Vercel AI SDK Telemetry

Implementation evidence

Vercel AI SDK Telemetry: from boundary to verified row

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

1
Choose the outcome
Streaming LLM analytics
2
Define the contract
feature, provider, model, prompt_version, and finish_reason
3
Instrument the boundary
Record the terminal generation result after the stream finishes so token usage and finish reason are complete.
4
Verify the evidence
Exercise a known fixture, then inspect llm_request_completed for one correctly typed terminal row.

Before you start

Prerequisites and boundaries

AI SDK generation or streaming calls running on a trusted server boundary
A stable feature and prompt-version taxonomy for comparing generations
An allowlist for telemetry metadata that excludes prompts, responses, headers, and tool payloads

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.

vercel-ai-sdk-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.

vercel-ai-sdk

Vercel AI SDK Telemetry event

javascript

await telemetry.log("llm_request_completed", {
  provider: "openai",
  model: "gpt-4.1-mini",
  feature: "draft_reply",
  prompt_version: "draft-reply-v4",
  input_tokens: 924,
  output_tokens: 218,
  latency_ms: 1380,
  finish_reason: "stop",
  status: "success",
  accepted: true,
});

Event contract

feature, provider, model, prompt_version, and finish_reason

input_tokens, output_tokens, total_tokens, estimated_cost_usd, and latency_ms

status, retry_count, time_to_first_token_ms, accepted, and release

Implementation checkpoints

Checkpoint 1

Record the terminal generation result after the stream finishes so token usage and finish reason are complete.

Checkpoint 2

Use AI SDK lifecycle callbacks or a telemetry integration for consistent coverage instead of duplicating ad hoc logging around every call.

Checkpoint 3

Review the SDK's experimental telemetry behavior during upgrades and explicitly exclude prompt, response, request-header, and tool-argument attributes.

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.

vercel-ai-sdk-verification

Vercel AI SDK Telemetry verification query

sql

SELECT *
FROM llm_request_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.

Tracking LLM costs LLM cost SQL Sensitive-data controls AI SDK telemetry AI SDK streamText 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