Discover the distributed system architecture powering TekTrakker's high-scale field service management platform.
How field updates travel from offline client applications to our database and accounting ledgers.
Technician performs an action (e.g. completes checklist). App captures local timestamp, generates a client UUID, and appends the change to the local SQLite queue.
Upon network reconnection, API Gateway authenticates, validates payload schemas, and prevents duplicate submissions via UUID deduplication.
Payloads route to a high-throughput Google Cloud Pub/Sub queue, partitioned by tech and job ID. Decouples client traffic spikes from database writes.
Events write to an append-only transaction database. Stream processors project changes onto read-views, enforcing business state rules (e.g., cancelled jobs reject completion updates).
Payment events write to an independent financial double-entry database, which reconciles Stripe webhooks asynchronously and exports records to QuickBooks.
Powered by a local SQLite cache on client devices. Technicians can view schedules, create invoices, and log time without cell coverage. Payloads queue locally and sync automatically.
Uses Capacitor SQL local-store with store-and-forward transaction tracking.
Instead of full-document overwrites, the system runs CRDT-inspired field-level merging. Parallel edits to different fields on a single job merge without conflict.
Collisions on identical fields default to Timestamp Last-Write-Wins (LWW) and flag a dispatcher review.
Every status change, invoice edit, or GPS ping is logged as an immutable, append-only event stream. The server projects these events to build the live read-views.
Ensures complete history replayability and eliminates silent semantic data drift.
Financial data (invoices, Stripe/Square payments) is decoupled from operational states. Operates as an independent double-entry ledger domain.
Reconciles Stripe transactions instantly via webhook listeners, protecting accounting from network failures.
Direct, native bi-directional integration with Intuit's REST API without third-party middleware or Zapier bridges.
Uses Google Cloud Tasks queues to manage API rate limits with exponential backoff retries.
Hosted on multi-region serverless Google Cloud architecture. Backed by a 99.99% uptime history, secure backup schedules, and 100% US-based support.
SDVOSB-certified with response SLAs under 15 minutes and direct escalation paths to DevOps.
How we resolve semantic conflicts, define data ownership, and isolate system failures at enterprise scale.
To prevent "ghost states" or operational disagreements between office teams and field crews, TekTrakker clearly partitions system-of-record authority across three logical planes:
Payments and invoice amounts are locked inside a strict transactional double-entry ledger. Operational dispatch changes can never mutate completed billing states.
The append-only log stores every single mutation with millisecond timestamps, ensuring a perfect, replayable audit trail of who did what, when.
Derived eventual-consistency states representing active dispatch boards. Projections update asynchronously from the event streams.
Core business events (payments, completions, and cancellations) enforce strict finality. Once closed, the server locks state mutations, keeping offline client synchronization from silently overwriting canonical states.
Instead of applying silent merges on conflicting field assignments, TekTrakker routes semantic disputes to a dedicated Resolution Dashboard. Dispatchers can review contradictory entries side-by-side and execute authoritative overrides.
In case of connectivity drops or API gateway saturation, the platform runs in read-only backup mode. Dispatchers can inspect active fleet maps and technician calendars while write operations shape traffic in Pub/Sub streams to preserve platform responsiveness.
Direct answers to technical queries regarding our distributed design patterns and operations.
A: We enforce **client-side UUID generation** on all transaction and invoice mutations. When the client syncs, the server-side processor maps this UUID as the QuickBooks `DocNumber` or Stripe request key. Before any write operation, our sync worker checks for this UUID. If the transaction already exists, the server confirms sync success without duplicating the record.
A: The server-side projector enforces strict state transitions. Since a cancellation is marked as a *terminal state*, any incoming offline `COMPLETED` event will not overwrite the status. Instead, the sync engine appends the technician's completed checklists and photos to the immutable audit log and flags the ticket in the dispatcher's queue as a *Post-Cancellation Work Claim* for manual review, preventing silent data overwrite.
A: TekTrakker's integration with QuickBooks Online is **100% native and direct**, communicating straight with Intuit's REST API. We do not use Zapier, Make, or other middleware, avoiding additional subscription costs and sync delays. Outbound writes queue dynamically through Google Cloud Tasks with exponential backoff retry algorithms to handle Intuit rate limits safely.
A: The platform runs on fully managed, serverless Google Cloud structures (Google Cloud Run and Firestore) distributed across multiple geographic regions. If a region experiences issues, traffic routes automatically to an active region. Data stores replicate synchronously, and the client applications' local SQLite databases act as a failsafe so field operations continue uninterrupted.
Our Enterprise tier includes complete open API access, webhooks, and dedicated developer support.
Contact Systems Engineering