Sending Documents

Every endpoint that takes a résumé or job description accepts the same three input forms: text, url, or file. Send exactly one.

{ "text": "Jane Doe — Senior Backend Engineer\n..." }
{ "url":  "https://example.com/resumes/jane-doe.pdf" }
{ "file": "JVBERi0xLjQKJcOkw7zDtsO..." }

This applies to /parse/resume, /parse/jd, /extract/*, and /redact. /match takes two documents at once, so its fields are prefixed — see below.

No matter which form you use, the document is processed transiently and deleted the moment the call ends. Nothing résumé-shaped is ever stored.

text — raw text

The simplest form, and the fastest: nothing to fetch or decode.

curl -X POST "$BASE/parse/resume" \
  -H "x-access-key: $REZMATCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Jane Doe\njane@example.com | +1 (415) 555-0142 | San Francisco, CA\n\nEXPERIENCE\nSenior Backend Engineer — Stripe, San Francisco, CA\nMar 2021 – Present\n- Own the payments ledger service in Go.\n\nEDUCATION\nUC Berkeley — B.S. Computer Science, 2017\n\nSKILLS\nGo, Python, PostgreSQL, Kubernetes"
  }'

Remember to JSON-escape newlines as \n. If what you send is actually HTML, it’s stripped to visible text before processing — you don’t have to clean up a page you scraped yourself.

url — fetch it for me

Point at a PDF or a web page and Rezmatch.ai fetches it. This is the natural form for job postings: Greenhouse, Lever, Workday, and company career pages all work directly.

curl -X POST "$BASE/parse/jd" \
  -H "x-access-key: $REZMATCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://job-boards.greenhouse.io/acme/jobs/123456"}'

Fetch behavior:

• http(s) only, redirects followed, 25-second timeout.
• The fetcher identifies itself as RezMatchBot/1.0 — allowlist it if you’re pointing at a site you control that has bot mitigation.
• HTML is reduced to its visible text before anything else happens. A job posting is overwhelmingly markup, scripts, and styles; stripping it first means your size limit is measured against the words that matter, not the page weight.
• The URL must be reachable from the public internet. Documents behind a login, a signed S3 URL you haven’t shared, or a private network won’t work — fetch them yourself and send file instead.

A URL that times out, 404s, or refuses the connection comes back as invalid_request with the reason in the message, and costs no credits.

file — base64 bytes

Send the document itself. Use this for anything you already have in hand, or anything not publicly reachable — the common case for résumés.

curl -X POST "$BASE/parse/resume" \
  -H "x-access-key: $REZMATCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"file\": \"$(base64 -i resume.pdf)\"}"

import base64, requests
 
with open("resume.pdf", "rb") as f:
    encoded = base64.b64encode(f.read()).decode()
 
res = requests.post(
    f"{BASE}/parse/resume",
    headers={"x-access-key": API_KEY},
    json={"file": encoded},
)

import { readFile } from 'node:fs/promises';
 
const file = (await readFile('resume.pdf')).toString('base64');
 
const res = await fetch(`${BASE}/parse/resume`, {
  method: 'POST',
  headers: { 'x-access-key': API_KEY, 'Content-Type': 'application/json' },
  body: JSON.stringify({ file }),
});

Base64 only — plain JSON, no multipart/form-data. A data-URI prefix (data:application/pdf;base64,) is not accepted; send just the payload.

Supported: PDF (processed natively, layout and all) and plain text. Not yet supported: DOCX, RTF, and images, which return unsupported_format. An image-only scan with no text layer returns parse_failed.

Size limits

Limits depend on your plan and on which form you used, because each form measures size differently:

A hard 10 MB ceiling applies to every form on every plan, before the per-plan limit is checked.

Oversized documents fail with invalid_request before any processing happens — no credits are charged. See Credits & Pricing for the full plan table.

Precedence

If you send more than one form, they’re resolved in this order and the rest are ignored:

text  →  file  →  url

Sending none of them returns invalid_request: One of text, url, or file is required.

Two documents at once: /match

/match and /score/requirements score a candidate against a role, so each side gets its own prefixed field. Both sides accept all three forms, plus a fourth option: already-parsed JSON.

Mix them freely — a base64 résumé against a live job posting:

curl -X POST "$BASE/match" \
  -H "x-access-key: $REZMATCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"resume_file\": \"$(base64 -i resume.pdf)\",
    \"jd_url\": \"https://job-boards.greenhouse.io/acme/jobs/123456\"
  }"

Each document passed inline is parsed as part of the call and adds its full parse cost — the match above costs 3 + 6 + 4 = 13 credits. Passing candidate and role JSON you parsed earlier costs just the 3. When you’re screening a pool against one role, parse once and reuse the JSON; see Credits & Pricing for the arithmetic.