OpenAI Responses API Telemetry

Implementation evidence

OpenAI Responses API Telemetry: from boundary to verified row

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

1
Choose the outcome
Responses API cost and latency analysis
2
Define the contract
operation_id, response_id, feature, workflow, provider, model, and release
3
Instrument the boundary
Wrap the application-owned Responses API call and emit one terminal request event after usage and outcome are known.
4
Verify the evidence
Exercise a known fixture, then inspect openai_response_completed for one correctly typed terminal row.

Before you start

Prerequisites and boundaries

The official OpenAI JavaScript SDK and telemetry-sh initialized in trusted server code
An application-owned operation ID plus stable feature, workflow, model-alias, and release names
A reviewed policy for prompt, response, tool, and identifier data

Delivery setup

Install and initialize server-side

Import telemetry-sh in server-only code and initialize it once with process.env.TELEMETRY_API_KEY. Keep ingestion credentials out of browser bundles, client-visible environment variables, source control, logs, and exception messages.

openai-responses-api-install

npm installation

bash

npm 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.

openai-responses-api

OpenAI Responses API Telemetry event

javascript

import OpenAI from "openai";
import { Telemetry } from "telemetry-sh";

const openai = new OpenAI();
const telemetry = new Telemetry(process.env.TELEMETRY_API_KEY);

const startedAt = performance.now();
let response;
let status = "success";
let errorType;

try {
  response = await openai.responses.create({
    model: process.env.OPENAI_MODEL ?? "gpt-5.6",
    input: approvedInput,
  });
  return response.output_text;
} catch (error) {
  status = "failed";
  errorType = classifyOpenAIError(error);
  throw error;
} finally {
  await telemetry.log("openai_response_completed", {
    operation_id: operationId,
    response_id: response?.id,
    feature: "support_draft",
    provider: "openai",
    model: response?.model ?? process.env.OPENAI_MODEL ?? "gpt-5.6",
    status,
    error_type: errorType,
    duration_ms: Math.round(performance.now() - startedAt),
    input_tokens: response?.usage?.input_tokens ?? 0,
    output_tokens: response?.usage?.output_tokens ?? 0,
    total_tokens: response?.usage?.total_tokens ?? 0,
    release: process.env.APP_RELEASE,
  });
}

Event contract

operation_id, response_id, feature, workflow, provider, model, and release

status, duration_ms, input_tokens, output_tokens, total_tokens, retry_count, and error_type

tool_call_count, accepted, human_handoff, estimated_cost_usd, and pricing_version when approved

Implementation checkpoints

Checkpoint 1

Wrap the application-owned Responses API call and emit one terminal request event after usage and outcome are known.

Checkpoint 2

Provider usage supports token accounting, but cost is an application-owned estimate that needs a dated price source and separate reconciliation.

Checkpoint 3

Keep prompts, generated output, tool arguments, tool results, API keys, and unrestricted exception messages out of the event by default.

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.

openai-responses-api-verification

OpenAI Responses API Telemetry verification query

sql

SELECT *
FROM openai_response_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.

Track OpenAI costs LLM cost SQL AI output economics SQL OpenAI developer quickstart OpenAI Responses API reference

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