DocMint documentation hub
DocMint: SEC Filings, HTML, Word, Excel & PPTX to Markdown
Local-tested — not production-verified on Apify, not market-validated. Live usage metrics are unavailable. Status is v0.2.0, unreleased.
Turn SEC filings (HTML / inline XBRL), web pages and Office files (Word, Excel, PowerPoint) into clean, LLM-ready Markdown and structured tables in one run. Text-based PDFs are supported too. Pay only for what converts. Approved public pricing is Model C: $0.008 per document + $0.004 per billable page.
What it is built for — and what it does not claim
- Built for: SEC EDGAR filings and other HTML (hidden inline-XBRL headers and
display:nonecontent are dropped and never billed), Word (DOCX), Excel (XLSX) and CSV (every sheet), and PowerPoint (PPTX, including speaker notes). - Also supported: text-based PDFs (PDFs with a text layer) and TXT/Markdown. Text-based PDFs supported, complex tables may not extract cleanly.
- Tables it reads from the file’s own structure: HTML tables (including
rowspan/colspan), Word tables and spreadsheet sheets, returned as rows, CSV and Markdown. - Not claimed: optical character recognition. Image-only (scanned) PDFs with no extractable text layer return a
no_text_layerrow, no text, and cost $0. - Private and local-network URLs are blocked, including via redirects. Each file is up to 50 MB (HTML up to 10 MB).
Guides: SEC filing to Markdown, Word to Markdown, text-based PDFs to Markdown, Markdown for RAG, Model C worked examples.
Model C pricing and the 3,000-character page-equivalent rule
Every successfully converted document is billed $0.008 (document-processed) plus $0.004 per billable page (page-processed). Failed documents, documents skipped by limits or the safety breaker, and image-only PDFs with no text layer, cost $0. You never receive pages you were not charged for, and you never pay for pages you did not receive.
A billable page is:
- a PDF page;
- a PPTX slide;
- for DOCX, HTML and TXT: 1 page per 3,000 characters of extracted text, rounded up, minimum 1 per document;
- for CSV and Excel: each sheet is counted on its own, at 1 page per 3,000 characters of the sheet’s values (as CSV, without table formatting; empty rows and trailing empty cells are not counted), minimum 1 per sheet. A CSV file is one sheet.
The math below uses only that rule and Model C rates. Figures are exact.
| Input | Billable pages | Document fee | Page fees | Total |
|---|---|---|---|---|
| 1-page text-based PDF | 1 | $0.008 | $0.004 | $0.012 |
| 10-page text-based PDF | 10 | $0.008 | $0.040 | $0.048 |
| 30,000-character DOCX or HTML (including an SEC filing) | 10 | $0.008 | $0.040 | $0.048 |
| 3,001-character DOCX or HTML | 2 | $0.008 | $0.008 | $0.016 |
| Short DOCX or HTML (minimum 1 page) | 1 | $0.008 | $0.004 | $0.012 |
| XLSX with 2 small sheets | 2 | $0.008 | $0.008 | $0.016 |
| Workbook with three small sheets (minimum 1 page each) | 3 | $0.008 | $0.012 | $0.020 |
| One 60,000-character sheet (60,000 ÷ 3,000 = 20) | 20 | $0.008 | $0.080 | $0.088 |
| Failed document, or image-only PDF with no text layer | 0 | $0 | $0 | $0 |
maxPagesPerDocument (default 200, maximum 500) caps billable pages per document. A run-level maximum charge also caps output: the last document is cut to the pages that fit, with a warning, and the run then stops.
Safety limits (local-tested, not production-verified on Apify)
The Actor code includes a per-document processing limit of 120 seconds (or a memory cap), and a safety breaker that cannot be turned off: the run stops after 3 documents in a row hit that limit (configurable via breakerConsecutiveLimit, default 3), or once limit-hit documents exceed 20% of attempted documents, checked from the 5th attempted document (breakerFailurePercent, default 20). Remaining documents get a skipped row and cost $0. These behaviors were exercised in local security and billing tests on 19 November 2026 (see the project’s SECURITY notes). They have not been production-verified on the Apify platform.
Actor inputs and outputs
Inputs (from the Actor input schema): documents (public http(s) URLs), base64Documents ({filename, data, contentType} for API uploads), extractTables (default true), includePages (default false), maxPagesPerDocument (default 200), breakerConsecutiveLimit (default 3), breakerFailurePercent (default 20).
Each output row includes markdown, text, tables (rows, CSV and Markdown, with page or sheet), optional per-page pages, billable_pages, metadata, status, and warnings.
Integration quickstart
The Actor is not published on the Apify Store as of 20 November 2026. Set DOCMINT_ACTOR_ID to the published Actor id after the first Store publish. The sample URL below is the public IRS W-9 PDF used in the project’s own input prefill. The same input shape accepts a public SEC HTML filing URL.
Apify API
import { ApifyClient } from "apify-client";
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor(process.env.DOCMINT_ACTOR_ID).call({
documents: [{ url: "https://www.irs.gov/pub/irs-pdf/fw9.pdf" }],
extractTables: true,
includePages: false,
maxPagesPerDocument: 200,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const item of items) {
// item.markdown, item.text, item.tables, item.billable_pages, item.warnings, item.status
console.log(item.status, item.billable_pages);
}
n8n
Use n8n’s Apify integration (or an HTTP Request node) to start the Actor with the same input fields: documents or base64Documents, plus any of extractTables, includePages, and maxPagesPerDocument. In a later node, read the default dataset and map markdown, tables, and billable_pages. Make and Zapier can call the same Actor through their Apify integrations using the same input object. These recipes assume the Actor id is available after publish; they have not been production-verified against a live Store listing.
LangChain
from langchain_core.documents import Document
# `items` is the Actor dataset: each row has markdown, text, tables, billable_pages, warnings
docs = [
Document(
page_content=item["markdown"],
metadata={
"source": item.get("source"),
"billable_pages": item.get("billable_pages"),
"status": item.get("status"),
},
)
for item in items
if item.get("status") == "ok" and item.get("markdown")
]
LlamaIndex
from llama_index.core import Document, VectorStoreIndex
documents = [
Document(
text=item["markdown"],
metadata={
"source": item.get("source"),
"billable_pages": item.get("billable_pages"),
},
)
for item in items
if item.get("status") == "ok" and item.get("markdown")
]
index = VectorStoreIndex.from_documents(documents)
There is no first-party DocMint LangChain or LlamaIndex package. The snippets wrap the real Actor output fields only.
Join a documentation waitlist
If you want a note when the Apify Store listing is published, leave an email. This is optional and is not a paid signup.
See the site privacy notice.