Project / Infrastructure & Microservices

Radar de Voos NorthCore

Autonomous airfare intelligence and continuous airline network monitoring with direct Telegram checkout.

Snapshot

Current record

Role
Microservice architecture, statistical automation and infrastructure
Stage
In Production (24/7); hosted on Oracle Cloud Infrastructure
Stack
TypeScript, Node.js, Sqlite, Docker
Status
Active

Record

A dated view of the project.

  1. Started
  2. Published
  3. Updated

Current state

Radar de Voos NorthCore is an autonomous microservice operating continuously 24/7, built to eliminate the drag of manual flight shopping and reliably pinpoint statistical airfare deals.

The service runs containerized on an Oracle Cloud Compute VM, continuously indexing target flight corridors across full 365-day calendar horizons and streaming actionable alerts to a dedicated Telegram interface.

The problem

Manual flight search burns valuable hours, suffers from opaque fare fluctuation, and forces repetitive interactions with sluggish, ad-bloated airline websites.

Furthermore, dynamic pricing algorithms make it virtually impossible for human shoppers to catch fleeting fare drops along an entire annual window, often falling for short-term fare noise.

The solution

A deterministic, autonomous scanning pipeline that polls routes three times daily (08:30, 13:00, and 19:30 BRT), accounts for temporal fare dispersion to filter noise, and delivers the year's Top 3 best bargains directly to Telegram with zero-friction direct booking links.

Engineering and architecture

Search engine and structured ingestion

  • Direct integration with optimized RapidAPI endpoints, filtering Brazilian carriers (Azul, Latam, Gol), maximum travel times, and connection counts per leg.
  • Normalization of complex flight payloads into strictly typed TypeScript domain models.

Historical ledger and anomaly detection

  • Local persistence using structured SQLite via persistent Docker volumes, ensuring lightning-fast reads and writes with zero third-party database overhead.
  • Statistical engine computing 30-day rolling averages per route/IATA code.
  • Automatic deal dispatch with "HISTORIC AIRFARE DETECTED" badges whenever a quoted price drops ≥ 15% below rolling averages or breaks the route's all-time historical low.

Conversational UI (Telegram UX)

  • Persistent custom keyboard: pinned bottom buttons for 🔍 Search Best Fares and 📊 Route Statistics.
  • In-flight live feedback: seamless progress indicators updating the temporary message (editMessageText) without spamming user conversation history.
  • Deep-linked booking: inline callback buttons routing directly into airline and official ticketing flows, completely bypassing repetitive web form entries.

Telemetry and observability (/stats)

  • Internal command reporting cumulative scan volumes, all-time historic record lows with exact round-trip departure dates, and moving average trends per airport pair.

DevOps, resilience and backups

  • Multi-stage container: lightweight and secure Node.js 20 build over Alpine Linux, orchestrated with Docker Compose.
  • Production VM: Oracle Cloud Infrastructure Compute instance monitored by systemd and Docker daemons.
  • Atomic backup automation: dedicated routine (backup-db.sh) capturing daily SQLite snapshots under a 7-day cyclical retention policy.
  • Zero no-code overhead: entirely first-party code without reliance on fragile third-party automation tools or costly LLM wrappers.

Keep exploring

Follow the work behind the project.

See the other projects, read the engineering notes, or get in touch about a problem worth building around.