Open source · OpenTelemetry · TypeScript

Keep your logs.
Spend on the signal.

Meet Jev: a tiny robot with headphones and a strong opinion about noise. Your LLM doesn’t need every health check. Let Jev find the logs worth thinking about.

Start building ↗Sort a pile with Jev ↓

v0.2.0 · MIT licensed · npm install jevlogs

Jev, a relaxed ivory robot in cobalt headphones, sorting log slips into a gray background box and a blue signal box. — Every log is kept. Background stays cheap. Signal gets the thinking.

Fits your existing stackOpenTelemetry+Jev by TypeSafe+Vercel AI SDKYour exporters. Your infrastructure.

00 / Jev’s desk

Sort a tiny pile.

Health checks belong in background. Failures belong in signal. Jev keeps every record — it just decides what deserves an expensive second look. Same four sample logs as npx jevlogs.

Close-up of Jev smiling while holding a blue signal card at a messy desk of log papers.

Headphones on. Show me the pile.

Incoming logs

INFO

GET /health → 200api-gateway · routine heartbeat

DEBUG

Cache hit: product:482catalog · expected operation

ERROR

Payment capture failedcheckout · protected by severity

WARN

Connection pool near capacitydatabase · investigate degradation

backgroundkeep · skip LLM0

signalkeep · send to analysis0

4 logs in the pile. Drag them, tap a bin, or let Jev sort.

01 / A smaller first decision

Think less about noise.
Learn more from failures.

Put a fast decision model before your reasoning model. Keep your observability pipeline, and choose what deserves a deeper look.

Give every log a value.

Jev grades diagnostic value on a 0–100 rubric, assigns urgency, and estimates whether further investigation is useful.

Keep the uncertain ones.

Only confidently low-value events can skip analysis. Errors and protected records always pass through. A timeout preserves the record.

Make fewer expensive calls.

Annotate first. Then enable routing on a separate LLM branch while your existing telemetry archive keeps everything.

Built for decisions, not paragraphs

Why Jev?
The math is different.

Jev holding up a blue signal card under the words Introducing jev logs.

$0.042per million input tokens
TypeSafe’s published launch price

Typed choices, rubric scores, and probabilities in one request. No generated explanation needed for a routing decision.

Parallel questionsShared log contextFree outputAt TypeSafe launch pricingExplicit uncertaintyKeep ambiguous events

Gateway lists $0.04/M; billing may differ. Jev is early access. Provider speed benchmarks are not Jev Logs benchmarks. Typed outputs can still make incorrect decisions.Read TypeSafe’s announcement ↗

02 / Less noise. A smaller bill.

What could you save?

A transparent estimate, not a benchmark. Adjust the fraction of logs still sent to your LLM, including errors, uncertain decisions, and failures.

Estimated monthly savings

$1,05187.6% less analysis spend

Analyze everything$1,200.00

Jev + selected LLM calls$149.40

Includes Jev input and question overhead, plus selected LLM input and output. Excludes storage, ingestion, hosting, retries, discounts, and prompt caching. Assumes equal average log sizes; savings depend on your workload.

03 / Small integration. Real control.

Your pipeline.
One new layer.

Install with npm install jevlogs. Try npx jevlogs for the offline demo, or add jevlogs.config.json and run npx jevlogs --live to start a local OpenTelemetry receiver. Send HTTP/JSON logs or wrap your existing exporter.

Read the quick start ↗

telemetry.ts

import { JevLogExporter } from 'jevlogs';
import { BatchLogRecordProcessor } from
  '@opentelemetry/sdk-logs';

const processor = new BatchLogRecordProcessor({
  exporter: new JevLogExporter({
    exporter: yourExporter
  })
});
// Every log kept. Value + priority attached.

Node.js 22+ · AI SDK 7.0.105 · OTel SDK Logs 0.222.0

Open source, from day one

Save the reasoning
for the interesting part.

Explore the source ↗

Independent project. Not affiliated with TypeSafe, Vercel, or OpenTelemetry.

Keep your logs.Spend on the signal.