Playground/Language & data/YouTube extract
API ACTIVITY

Usage & budget

PLAYGROUND SAFETY LIMITS

Steady requests. Visible limits.

0 active · 0/20 queued

Browser: one start every 1.2 seconds (at most 50/minute), up to 2 active calls and 20 waiting. Shared by examples in this tab. Waiting requests cancel when their run is stopped; a full queue rejects new work without an API call.

Server: 60 requests per rolling 60 seconds and 2 concurrent calls per key, per server instance. Other tabs using the same key share that instance’s allowance. This in-memory safeguard resets on restart and is not a distributed or account-wide quota.

Source: playground policy in lib/requestRateLimit.ts. TypeSafe may impose separate limits. Provider HTTP 429 pauses calls until its reported reset; without a reset time, retry requires your action. HTTP 402 is a budget/billing failure, not a pacing limit. No automatic retries or background replay.

Your TypeSafe API key

Import your key to override the shared server default.

Saved on this browser until removed or site data is cleared. Masked on screen; sent only with Jev requests through this app. The saved value is never displayed.

Connecting
Language & data · extractive prototype

Keep the words. Find the point.

Existing YouTube captions → Jev relevance decisions → a chronological extract. No generated prose, transcription, or translation.

WORKSPACE GUIDE

YouTube extract

Select original caption passages with Jev scores, cost tracking and timestamp-linked verification.

CaptionsJev scoresExtract

What you provide

A public YouTube URL whose video already has a caption track. English is preferred, manual tracks before automatic ones; the track's origin is shown.

What happens

Captions are split at gaps, terminal punctuation and speaker markers. Each chunk gets one request carrying a relevance score and a key-claim choice, with the bounded transcript as topic context. No transcription, translation or text generation runs.

What you get

Chunks scoring at least 0.5, minus near-duplicates, restored to chronological order and joined by whitespace. Every word is original caption text. Missing or failed scores stay unknown and stop the run rather than being filled in.

Walk through a run

  1. 01

    Load captions

    Paste a public YouTube URL with an existing caption track.

  2. 02

    Score passages

    Jev scores relevance and classifies key claims. Adjust extract length locally.

  3. 03

    Verify omissions

    Inspect all scored chunks and open timestamp links to compare with the source.

Try this

Extract a talk, then open the raw list and read the dropped passages: a high score cannot prove that nothing important was omitted.

Execution limits

Source-only extract, not a generated summary. Scores cannot prove that nothing important was omitted. Public caption retrieval may be blocked.

Explore the example. Check the evidence.

YouTube extract

Live Jev · two questions per call

Public captions only, never transcription or translation. Caption access can be blocked by YouTube. Running sends caption text to TypeSafe.

Limits before a run
Caption lines 5,000
Cue count in the track, before cleanup.
Characters 60,000
Whole transcript. Oversized tracks fail; nothing is truncated.
Natural chunks 200
Split at gaps, terminal punctuation and speaker markers.
Characters per chunk 6,000
An unpunctuated run stays whole, so it can exceed this alone.
Watch page 2 MiB
Upstream body bound. A heavily decorated page can exceed it.
Retrieval deadline 20 s
Covers the watch page, player and caption fetches together.

A track over any of these fails and names which one. Nothing is truncated to fit.

Recomputes locally without more calls. Relevance ≥ 50%; word-overlap duplicates above 60% are dropped. Key claims are shown for inspection, not used as a selection gate.

Ready

Chunks processed
0 / 0
Chunks kept
0
Average kept relevance
—
Jev calls attempted
0
Input cost estimate
$0.000000
Elapsed
0.0s

Reported input tokens only; excludes output and hosting costs. Not an invoice. 0 calls with unknown usage.

Extract preview

Your source-linked extract will appear here.

An extract can omit critical context. Jev scores and confidence are not verification. Compare kept and dropped passages with the video. Cleanup removes fillers conservatively; joining adds whitespace only, preserving existing source transitions.