Home / Patra · In development — not yet publicly available
Every tool gives you a file. Patra gives you the file and the proof of what was done to it — for anyone who needs dependable PDF tools, for firms that manage documents, and for organisations preparing documents for their knowledge systems.
01 / 07
A document arrives — here a synthetic three-page credit policy from a fictional lender — and Patra inspects it before anything else: how many pages, whether it is encrypted, what it contains, what could be risky about it.
02 / 07
Every operation runs in a disposable worker with no network access and fixed limits — 512 MiB of input, 10,000 pages, 300 seconds — enforced by the operating system, not by the program politely declining.
03 / 07
A person marks what must go; Patra burns those regions into the page so the text beneath is gone, not hidden under a box. Redaction is geometric and human-marked — Patra does not guess what is sensitive.
04 / 07
The file is repacked without losing anything, and a page is taken out and put back where it belongs. After each operation a slip records what ran, on which input, and what it produced.
05 / 07
The slips join into one processing receipt: the source digest, every step with its input and output digests, how exact it was, where it ran, and retained: false. Its own digest detects accidental or casual tampering; it is not a signature.
06 / 07
A private library keeps each document with its receipts and makes it searchable with citations — a single-owner workspace, built but not yet hosted. The document-management layer firms need — versions, matters, permissions, retention and workflows — is drawn in outline because it is in development In development
07 / 07
Patra prepares a document for a knowledge system: its text, layout and tables, each with the place on the page it came from. The hand-off itself — into BanyanGraph or an organisation's own knowledge system — is in development In development Patra prepares and hands off; the knowledge base does the ingesting.
The processing receipt records provenance for document work. The example below shows a configurable profile with optional identifiers for regulated, legal and client-matter workflows.
For individuals and professionals
Convert to and from PDF, compress, redact, merge, split and re-order, OCR in English and Hindi, protect and unlock, watermark and compare — each in an isolated worker, returned with a receipt.
31 operations built and tested (registry, 25 Sep 2026) · a hosted version for the public In development
For legal, CA and CS firms, MSMEs and banks
A private library with cited search, export and a receipt for every document, growing into versions, matters, permissions, retention and workflows.
Single-owner workspace built, not hosted · document management In development
For organisations with knowledge systems
Documents and data normalised and structured with provenance, ready for a knowledge system — BanyanGraph or your own.
Normalisation with provenance built · the hand-off In development
Real footage of Patra's current workbench: a document is received, inspected and extracted, and the receipt is opened. It runs on our own machines; it is not a hosted service.
Core provenance fields are shown with optional profile identifiers. The identifiers label a record; they do not change what processing the receipt claims occurred.
| Field | Records |
|---|---|
source_digest | A fingerprint of the file you gave it |
steps[] | Each operation, its version, its input and output fingerprints, and how exact it was: byte-identical, structural or semantic; lossless, vector-preserving, rasterised or reconstructed |
executed_on | Where it ran — an isolated worker |
human_confirmed | Whether a person confirmed the step, as with redaction |
retained | Whether anything was kept |
receipt_digest | A fingerprint of the receipt itself — it detects accident and casual tampering; it is not a signature |
regulatory_identifiers[] | Optional regulator, instrument, section or clause identifiers |
legal_references[] | Optional legal matter references, such as a case or filing identifier |
client_matter_ids[] | Optional identifiers from the client's matter system |
custom_json | Optional client-defined JSON metadata, subject to the configured profile |
This is an example profile, not a claim that every receipt contains every identifier or that Patra exposes an API. Transient processing keeps nothing: the anonymous workbench has no storage behind it by construction. The private library, when it is hosted, will have its own stated storage, access and deletion rules.
A synthetic two-clause board resolution, written for this demonstration. Patra 0.1.0 produced this receipt on . It is a demonstration run, not a customer document.
—
The document core inspects, transforms and verifies files with processing receipts and no-egress controls. A private knowledge workspace over material a professional controls is in development. Current product stages are listed on the capabilities page.
Tell us what your documents go through and what you need to be able to prove afterwards. We will tell you plainly what Patra does today and what is still being built.
