accepted, human_handoff, error_type, release, and environment

Implementation checkpoints

Checkpoint 1

Wrap Agent.run at the application boundary and emit one terminal outcome; use Pydantic AI instrumentation for detailed model and tool traces instead of duplicating them.

Checkpoint 2

Treat output validation retries and provider retries as distinct categories when they lead to different engineering decisions.

Checkpoint 3

Do not serialize RunResult, messages, dependencies, tool arguments, or validated output into a Telemetry 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.

pydanticai-verification

Pydantic AI 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.

Instrumentation testing in CI Sensitive-data controls Agent loop detection SQL Pydantic AI agents Pydantic AI testing

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