Dokumentli is a vision language model built for business documents. Send pages and a spec, get structured JSON back. OpenAI-compatible, billed per page.
from openai import OpenAI
client = OpenAI(
base_url="https://api.dokumentli.ai/v1",
api_key="dk_live_...",
)
r = client.chat.completions.create(
model="dokumentli",
messages=[{"role": "user", "content": [
{"type": "file", "file": {
"filename": "invoice.pdf",
"file_data": f"data:application/pdf;base64,{pdf}"}},
{"type": "text", "text": json.dumps(spec)},
]}],
)
print(json.loads(r.choices[0].message.content))
The setup you know from the OpenAI API. Change the base URL, keep your SDK.
E-mail, six-digit code, company name. No password.
Shown once, stored hashed. Revoke any time.
PDF or images plus a JSON spec to /v1/chat/completions.
Top up by card when your free pages are used. Credit burns down per page.
Prepaid credit in EUR. Failed requests are not billed. Higher tiers bring lower prices and higher rate limits.
100 pages once, no card
No minimum, top up from €50
€500 monthly minimum
€2,000 monthly minimum
All prices excl. VAT. One page is one PDF page or one image. Credit is valid for 12 months.
Invoices, contracts, claims, patient files. Processed in Germany, by a European company.
Gateway and GPU nodes run in data centres in Germany. No US cloud in the document path.
Documents are held in memory for the request and discarded. Logs contain metadata only: pages, latency, status.
API traffic is never used to train or evaluate models.