Extract

POST /v1/extract reads a GAEB DA XML file (the German Leistungsverzeichnis exchange format, .x81 to .x86) and returns its bill-of-quantities structure as a faithful, nested tree. It makes no decisions for you: the hierarchy, every text element and every quantity come back exactly as the file carries them. Composing that structure into embed-ready text is the job of Chunk, where you control the knobs.

Request

Multipart upload, one file per request, up to 20 MB:

curl

curl https://api.flinq.ai/v1/extract \
  -H "Authorization: Bearer flq_your_key_here" \
  -F "file=@LV_Rohbau.x83"

Python

import requests

with open("LV_Rohbau.x83", "rb") as f:
    r = requests.post(
        "https://api.flinq.ai/v1/extract",
        headers={"Authorization": "Bearer flq_your_key_here"},
        files={"file": f},
    )
tree = r.json()["tree"]  # nested group/item/remark nodes

JavaScript

import { readFile } from 'node:fs/promises';

const form = new FormData();
form.append('file', new Blob([await readFile('LV_Rohbau.x83')]), 'LV_Rohbau.x83');

const r = await fetch('https://api.flinq.ai/v1/extract', {
  method: 'POST',
  headers: { Authorization: 'Bearer flq_your_key_here' },
  body: form,
});
const { tree } = await r.json(); // nested group/item/remark nodes

Response

{
  "object": "extract",
  "format": "gaeb",
  "project": {
    "name": "Neubau Testhalle", "label": "TH-01", "currency": "EUR",
    "date": "2026-03-01", "owner_name": "Stadt Musterstadt",
    "owner_city": "Musterstadt", "boq_name": "LV Rohbau",
    "boq_label": "Leistungsverzeichnis Rohbau"
  },
  "tree": [
    {
      "type": "group", "rno": "01", "label": "Rohbau",
      "description": "Rohbauarbeiten Alle Rohbauarbeiten gemäß VOB/C.",
      "children": [
        {
          "type": "item", "rno": "01.02.0010",
          "short_text": "Ortbeton C25/30",
          "long_text": "Ortbeton der Festigkeitsklasse C25/30, Expositionsklasse XC2.",
          "additional_texts": ["Einbau lagenweise, verdichtet mit Innenruettler."],
          "qty": 12.5, "unit": "m2", "lump_sum": false,
          "unit_price": null, "total_price": null
        },
        { "type": "remark", "text": "Hinweis: Ausschreibung nach VOB/A." }
      ]
    }
  ],
  "warnings": [],
  "usage": { "prompt_tokens": 1234, "total_tokens": 1234 }
}

The tree

Four node types cover everything a GAEB file contains:

Type	What it is
group	A category (Gliederungsebene) with its own label, optional description and nested children.
item	A position: short_text, long_text, additional_texts (sub-descriptions, in document order), qty, unit, lump_sum, and prices.
remark	A Vorbemerkung: position-less prose between categories or items.
unparsed	Content the parser did not recognize, preserved as raw text. Always paired with a warning.

rno is the hierarchical position number: a position 0010 inside category 02 inside category 01 comes back as 01.02.0010.

Prices are preserved when present. Tender files (X83) carry no prices, so unit_price and total_price are null. Bid and award files (X84, X85) carry them, and the tree keeps them.

Tolerant parsing

The parser recognizes GAEB DA XML 3.2 and 3.3 automatically and does not require an ideal file. Whatever it cannot map to a known node is preserved as unparsed raw text with an entry in warnings, and an unparseable quantity becomes qty: null with a warning rather than a silent zero. A request fails with 422 in only two cases: the file is not well-formed XML, or its root element is not <GAEB>.

Billing

Billed flat per document: one parsed file costs EUR 0.02, whatever its size (see Pricing). The usage object still reports the token count of the extracted text (all labels, descriptions, position texts and remarks); treat it as a planning figure for the downstream chunk and embed calls, not as what this request billed.