PRODUCTION / WORKFLOWS
Olive — Deliverable Tracking Platform
A full-stack deliverable-tracking and workflow-orchestration platform built for production at CognoVerse — event-driven workers, dual-cloud storage, and dual auth models.
Overview
Olive is a production deliverable-tracking and workflow-orchestration platform architected and solo-built at CognoVerse. It coordinates cloud-folder provisioning, email notifications, automated deadline escalation, and file-submission detection through an event-driven pipeline of four decoupled workers — with a dual-cloud storage abstraction (Google Drive and Microsoft OneDrive) behind a single interface, tokenless magic-link workflows for external stakeholders alongside JWT role/department access for internal users, a hand-built zoomable Gantt chart, and an immutable audit log covering 30+ action types.
Problem
Deliverable tracking across external stakeholders and internal teams needs reliable orchestration — folder provisioning, notifications, deadline escalation, and file detection — without coupling every workflow to a single request path or a single cloud vendor.
Result
Shipped from first commit to production in under 2 weeks; still owned as the sole engineer in production, with an event-driven pipeline across 4 decoupled workers.
Architecture
Key engineering decisions
Event-driven worker pipeline
Four decoupled workers handle cloud-folder provisioning, email notifications, automated deadline escalation, and file-submission detection so each concern can fail and retry independently.
Dual-cloud storage abstraction
Live Google Drive and Microsoft OneDrive (Graph API) backends sit behind one interface so the product is not locked to a single storage vendor.
Two coexisting auth models
Tokenless magic-link workflows for external stakeholders alongside JWT-based role/department access for internal users.
Immutable audit trail
An audit-log system tracks 30+ distinct action types so operational history stays reconstructable.
Failure modes
Coupled request-path workflows
Redis Streams and decoupled workers keep long-running provisioning and escalation off the critical API path.
Single-vendor storage lock-in
The dual-cloud abstraction lets Drive and OneDrive backends coexist behind one interface.
External vs internal access mismatch
Magic links for stakeholders and JWT role/department auth for internals avoid forcing one auth model on both audiences.
Implementation
- Fastify on Bun with TypeScript for the core API
- PostgreSQL for durable state and audit history
- Redis Streams powering the event-driven worker pipeline
- Four workers: folder provisioning, email, deadline escalation, file-submission detection
- Google Drive + Microsoft OneDrive (Graph API) behind a unified storage interface
- Magic-link auth for external stakeholders; JWT role/department auth for internals
- Hand-built zoomable Gantt-chart visualization in React
- Immutable audit log covering 30+ action types
Results & metrics
< 2 weeks
First commit → production
Sole engineer
Ownership
30+
Audit action types
Lessons
- —Production workflow platforms need event boundaries early — provisioning and escalation should not live on the request path.
- —External stakeholders and internal users often need different auth models, not a forced compromise.
- —Auditability is a product feature when deliverables and deadlines have operational consequences.