PDF Central

Agent use case

PDF ingestion for RAG

Extract page-aware text, Markdown, structure, fields, or bounded chunks before indexing.

Recommended pipeline

Inspect first, choose text or Markdown based on layout, chunk with explicit overlap, and retain source hashes plus page evidence with each vector record.

  1. Read the live capability detail and project limits.
  2. Upload and complete an immutable input.
  3. Submit the operation with a stable idempotency key and hard cost ceiling.
  4. Read terminal job state, artifact records, hashes, and receipt.

Relevant operations

Start with the SDK

const pdf = new PDFCentralClient({ apiKey: process.env.PDF_CENTRAL_API_KEY });
const input = await pdf.upload(bytes, { filename: "input.pdf" });
const job = await pdf.run("inspect_pdf", {
  fileIds: [input.file_id],
  maxCostMicrousd: 200000
});

Availability and parameters may differ by operation. Use the linked capability detail as the source of truth.

Run the quick start