> ## Documentation Index
> Fetch the complete documentation index at: https://docs.models.dunsky.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Understand the platform architecture

> Learn how to maintain and develop the Dunsky Cloud Platform: understand the platform architecture. Guidance based on the core-app repository.

core-app is a two-package monorepo. Its React interface calls a FastAPI backend that coordinates application data, dashboard queries, model execution, and output access.

| Component | Responsibility |
| - | - |
| `src/platform` | React, TypeScript, Vite, Mantine, and Plotly interface |
| `src/api` | FastAPI routes, Pydantic validation, and application services |
| PostgreSQL / Lakebase | Application, tenancy, and measures schemas |
| Databricks | SQL queries and heavy or scheduled computation |
| Models API | Model execution service used by the application |
| Azure Blob Storage | Model input and output files |
| Redis | Caching and distributed locks |
| nginx | Built frontend serving and API proxying |

## Code map

Backend code lives under `src/api/src/core`: `routes`, `services`, `clients`, `db`, `auth`, `runners`, and `formula_engine`. Frontend code under `src/platform/src` includes `pages`, `components`, `clients`, `contexts`, `hooks`, and `config`.

## Execution boundaries

Request-scoped work belongs in FastAPI routes and their services. Heavy or scheduled modelling computation belongs in Databricks. Redis is not a task queue, and there is no Celery worker to start.

FastAPI starts a reconciliation loop through its lifespan handler. Application resources mount under `/rest/v1`; `/health` is a separate health endpoint.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.