ChemTrace Pro: A Multi-Stakeholder Chemical Traceability Platform

Subtitle: From Import Declaration to Public and Environmental Safety: An Integrated Approach to Chemical Management Compliance in Suriname

Authors: Gabor Tjong A Hung, Donovan Bogor

Date: June 2026

Version: 1.0

Executive Summary

ChemTrace Pro is an integrated chemical tracking and compliance platform built for Suriname's National Environment Authority (NMA) for the execution and control of Article 35 from the Environmental Framework Act. It replaces disconnected paper-based workflows with a unified system that traces every chemical container from the moment it is declared for import to the moment a first responder scans its QR code at an emergency scene.

The platform serves four distinct groups — NMA oversight staff, importing companies, warehouse inspectors, and the general public — through five purpose-built applications connected to a single source of truth. Rather than treating regulatory compliance as a back-office paperwork exercise, ChemTrace Pro embeds compliance into the physical supply chain: every can, cylinder, and drum will receive a unique digital identity encoded on both an RFID tag and a printed QR code — a capability the NMA has never had before. That identity binds the container to its import approval, its Safety Data Sheet (SDS), and its GHS hazard classification for its entire lifecycle.

This document presents the architecture, design rationale, and operational flows of ChemTrace Pro as a reference model for national-scale chemical traceability.

1. Problem Statement

Suriname's chemical import regulatory workflow has historically operated across disconnected systems: paper SDS binders, spreadsheet inventories, manual warehouse ledgers, and ad-hoc communication between consignees and regulators. For the NMA, this fragmentation creates critical operational gaps.

The Traceability Gap

A chemical container sitting on a warehouse floor had no reliable digital tether to its import declaration, its approval status, or its safety documentation. A regulator conducting an inspection could see a can of pesticide but could not confirm, in real time, whether that can was legally imported, whether its SDS had been reviewed, or whether it had been approved for release. The only way to answer those questions was to cross-reference multiple paper documents, often stored in different locations.

Slow and Error-Prone SDS Management

Safety Data Sheets arrived as PDFs, were filed manually, and were difficult to retrieve during inspections or emergencies. Extracting structured hazard data — GHS pictograms, H-codes, P-codes, UN transport numbers — was done by hand, introducing transcription errors and delays. In many cases, the extracted data never made it into a searchable digital format at all.

No Instant Hazard Awareness

Firefighters, medical personnel, or concerned citizens at an incident scene had no direct way to look up the contents of a chemical container. The information existed, but it was locked in PDFs or internal databases that required logins, phone calls, or physical visits to access. In an emergency, minutes matter.

Offline Inspections Were Impractical

Field inspectors conducted warehouse audits with clipboards and printed lists. Remote warehouses often lacked reliable mobile connectivity, so inspectors could not access live records during the audit. Discrepancies between declared and physical inventory were noted manually and transcribed later, creating a second opportunity for error.

Undeclared Imports Went Undetected

A warehouse might contain more physical inventory than was declared on import paperwork. Without scanning individual containers and comparing them against a digital manifest, there was no systematic way to detect undeclared or excess stock.

Absence of System-Generated Labels

The NMA does not currently produce RFID labels from approved batch data. Chemical containers arrive without scannable digital identities that bind them to their import approval and Safety Data Sheet. Introducing integrated label printing is a planned enhancement: labels will be generated only after batch approval and will reflect the exact approved data, preventing any mismatch between the physical label and the system record.

Who Is Affected

Existing generic ERP or inventory systems do not solve these problems because they treat chemicals as ordinary stock-keeping units. They lack native understanding of GHS classifications, UN numbers, regulatory approval workflows, or RFID label generation. ChemTrace Pro was therefore designed as a domain-specific platform rather than a configuration of generic software.

2. Background and Related Work

Regulatory Context

Chemical compliance sits at the intersection of environmental law, occupational safety, and supply-chain logistics. Two international frameworks are directly relevant to ChemTrace Pro's domain:

• The Globally Harmonized System (GHS) of Classification and Labelling of Chemicals standardizes hazard pictograms, signal words, hazard statements (H-codes), and precautionary statements (P-codes). Every SDS is expected to follow GHS conventions.
• The Montreal Protocol on Substances that Deplete the Ozone Layer imposes strict controls on the import and consumption of ozone-depleting substances. Suriname, as a party to the Protocol, must monitor and report ODS imports annually.

In practice, regulators must reconcile three data sources for every imported chemical: the import declaration, the manufacturer's SDS, and the physical container label. When these sources disagree — or when one is missing — enforcement becomes difficult and public safety is compromised.

Existing Approaches and Their Limitations

Several categories of tools address fragments of this problem:

• Generic ERP systems (SAP, Oracle NetSuite) track inventory and suppliers but treat chemicals as ordinary SKUs. They do not natively understand GHS classifications, UN transport codes, or regulatory approval workflows.
• SDS management platforms (Chemwatch, 3E Protect) centralize Safety Data Sheets but are typically siloed from import-declaration systems and warehouse operations.
• RFID asset-tracking systems track physical containers but rarely tie tags to regulatory documents or approval states.
• Custom government portals are common in customs and environmental agencies, but they usually serve a single stakeholder (e.g., the importer) and do not integrate field inspection tools or public-facing hazard disclosure.

The Missing Piece

What is absent from the landscape is a platform that treats the physical chemical container as the center of the compliance record. Such a platform must:

1. Link each container to its regulatory paperwork (SDS, import approval).
2. Make hazard data instantly accessible to the public via a durable, scannable code.
3. Support offline warehouse inspections with automatic discrepancy detection.
4. Provide separate, role-appropriate interfaces for regulators, importers, inspectors, and the public.

ChemTrace Pro closes this gap by combining an edge-hosted API, offline-capable mobile scanning, and static public hazard pages into one coherent, multi-actor system.

3. Solution Overview

Core Concept: One Container, One Identity

ChemTrace Pro applies a simple but powerful rule: every physical chemical container will receive a unique TagIdentifier (TID). That TID is encoded on an RFID chip and printed as a QR code on the container's label — a completely new process for the NMA. Scanning either identifier resolves to a single digital record containing:

• the chemical product and batch,
• the Safety Data Sheet and extracted hazard data,
• the supplier and consignee,
• the NMA approval status,
• the current location and safety status.

This creates a closed loop: consignees declare imports digitally, NMA approves them digitally, the system generates RFID labels digitally (a new capability), and field inspectors verify physical inventory against the same digital record.

System Metaphor: A Digital Passport

ChemTrace Pro functions like a passport system for chemical containers. Each container gets a passport (the TID) that follows it from the port of entry to the warehouse shelf to the point of sale or use. Anyone with the right credentials — regulator, inspector, business user, or citizen — can scan the passport to see the container's history and hazard profile.

Key Design Principles

1. Single source of truth. All five applications read from and write to the same database via a unified API. There are no ad-hoc data copies.
2. Offline-first field operations. The Android scanner downloads inspection data before entering a warehouse and syncs results afterward. Connectivity is not required during the audit.
3. Public safety by default. QR hazard pages are intentionally public and static. In an emergency, there is no login barrier between a first responder and the information they need.
4. Domain-driven data model. Hazard classifications, UN numbers, container types, and approval states are first-class concepts, not afterthoughts.
5. Edge deployment. The backend runs on Cloudflare Workers, and static hazard pages are served from Backblaze B2 via Cloudflare's CDN. This places data close to users in Suriname and keeps operational costs predictable.

The Five Applications

4. Architecture and Design

High-Level System Architecture

Operational Flows

The following diagrams illustrate how the system operates across its multiple stakeholders. Each diagram is presented as a sequence flow with swimlanes to make actor responsibilities explicit.

Flow 1: Batch Rejection with Feedback and Resubmission

Not every import request is approved on the first submission. This flow shows how NMA staff communicate rejection reasons back to consignees, who can then edit and resubmit without losing their original draft.

Key design points:

• Rejection feedback is stored as a text field on the batch record, making it auditable and visible to both staff and consignees.
• Resubmission reuses the same batch record with updated data, preserving the submission history rather than creating a duplicate.
• The portal highlights which fields were flagged by NMA, reducing guesswork for the consignee.

Flow 2: Import Request, Approval, and Label Generation

This is the central regulatory workflow. A consignee submits a chemical import request; NMA reviews and approves it; the system then generates RFID labels and public hazard data as an integrated, approval-gated step.

Key design points in this flow:

• Integrated label printing is entirely new. ChemTrace Pro introduces BarTender-driven RFID label generation as a capability the NMA has never had before. Labels are produced only after batch approval, ensuring the physical label always matches the approved system record.
• The batch record stores a snapshot of the consignee and supplier at the time of creation. This preserves historical accuracy even if the parent company record is later updated.
• Each TagIdentifier gets its own static JSON file on B2. This decouples public hazard access from the backend API — the QR PWA never hits the API, it fetches static files directly from the CDN.
• BarTender receives a print job containing the list of TIDs and chemical metadata. The printer encodes the TID onto the RFID chip and prints the QR code URL (qr.nma.sr/{tid}) on the label surface.

Flow 3: Warehouse Inspection with Offline Scanning

This flow supports NMA field inspectors who audit warehouses, often in locations with poor or no mobile connectivity.

Key design points in this flow:

• The Android app uses a local SQLite database (Room) that mirrors a subset of server state for the active inspection. All scan operations read and write locally.
• Discrepancy detection is immediate. The app maintains a running count per chemical product. If the inspector scans 45 units of Mebrom 98 but only 42 were declared, the app flags the +3 difference instantly.
• Sync is explicit and reportable. The inspector chooses when to upload. The backend generates a formal inspection report with discrepancies and safety flags for NMA review.

Flow 4: Public QR Scan at an Emergency Scene

This flow is optimized for speed and reliability. No login, no backend API call, no dynamic rendering.

Flow 5: Safety Incident Reporting from Field to Follow-Up

This flow shows how an inspector who discovers a damaged or leaking container triggers a safety alert that reaches NMA staff and the responsible consignee.

Key design points:

• Safety flags are recorded locally during offline inspection and synced with the rest of the scan data. The inspector does not need connectivity to flag a problem.
• Photos and notes travel with the sync payload and are stored in D1, giving NMA staff visual evidence without requiring a return visit.
• The safety alert system creates an auditable follow-up trail from discovery through resolution, with both NMA staff and the consignee held accountable.

Flow 6: User Onboarding and Granular Permission Assignment

This flow illustrates how NMA administrators create accounts for external company users and assign them fine-grained permissions within their organizations.

Flow 7: Chemical Spill Emergency Response and Origin Investigation

This flow demonstrates the system's most critical public-safety and regulatory-enforcement scenario: a chemical spill at an incident site. It shows how two different actors — a first responder and an NMA investigator — extract different but complementary information from the same container identifier.

Key design points:

• Same container, two different views. The first responder sees hazard data, SDS, and emergency contacts. The NMA inspector sees regulatory metadata: batch ID, importer, approval history, storage assignment, and transfer chain. Both actors scan the same QR code, but the system serves context-appropriate information.

• The QR PWA is intentionally stateless. It fetches static JSON from B2 via the CDN. This means it works even if the Workers API is down, congested, or physically unreachable from the incident site. First responders get hazard data in seconds without authentication barriers.

• Route inference vs. confirmation. The system can suggest a likely distribution path by matching the chemical and unit size against approved transfer requests. However, without mandatory RFID scans at every handoff point, this remains an inference. The whitepaper explicitly notes this limitation because it informs both investigation procedures and future policy: enforcing RFID scans at every transfer would upgrade these inferences to confirmations.

• Incident reports create audit trails. When an investigator logs a spill, the incident record is linked to the batch, the chemical, and the inferred transfer history. This enables NMA compliance staff to review patterns — for example, whether a specific warehouse, supplier, or transport method appears repeatedly in incident reports.

Flow 8: Chemical Product Lifecycle and Retest Management

Chemicals are not static records. They move through states from initial registration to approval, then through periodic retesting, and eventually to expiration or disposal. This flow shows how the system tracks that lifecycle.

Key design points:

• Status fields drive the lifecycle. A chemical product has status (pending_approval, approved, rejected). A batch has its own status (pending_approval, approved, delivered, rejected, quarantined, disposed, expired). This separation allows the chemical master data to remain approved while individual batches move through their own lifecycles.
• Retest scheduling is proactive. The system tracks retest_date and expiration_date on each batch. As these dates approach, staff receive alerts and can schedule inspections before the chemical becomes non-compliant.
• Quarantine as an intermediate state. A failed retest does not immediately delete the batch; it moves to quarantined, giving the consignee a documented window to arrange disposal.
• Disposal closes the audit loop. The consignee uploads a disposal certificate, which is stored on the batch record. This creates an unbroken chain from import declaration through lab retesting to final disposal.

Key design points:

• Company admins can manage their own team's permissions within boundaries set by NMA. This decentralizes day-to-day access control without sacrificing oversight.
• Permissions are stored as JSON in the user_companies table, making the system extensible: new permissions can be added without schema migrations.
• effectivePermissions is computed at login: for admins it expands to all permissions; for other roles it returns the explicit allow-list. The frontend uses this computed list to show or hide UI elements.
• Users can belong to multiple companies, each with a different role and permission set. A logistics contractor who works with two importers, for example, can switch between company contexts within the same login session.

Key design points in this flow:

• The QR PWA is fully static. It is served from Cloudflare Pages (or equivalent static hosting) and fetches per-tag JSON files from B2 via the CDN.
• No backend API call is required to display hazard information. This means the page loads even if the Workers API is temporarily down or congested.
• Edge caching is critical. Cloudflare caches the small JSON files (~2–5 KB each) at its global edge locations. After the first fetch, subsequent requests for the same tag are served from the edge without hitting B2.
• The optional Report feature is the only step that contacts the backend API. It POSTs GPS coordinates and a message, creating an incident record in D1.

Data Model Overview

The database schema is organized around the container-as-identity concept. Key entities and their relationships:

Notable patterns:

• Hybrid ID strategy. API-facing identifiers are UUIDs (prevent enumeration, easy distributed generation). Internal database joins use integer IDs where appropriate for smaller indexes.
• JSON columns for flexible structures. Chemical names, CAS numbers, inventory line items, hazard statements, and granular permissions are stored as JSON to accommodate variable-length arrays without schema migrations.
• Polymorphic contacts and addresses. A contact can belong to either a company or a specific address. This is implemented via contactable_type + contactable_id, enabling different points of contact for headquarters versus warehouse locations.
• Snapshot pattern. Batch records store a JSON snapshot of the consignee and supplier at creation time. This preserves the historical record even if the parent company is later renamed or restructured.

5. Implementation Approach

Technology Stack Summary

Runtime Constraints and Implications

Cloudflare Workers run in V8 isolates with no shared memory between requests. This architecture provides automatic horizontal scaling but imposes constraints:

• No long-lived processes. Background jobs must use scheduled Workers or external queues.
• No in-memory caches across requests. Shared state uses Cloudflare KV or D1.
• Request timeouts. Long-running operations (e.g., large batch imports) must be chunked or offloaded.

For ChemTrace Pro's workload — REST API calls, small database transactions, and B2 file uploads — these constraints are acceptable. The heaviest operation, batch approval with label generation, is bounded by the number of physical units in a single batch and completes well within worker limits.

Testing Strategy

• Backend integration tests run against the actual Hono application inside a local Miniflare sandbox using @cloudflare/vitest-pool-workers. Tests cover user creation, company CRUD, batch lifecycle, and approval state transitions.
• Frontend E2E tests use Playwright to exercise login, navigation, batch workflows, and dashboard rendering.
• Performance benchmarks measure endpoint latency inside the sandbox and report percentile statistics for architecture evaluation.

6. Evaluation and Results

Functional Validation

The backend integration test suite contains 30 tests across 7 test files covering the core API surface:

All tests pass against the local Miniflare sandbox with a seeded D1 database.

Operational Targets

The project defines clear success metrics tied to real-world operations rather than synthetic benchmarks:

7. Discussion

Key Trade-Offs

D1 SQLite vs. PostgreSQL. D1 was chosen for operational simplicity and edge distribution. For the expected scale — thousands to low millions of records — SQLite is sufficient. If the NMA later needs complex analytics, multi-region writes, or millions of concurrent connections, a migration to PostgreSQL would be warranted. The Drizzle ORM layer makes that migration feasible without rewriting business logic.

Cloudflare Workers vs. traditional servers. Workers eliminate server maintenance and scale automatically, but they impose constraints: no long-running processes, limited execution time per request, and a non-standard runtime. The team accepted these constraints because the workload maps cleanly onto the Workers model.

Static QR PWA vs. dynamic rendering. Serving QR pages as static JSON from B2 means the public-facing site scales infinitely without backend load, but it also means hazard pages are eventually consistent. If a batch is corrected after labels are generated, the static files must be regenerated and re-uploaded.

Offline-first Android vs. real-time sync. The scanner app stores data locally and syncs explicitly. This enables inspections in warehouses without connectivity but complicates conflict resolution if two inspectors modify the same tag. The current design avoids this by assigning inspections to a single inspector and a single warehouse at a time.

Known Limitations

• The consignee portal is currently minimal and needs full submission form implementation.
• SDS extraction is partially automated but still requires manual validation for unsupported PDF layouts.
• Multi-language support exists in the locale files but is not fully deployed across all public-facing content.
• The iOS version of the inspection app is not planned for the initial release.
• Real-time GPS tracking of containers is out of scope; location is updated only at scan time.

When Not to Use This System

ChemTrace Pro is purpose-built for national-scale chemical import regulation with RFID/QR traceability. It is not a general warehouse management system, a customs/border control integration, or a chemical synthesis/manufacturing platform.

8. Future Work

Planned enhancements include:

• Full consignee portal with SDS upload, hazard extraction, and approval tracking.
• Offline service worker for the QR PWA so hazard pages remain accessible after initial load without network.
• Enhanced SDS extraction with broader PDF layout support and automated H/P code validation against the GHS reference dataset.
• Inspection scheduling and analytics in the staff dashboard, including trend reporting on undeclared inventory.
• Audit logging with immutable log streams for regulatory compliance.
• Expanded RFID hardware support beyond the current Zebra TC2x target devices.

9. Conclusion

ChemTrace Pro addresses a concrete regulatory gap by treating the chemical container — not the document, not the company — as the unit of compliance. By binding each container to a unique identifier, an approval record, a Safety Data Sheet, and a hazard profile, the system creates end-to-end traceability from import declaration to public safety disclosure.

The architecture reflects the domain: an edge-distributed backend connects role-specific frontends and an offline-capable Android scanner around a single database. Shared TypeScript types, Drizzle ORM, and workspace packages keep the five applications aligned as the platform evolves.

The system is designed for extension. New regulatory workflows — retest scheduling, disposal tracking, expanded analytics — can be added as new domain routes and schema tables without reworking the core identity model.

References

1. ChemTrace Pro System Specification
2. ChemTrace Pro README
3. Introduction to ChemTrace Pro
4. Cloudflare Workers Documentation — https://developers.cloudflare.com/workers/
5. Cloudflare D1 Documentation — https://developers.cloudflare.com/d1/
6. Drizzle ORM — https://orm.drizzle.team/
7. Hono Web Framework — https://hono.dev/
8. Vue.js 3 — https://vuejs.org/
9. Nuxt 3 — https://nuxt.com/
10. Globally Harmonized System (GHS) — United Nations, https://unece.org/ghs-pictograms
11. Montreal Protocol — United Nations Environment Programme, https://ozone.unep.org/