API Reference

POST /extract

Extracts structured data from a PDF file according to a configured schema. Only the fields defined in the schema are returned — sensitive content in the document is never exposed.

Endpoint

http

POST https://reliant-production.up.railway.app/extract Headers: X-Reliant-Key: rel_... Content-Type: multipart/form-data ← set automatically by your HTTP client

This endpoint accepts multipart/form-data (not JSON). The PDF is sent as a file field alongside the other parameters as regular form fields. Do not set Content-Type manually — your HTTP client must set the correct multipart boundary.

Form fields

Field	Type		Description
file	file	required	The PDF to extract data from. Max 10 MB.
schema_id	string	required	ID of the schema that defines which fields to extract.
provider	string	required	anthropic \| openai \| gemini \| groq \| mistral
model	string	required	Model to use (e.g. `claude-sonnet-4-6`, `gpt-4o`).
user_id	string	required	Your Reliant user ID.
end_user_id	string	optional	Enables persistent memory scoped to this end user.
external_id	string	optional	Extra scope key to isolate memory threads for the same user.

How it works

Reliant extracts all text from the uploaded PDF and sends it as the user prompt to the configured LLM provider. The schema's JSON Schema definition is used as the output contract — the model is instructed to return only the declared fields. The extraction goes through the same reliability pipeline as /execute: automatic retries, provider fallback, quality scoring, and full audit logging.

If the PDF text exceeds approximately 20 000 tokens, the content is truncated and the response will include "truncated": true. Image-based (scanned) PDFs that contain no extractable text will return a 422 error.

Schema example

Define a schema with only the fields you want to expose. For a purchase order you might use:

json — schema definition

{ "type": "object", "required": ["supplier", "total_amount", "issue_date"], "properties": { "supplier": { "type": "string" }, "total_amount": { "type": "number" }, "issue_date": { "type": "string", "format": "date" }, "invoice_number": { "type": ["string", "null"] } } }

Only these four fields will ever appear in the response — any other content in the PDF (bank details, personal data, internal notes) is discarded.

cURL

curl

curl -X POST https://reliant-production.up.railway.app/extract \ -H "X-Reliant-Key: rel_..." \ -F "file=@invoice.pdf" \ -F "schema_id=YOUR_SCHEMA_ID" \ -F "provider=anthropic" \ -F "model=claude-sonnet-4-6" \ -F "user_id=YOUR_USER_ID"

TypeScript / Node.js

typescript

import { readFileSync } from 'fs' const form = new FormData() form.append('file', new Blob([readFileSync('invoice.pdf')], { type: 'application/pdf' }), 'invoice.pdf') form.append('schema_id', 'YOUR_SCHEMA_ID') form.append('provider', 'anthropic') form.append('model', 'claude-sonnet-4-6') form.append('user_id', 'YOUR_USER_ID') const response = await fetch('https://reliant-production.up.railway.app/extract', { method: 'POST', headers: { 'X-Reliant-Key': 'rel_...' }, body: form, }) const data = await response.json() console.log(data.output) // { supplier: "Acme Corp", total_amount: 1250.00, issue_date: "2026-07-07", invoice_number: "NF-00412" }

Python

python

import requests with open('invoice.pdf', 'rb') as f: response = requests.post( 'https://reliant-production.up.railway.app/extract', headers={'X-Reliant-Key': 'rel_...'}, files={'file': ('invoice.pdf', f, 'application/pdf')}, data={ 'schema_id': 'YOUR_SCHEMA_ID', 'provider': 'anthropic', 'model': 'claude-sonnet-4-6', 'user_id': 'YOUR_USER_ID', }, ) data = response.json() print(data['output'])

Response — 200 OK

json

{ "success": true, "status": "SUCCESS", "output": { "supplier": "Acme Corp", "total_amount": 1250.00, "issue_date": "2026-07-07", "invoice_number": "NF-00412" }, "truncated": null, "errors": null, "metadata": { "execution_id": "cmonavlcs0004ph01ska7i1tq", "schema_version": 1, "attempts": 1, "latency_ms": 1240, "tokens_used": 834, "provider": "anthropic", "model": "claude-sonnet-4-6", "quality_score": null, "quality_feedback": null, "quality_passed": null } }

Error responses

Status	Error	Cause
`400`	Validation Error	Missing required field or invalid provider.
`413`	File Too Large	PDF exceeds 10 MB.
`422`	PDF Parse Error	File is corrupted or image-based (no extractable text).
`207`	—	All retries failed; safe_fallback was used if configured.

← PreviousPOST /execute Next →POST /execute-batch