Field extraction, PII detection and exact quotes, built on TypeSafe’s Jev. Jev only answers multiple choice, so this library numbers the words of your text, lets Jev pick numbers, and returns the exact substring with its offsets and a probability.
Live demo of the jeveryword libraryNewHosted API, free to try, then pay per call with x402Hosted API: free to try, then x402↓Ready to run.
Each word of your text gets a number. Jev answers every question with numbers, and the page copies the matching words back out of your text.
Cost at $42 / billion input tokens.
Measured for this run.
In the second sample, the first-name field asks for the speaker’s first/given name, excluding other people’s names, so the result is Maya and Alex is skipped. Jev receives each field’s instruction together with your text.
To add a field of your own, such as Middle name, give it a name, a JSON key and an instruction. Every value is text copied from the input, or null. Run it with Extract fields above. Your fields stay in place across the samples until you reset them.
The actual model, state, instructions, and choices sent to TypeSafe. API credentials are excluded.
Jev scores every word of your text, and the words likely to be personal information are highlighted in place.
Each word is scored by Jev“PII only” asks one yes/no question per word and uses fewer tokens. The slider and highlights work the same way, without category labels.
Ready to scan.
The slider re-filters the same scan without another request. A score is 1 minus Jev’s probability that the word is not personal, and it does not guarantee a correct detection.
Select a highlight to see its category and score.
Experimental review aid. Lower thresholds reveal weaker candidates; missed PII can remain unhighlighted. Redacted copy removes only the currently highlighted spans. Review before sharing.
Every non-whitespace chunk is classified in context, including all people, repeated mentions, and corrected values. Adjacent chunks with the same category are joined visually. The displayed span score is the lowest score among its chunks. No regex-only candidate filter and no generative model.
This page runs on jeveryword, an MIT-licensed JavaScript library with no dependencies. It needs a TypeSafe API key, or no key if you deploy on Vercel.
Teach your coding agentnpx skills add jkrup/jeverywordnpm install jeverywordThe same field extraction and PII detection as an HTTP API, with no account or API key. Each request is paid in USDC over x402, and a request that fails is not charged. It suits agents that hold a wallet. The first 25 calls a day from an IP address are free, so you can try it with curl.
| Pull fields out of text | $0.005 per call + $0.0005 per field |
|---|---|
| ten fields from a message: 1¢ | |
| Find personal data | $0.004 per 1,000 characters |
| $0.006 with a category for each hit |
curl https://jeveryword.vercel.app/v1/extract \
-H 'Content-Type: application/json' \
-d '{
"text": "Hi, I am Maya Chen. Reach me at maya@fern.example",
"fields": [
{"id": "name", "description": "The speaker full name."},
{"id": "email", "description": "The speaker email address."}
]
}'{
"results": {
"name": { "status": "extracted", "value": "Maya Chen",
"start": 9, "end": 18, "probability": 1 },
"email": { "status": "extracted", "value": "maya@fern.example",
"start": 32, "end": 49, "probability": 1 }
},
"calls": 1, "inputTokens": 1095, "durationMs": 418
}For PII, post {"text": "…"} to /v1/pii. Text can be up to 4,000 characters, with up to 16 fields.
npm install @x402/fetch @x402/evm viemimport { wrapFetchWithPayment } from '@x402/fetch';
import { x402Client } from '@x402/core/client';
import { ExactEvmScheme } from '@x402/evm/exact/client';
import { privateKeyToAccount } from 'viem/accounts';
const account = privateKeyToAccount(process.env.EVM_PRIVATE_KEY);
const client = new x402Client();
client.register('eip155:*', new ExactEvmScheme(account));
// pay() behaves like fetch, but pays a 402 and retries
const pay = wrapFetchWithPayment(fetch, client);
const url = 'https://jeveryword.vercel.app/v1/extract';
const response = await pay(url, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
text,
fields: [
{ id: 'name', description: "The speaker's full name." },
],
}),
});
const { results } = await response.json();EVM_PRIVATE_KEY is the key of a wallet holding a little USDC on Base. Each call costs a cent or less. Keep the key in an environment variable on a server, and use a wallet that only holds a small balance. Coding agents can run npx skills add jkrup/jeveryword, which includes this.