Privacy and PDF handling
TradeGuard is a static public demonstration workspace. It is intended for fictional, synthetic or anonymised information—not confidential customer, bank or trade documents.
How the public PDF review works
- Up to five selected PDFs are read in the browser using the pinned self-hosted PDF.js 6.3.289 build included at repository root in this public package; each file is limited to 10 MB and the combined selection to 40 MB.
- The browser extracts available selectable text and applies the visible transaction-monitoring rules locally, with PDF.js evaluation disabled for this parser path. OCR for scanned/image-only documents is not included in this public build.
- The app does not provide a TradeGuard server upload endpoint, user account, case database or cloud case store. The PDF bytes are passed to the in-page parser rather than to a TradeGuard API.
- After analysis, the file input, temporary text reference and PDF.js document object are cleared by the page.
The file reference and temporary extracted text are cleared by the page after analysis. Identifier-like text can produce a visible warning, but the public build cannot prove whether a value is real or fictional; the user remains responsible for uploading only fictional, synthetic or anonymised data. This is not a promise that the device, browser, hosting provider, network provider or an uploaded source document has no separate history. Confirm the deployed build in the browser Network tab before any controlled pilot, and do not use confidential information in this public demo.
Two PDF input paths
Full statement mode
Use this only with fictional or anonymised statements. Available transaction values and limited context signals may be pre-filled. Review and correct the fields before relying on the indicative output.
Transaction history only mode
Use a PDF containing only anonymised transaction rows. The system extracts transaction patterns and amounts, then lets the reviewer enter an anonymised reference, profession, profile, baseline and supporting context manually. If no account reference is supplied, the report uses an anonymised case reference and the internally generated Case ID remains the primary identifier.
What this public build does not provide
- Login, role-based access or user management.
- Live sanctions, PEP, adverse-media or restricted-goods screening.
- Bank, SWIFT, core-banking or trade-platform integration.
- Saved cases, server-side audit logs or institutionally certified records.
- Regulatory, legal or final financial-crime determinations.
Analytics and external services
The public app does not intentionally send transaction content to an analytics service. The site loads from GitHub Pages and the transaction-monitoring page uses the self-hosted PDF.js build included in this package. GitHub Pages, the browser and network providers may have their own infrastructure logs governed by their respective policies.
Institutional deployment
Before a pilot uses real data, the deployment should have a formal privacy notice, lawful processing basis, retention and deletion policy, data-residency decision, access controls, encryption, vendor agreements, security testing and an approved incident-response process.
Contact
TradeGuard is an IT product of RegTech Nexus AI, a sister concern of Travel To Know. Corporate office: 28/A-2, Toyenbee Circular Road, 1st Floor, Motijheel Commercial Area, Dhaka-1000, Bangladesh. Phone: +880 1772 282925. For a pilot discussion, contact regtechnexusai@gmail.com.
The public build does not establish a formal retention period for pilot-request emails or a complete institutional data-controller notice. Do not send confidential or personal customer information by email; confirm the controller, lawful basis, retention and cross-border handling in a written pilot arrangement.
Document Assessment and OCR
The Document Assessment workspace processes selected PDF and image files in the browser. It uses selectable-text extraction and optional OCR-assisted extraction for scanned or image-based pages. Files are not uploaded to a TradeGuard server, but the optional OCR engine and language resources are loaded from a CDN. Use fictional, synthetic or anonymised documents only and verify all OCR results against the original file.