Skip to content

Tech Stack


Last updated: November 2026


This document outlines the current technology choices for Crestr. The stack prioritises simplicity and performance for geospatial operations, and maintainability while working within the constraints of a solo-developed project.

Overview

Layer Technology Notes
Frontend HTML, CSS, TypeScript, OpenLayers Client-side mapping and UI
Backend Python and FastAPI Web framework layer
Database PostgreSQL + SQLModel ORM Relational storage with ORM
Pathfinding BRouter Expert routing API hosted on a docker container
Map Data OpenStreetMap (OSM) for nodes, OpenTopoMap for tiles, LocationIQ for reverse geocoding Source of geographic data
Authentication Sessions PostgreSQL table + Cookies User session management

Detailed Breakdown

Frontend

  • HTML, CSS, TypeScript: Core web technologies. No heavy frameworks at present to keep bundle size small and performance high
  • OpenLayers: Chosen for interactive mapping capabilities, vector layer support, and strong OSM integration. Provides the foundation for route display and user interaction.

Backend

  • Python: Primary language due to strong ecosystem support for geospatial libraries and rapid development.
  • FastAPI: Current main web framework after migrating from Flask, which offers better performance, automatic OpenAPI documentation, and modern async support.

Database

  • PostgreSQL: Robust, open-source relational database with excellent GIS extensions (PostGIS planned for future spatial queries).
  • SQLModel: Provides an abstraction layer, reducing syntax and boilerplate, and supports migrations via Alembic.

Core Algorithms

  • BRouter: Expert open-source routing algorithm. This replaced my custom algorithm to allow routing across the UK and Ireland rather than just Cumbria. It handles elevation and geographic map tiles via .rd5 allowing me to easily provide international routing for Crestr should I upgrade the current server.

Authentication & Security

  • Session pSQL table + Cookies: Upon Login a cookie is sent containing a cryptographic session ID, this is compared to the session ID in the pSQL table upon each API call which requires authentication, this method keeps users logged in for 7 days (can be easily edited)

Design Principles

  • Keep the stack minimal to reduce operational overhead.
  • Favour open-source and community-supported tools.
  • Ensure the architecture supports incremental improvements without major rewrites.

Rationale

Choices were driven by the need for strong geospatial capabilities, and long-term maintainability. The current stack allows rapid iteration while avoiding unnecessary complexity.

The decision to replace my own algorithm with a BRouter docker container shows this the clearest, as I put aside my own work in the favour of a better product.