TECHNICAL WHITE PAPER · EDITION 01 · 1 OCTOBER 2026

Source-evidenced planning.
Explicit limits.

How Agdron Mixer connects Australian aerial-use evidence to mixing calculations and durable job records—and what it does not claim to do.

This project-authored paper describes the current implementation and recorded test evidence. It is not an independent certification, regulatory approval or substitute for the current product label.

01. Purpose and scope

Agdron Mixer supports planning for source-evidenced Australian agricultural aerial uses. Its goal is to keep a selected product/use, the calculation inputs and the evidence behind them connected. It does not establish an operator’s licence, aircraft approval, site suitability or legal permission to spray.

The packaged dataset snapshot of 29 September 2026 contains 875 app-ready products, 15,379 released paths and 117,020 exact approval records. Approval records are mapping occurrences, not distinct products, spraying events or universal permissions. A product’s inclusion does not mean every crop, pest, state, aircraft or application method is available. Registration checks describe the evidence at review time; the offline package is not a live registration service.

02. Evidence and release model

The source database is protected from modification. Product-specific evidence and validated release manifests are used to build a separate verified database. Source identity and hashes are retained, and the packaged database is checked against the derived database.

  1. Establish identity: distinguish a product registration number from label approval and PDF identifiers.
  2. Bind the source: preserve the official label, approval identity, hash and page/section or table-cell evidence.
  3. Describe the use: retain product-specific crop/situation, target, rate, state, method, carrier and component semantics.
  4. Hold ambiguity: do not promote unresolved values, conflicting directions or unsupported mappings.
  5. Verify the package: test code and data gates and independently check release integrity.

Deterministic parser families may be reused only where supported structure and semantics match. Shared active ingredients, manufacturers or visually similar labels are not proof of interchangeable directions. Ground information and permit-derived uses are not automatically promoted into aerial calculations.

03. Selection and calculations

The selector accepts product, crop and pest searches in different orders and narrows compatible choices. Aerial-rate authority is checked for the exact selected use and aircraft context. Changing relevant selections invalidates stale label-derived rates.

Calculations retain rate units and basis, operator inputs, carrier medium and mandatory components. Per-hectare amounts, concentration-based rates, water-plus-oil and other supported carrier branches must not be treated as equivalent. Required components without a supported numeric rule remain blocked rather than being estimated.

Job and mixing areas are distinct inputs. When positive areas disagree, reports disclose the mismatch and remain draft; the software does not silently overwrite either input. Loading a saved job or applying a template requires fresh executable label-rate selection.

04. Warning coverage is incomplete

Absence is not clearance

For complete warnings and restrictions, consult the current product label. An empty warning field is not evidence that a restriction is absent.

There are three different levels of information: raw source text, referenced evidence, and structured fields the app can display conditionally. Preserving text does not establish complete structured coverage. The remaining warning-backfill programme is deferred; no pending work is represented as certified.

The preserved checkpoint has 810 pending and 14 held warning reviews. Fifty-eight products were fully certified across their released paths in that warning-review checkpoint; this does not imply complete warning coverage for the rest of the 875-product package. Droplet, wind, inversion, temperature, drift, buffer, re-entry, withholding, aircraft and PPE requirements must still be checked on the applicable current label. Not every label specifies every category.

Missing or legacy warning data produces a label-referral notice. Known source, mapping, carrier and calculation holds remain in force. The app does not make a comprehensive or automatic compliance assessment.

05. Records, reports and privacy

Jobs and templates are saved in local browser storage. Job JSON export/import supports transfer and backup; invalid or unsupported records are rejected. Local storage is not account-based synchronisation. Clearing browser/site data or losing a device can remove local records, so export important jobs.

Printable Mix Sheets and Client Field Reports retain source references, cautions and readiness notices. Quotes and invoices support multi-line amounts, numbering, persisted status and printable output. Reports are jurisdiction-neutral, not statutory-compliance certificates. Maps, signatures, cloud accounts, subscriptions and automated weather services are not claimed as current delivered features.

The website does not need to download the label database just to explain the product. Opening the app starts the separate data-loading workflow. Hosting providers receive normal web requests; following external label links contacts the source site. “Local records” should not be read as a claim of anonymity or a security certification.

06. Offline and device behaviour

The app is a Progressive Web App. On supporting browsers it can be added to the home screen or installed without an APK. HTTPS hosting, successful loading and adequate storage are prerequisites for the intended offline experience.

The database is currently 691,609,600 bytes (about 692 MB). Its browser storage and in-memory footprint may be larger. Use a reliable connection for initial loading, wait for caching to complete and test offline restart before field use. External APVMA label links still require connectivity.

The offline database cache is versioned by content hash. App updates use an explicit prompt rather than silently replacing a running session. Browser storage can still be evicted by the device, and private/incognito offline caching is not supported by the recorded acceptance evidence.

07. Verification and remaining acceptance gates

The 29 September 2026 sealed app baseline recorded 5,456 backend tests and 460 frontend tests, with type/build checks and independent release verification. These counts describe that baseline, not an assurance that every future deployment is verified.

AreaRecorded evidence / boundary
Search and persistenceSelector orders, saved-job reload, template reuse and fresh-authority checks exercised in browser tests.
ReportsRepresentative PDF pagination and content-preservation checks. Actual printers and every device remain separate acceptance work.
Responsive layoutsPhone/tablet/desktop viewport checks in desktop Edge emulation; not physical-device certification.
Offline restartIsolated persistent Edge profile restarted offline with the test server stopped; retained records and fresh database lookup checked.
Website deploymentThe root website and /app/ package require hosting and deployment-specific checks. A local build is not proof of a live site.

Physical Android/iOS/tablet installation, storage/memory behaviour and field acceptance remain open. The user must assess current label instructions, registration/permissions, weather, site conditions, operator requirements and applicable jurisdiction rules. This document does not replace professional advice or those checks.

Source references

Official external starting points: APVMA PubCRIS and the APVMA website. Product-specific source references are surfaced in the app where available. Agdron is not affiliated with or endorsed by APVMA.

This edition is based on the project’s sealed release receipts and delivery reconciliation dated 29 September 2026. It will need revision when data coverage, verification or supported features change.

Open Agdron Mixer ↗Back to the website

To retain this paper, use your browser’s Print → Save as PDF option.