contexa.works
← tutorials

PDF to JSON

Convert any PDF into structured data with typed columns, tables, and sections. This API is asynchronous — submit a job, poll for status, then fetch the result.

1

Submit your PDF

POST a PDF to start an extraction job. The API returns immediately with a job ID while processing continues in the background. Multi-page documents are automatically split and extracted page-by-page with rolling context for accurate table continuation.

curl — upload a file
curl -X POST https://api.contexa.works/api/v1/pdf \
  -H "x-rapidapi-key: YOUR_API_KEY" \
  -F "file=@report.pdf"
response — 202 Accepted
{
  "jobId": "e5f6a7b8-...",
  "status": "processing"
}
curl — base64 with page range and callback
curl -X POST https://api.contexa.works/api/v1/pdf \
  -H "x-rapidapi-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "base64": "'$(base64 -i report.pdf)'",
    "pages": "1-5",
    "callbackUrl": "https://your-app.com/webhooks/contexa"
  }'

Use the pages parameter to focus on specific pages (e.g. "1-5"). This speeds up extraction and reduces cost for large documents.

2

Poll for completion

Check the job status by polling with the job ID. The status will be processing, completed, or failed. Or skip polling entirely by providing a callbackUrl.

curl — check status
curl https://api.contexa.works/api/v1/jobs/e5f6a7b8-... \
  -H "x-rapidapi-key: YOUR_API_KEY"
response — completed
{
  "id": "e5f6a7b8-...",
  "status": "completed",
  "documentType": "pdf",
  "fileName": "report.pdf",
  "processingTimeMs": 4210,
  "createdAt": "2024-06-15T10:30:00Z",
  "completedAt": "2024-06-15T10:30:04Z"
}
3

Fetch the result

Once the job is completed, fetch the structured data. The response includes document metadata, sections, tables with typed columns, and key-value pairs.

curl — get result
curl https://api.contexa.works/api/v1/jobs/e5f6a7b8-.../result \
  -H "x-rapidapi-key: YOUR_API_KEY"
response
{
  "id": "e5f6a7b8-...",
  "status": "completed",
  "processingTimeMs": 4210,
  "data": {
    "title": "Q4 2024 Financial Report",
    "author": "Finance Team",
    "date": "2024-12-31",
    "documentType": "report",
    "sections": [
      {
        "heading": "Revenue Summary",
        "content": "Total revenue grew 12% year-over-year...",
        "tables": [
          {
            "name": "Revenue by Region",
            "columns": [
              { "name": "Region", "type": "string", "unit": null },
              { "name": "Q4 Revenue", "type": "currency", "unit": null },
              { "name": "YoY Growth", "type": "percentage", "unit": null }
            ],
            "rows": [
              ["EMEA", 5200000, 14.2],
              ["North America", 4800000, 11.5],
              ["APAC", 3100000, 8.7]
            ]
          }
        ]
      }
    ],
    "keyValuePairs": [
      { "key": "Report Period", "value": "Q4 2024" },
      { "key": "Currency", "value": "USD" }
    ],
    "summary": "Q4 revenue totalled $13.1M across three regions..."
  }
}
Column types detected:
stringnumbercurrencypercentagedateboolean

Units like pence, bps, or cents are extracted into a separate unit field.

JavaScript — full flow
const BASE = "https://api.contexa.works/api/v1"

// 1. Submit
const submit = await fetch(`${BASE}/pdf`, {
  method: "POST",
  headers: { "x-rapidapi-key": "YOUR_API_KEY" },
  body: formData,
})
const { jobId } = await submit.json()

// 2. Poll until done
let job
do {
  await new Promise(r => setTimeout(r, 2000))
  const res = await fetch(`${BASE}/jobs/${jobId}`, {
    headers: { "x-rapidapi-key": "YOUR_API_KEY" },
  })
  job = await res.json()
} while (job.status === "processing")

// 3. Fetch result
if (job.status === "completed") {
  const res = await fetch(`${BASE}/jobs/${jobId}/result`, {
    headers: { "x-rapidapi-key": "YOUR_API_KEY" },
  })
  const { data } = await res.json()

  for (const section of data.sections) {
    for (const table of section.tables) {
      console.log(table.name, table.rows.length, "rows")
    }
  }
}
Python — full flow
import requests, time

BASE = "https://api.contexa.works/api/v1"
headers = {"x-rapidapi-key": key}

# 1. Submit
r = requests.post(
    f"{BASE}/pdf",
    headers=headers,
    files={"file": open("report.pdf", "rb")},
)
job_id = r.json()["jobId"]

# 2. Poll
while True:
    time.sleep(2)
    status = requests.get(
        f"{BASE}/jobs/{job_id}",
        headers=headers,
    ).json()
    if status["status"] in ("completed", "failed"):
        break

# 3. Fetch result
if status["status"] == "completed":
    result = requests.get(
        f"{BASE}/jobs/{job_id}/result",
        headers=headers,
    ).json()
    for section in result["data"]["sections"]:
        for table in section["tables"]:
            print(f"{table['name']}: {len(table['rows'])} rows")
Tip: For multi-page PDFs, tables that span pages are automatically merged by matching column names. You don't need to handle pagination yourself.
Try it in the playgroundAPI reference
TermsPrivacyContact