Every checked heading and figure comes back as Markdown

Superlinked gives your agent one API to turn PDFs, decks, papers and forms into Markdown it can chunk and index.

nvidia-q4-fy2025-cfo-commentary.pdf

Markdown

## First Quarter of Fiscal 2026 Outlook

Outlook for the first quarter of fiscal 2026 is as follows:

- Revenue is expected to be $43.0 billion, plus or minus 2%.
- GAAP and non-GAAP gross margins are expected to be 70.6% and 71.0%, respectively, plus or minus 50 basis points.
- GAAP and non-GAAP operating expenses are expected to be approximately $5.2 billion and $3.6 billion, respectively.
- GAAP  and  non-GAAP  other  income  and  expense  are  expected  to  be  an  income  of approximately $400 million, excluding gains and losses from non-marketable and publicly-held equity securities.
- GAAP and non-GAAP tax rates are expected to be 17.0%, plus  or  minus  1%,  excluding  any discrete items.

All 9 lines of the First Quarter of Fiscal 2026 Outlook section

25 of 25 checks passed, on four documents of four different kinds

First page of the two-column Docling technical report

The two-column paper comes back in reading order

## Abstract

This technical report introduces Docling , an easy to use, self-contained, MITlicensed open-source package for PDF document conversion. It is powered by state-of-the-art specialized AI models for layout analysis (DocLayNet) and table structure recognition (TableFormer), and runs efficiently on commodity hardware in a small resource budget. The code interface allows for easy extensibility and addition of new features and models.

Abstract, introduction and architecture arrive in that order, one column at a time, which is what a chunker needs and what a naive text dump of a two-column PDF interleaves. One hyphen does not survive the trip: MIT-licensed was broken across a line in the PDF and comes back as MITlicensed.

Docling Technical Report, page 1

SiriusPoint financial results presentation slide

Every figure in the highlights table comes back in its own cell

## Q1 2025 FINANCIAL RESULTS

| Financial Highlights                    | Financial Highlights   | Financial Highlights   | Financial Highlights   |
|-----------------------------------------|------------------------|------------------------|------------------------|
| $ numbers in USD millions               | Q1'24                  | Q1'24                  | Q1'25                  |
| Gross Premiums Written                  | $881                   | $881                   | $990                   |
| Net Premiums Written                    | $627                   | $627                   | $752                   |
| COR (%)                                 | 91.4%                  | 91.4%                  | 95.4%                  |

Showing 5 of the 13 rows the model returned for this table

Every number keeps the cell it was printed in, down to the percent signs. The header row is where it gives way: Financial Highlights repeats across all four cells and the Q1'24 column comes back twice, so nothing in the Markdown says which quarter a figure belongs to. Further down the same table the two copies of Q1'24 stop agreeing, $91 in one and Underlying 5 $108 in the other.

SiriusPoint Q1 2025 Investor Presentation, page 13

FEMA proof of loss form with sections, fields, and checkboxes

Every field and every checkbox choice on the form page comes back

- [ ] Individual or Household

- [ ] Business

- [ ] Government

- [ ] Tribal Nation/Tribal Government

- [ ] Not-for-Profit

- [ ] Other:

All twelve amount labels on the page shown come back, and so do the six claim-type choices, each as an unticked Markdown checkbox. The amounts arrive in a run of their own: 26 bare $ symbols on lines that carry no label, so nothing in the Markdown says which amount belongs to which category. The six choices land under `## Real Property`, 32 lines after the question they answer.

FEMA Hermit’s Peak/Calf Canyon Proof of Loss

25 of 25 checks passed over the 4 documents in this run: the three on these cards, and the financial report in the hero. 18 of the checks look for a named heading or figure in the Markdown that came back, 3 check that sections arrive in document order, and 4 count the tables. None of them looks at whether a table keeps its column headings, which is where this run thins out; the caption on each card says where.

The whole document comes back as one Markdown string

View on GitHub

from sie_sdk import SIEClient
client = SIEClient(
    api_key="sk-sie-…",
    base_url="https://api.superlinked.com",
)
# Parse a document → clean markdown, tables and layout kept.
result = client.extract(
    "modeldocling",
    {"document": filePDFnvidia-q4-fy2025-cfo-commentary.pdfbrowse},
)
print(result["data"]["markdown"])

document_bytes=$(base64 < 'nvidia-q4-fy2025-cfo-commentary.pdf' | tr -d '\n')
curl https://api.superlinked.com/v1/extract/docling \
  -H "Authorization: Bearer sk-sie-…" \
  -H "Content-Type: application/json" \
  -d "{\"items\":[{\"document\":{\"data\":\"$document_bytes\",\"format\":\"pdf\"}}]}"

Build the "Doc to Markdown" capability into my app using the Superlinked Inference Engine (SIE).

Context
- SIE is an OpenAI-style inference API. Python SDK: `from sie_sdk import SIEClient`; TypeScript: `@superlinked/sie-sdk`.
- Base URL: https://api.superlinked.com (or my regional endpoint). Auth: Bearer key from env `SIE_API_KEY` (never hard-code it).
- Model: docling (SIE primitive: /extract). Keep the model id configurable.

Task
- Input: an uploaded document (PDF / Office / scan).
- Behaviour: return clean markdown for the document, preserving tables and reading order
- Call the selected SIE primitive once per request and map the response into your domain type.

Deliverables
- A typed client wrapper, an application-level function for this task, error handling for timeouts/empty input, and unit tests with a stubbed client.
- Wire it into my existing stack (ask me which framework if unclear) and add a short usage example.

Run

Output

Q4 Fiscal 2025 Summary

GAAP	GAAP	GAAP	GAAP	GAAP	GAAP
($ in millions, except earnings per share)	Q4 FY25	Q3 FY25	Q4 FY24	Q/Q	Y/Y
Revenue	$39,331	$35,082	$22,103	Up 12%	Up 78%
Gross margin	73.0 %	74.6 %	76.0 %	Down 1.6 pts	Down 3.0 pts

Extracts clean, agent-ready markdown
Keeps tables and layout intact

PRICE

$ / 1k pages ↓

$0.397 SIE docling

$1.5 AWS Textract

$1.5 Google Enterprise Document OCR

$1.5 Azure Document Intelligence Read

Mistral OCR 4 $4

QUALITY

olmOCR ↑

SIE docling 0.32

0.30

0.29

LATENCY

p50 ms ↓

SIE docling 214ms

240ms

260ms

270ms

300ms

Deploy your way

Managed Cloud

Full compute toolkit for your agents with zero ops.

No idle GPUs, pay for what you use
Fits your stack: SDK, API, CLI, MCP
Zero lock-in, self-host the same stack
SOC 2 Type 2, US or EU data residency

$50 Example credit balance

1.3B tokens embedded

120k pages parsed

27M tokens generated (1:1 in:out)

Get started

no credit card required

Self-host with K8s

Easy & scalable deployment in your own cloud.

Terraform to your cloud in minutes
Apache-2.0, same engine as Cloud
Scales to zero, no bill between jobs
Per-tenant pools, no noisy neighbors

Agent prompt

Deploy SIE to our AWS account with the superlinked/sie/aws Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our GCP project with the superlinked/sie/google Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our Azure AKS cluster via helm install. Requirements: superlinked.com/docs/deployment

Deploy guide

Run locally

Run the same models on your own machine.

Runs on NVIDIA GPU or Apple Silicon
One command, no Docker or cluster
All 100+ Cloud models, fully offline
Same SDK and IDs, no code changes

pip install "sie-server[local]" && sie-server servepip install "sie-server[local]" && sie-server serve --device cuda

Quickstart