We’re building an open-source embedded analytics tool that enables your customers to analyse their data within your app. Think ChatGPT but it’s connected to your database and embedded within your application.

In complete transparency, we’ve gone through multiple pivots to reach this stage and here’s our thesis.

There are two ways companies are solving their customer facing analytics needs today - building in-house, or embedding traditional BI tools (Metabase, PowerBI, etc) as iFrames. Why does this matter?

* Traditional BI tools aren’t built for an external-facing use case. * Restrictions on RBAC rules limit the interactive ability that these tools can provide. If you give one user edit-access, they will then have the ability to go through the entire database, including data of other customers of yours! * They have a seat-based pricing, which isn’t friendly when embedding within your app. This means that you’ll be charged for every user of yours who visits the analytics section. * iFrames stick out like a sore thumb. They don’t seamlessly embed within your application and the level of UI customisation is very limited. * Although building in-house arguably offers the highest level of customisation, the trade-off is in engineering bandwidth. How much of engineering time are you ready to spend on something that isn’t your core product offering? It could take a team of 3 engineers up to 6 months to roll out an advanced analytics offering.

Solution

A modular, lightweight analytics engine built in Python to power customer-facing analytics applications. The platform provides a unified semantic layer for defining business data models, a dynamic query engine that integrates with heterogeneous data sources, and a robust user management and authorization system—all accessible via a FastAPI-powered REST API.

Overview

This platform is designed to abstract complex data sources into a business-friendly semantic layer. It enables developers to define data models in JSON (with YAML support planned), dynamically generate queries across multiple data sources, and securely expose analytics functionality to both admin/developer users and end users.

Key Features

* Semantic Layer * Define and manage data models in JSON * Dynamic schema generation with plugin support * Versioning and audit trails (integrated with Git) * Query Engine * Translates semantic definitions into optimized queries * Supports multi-source queries (SQL, NoSQL, files, APIs) * Integration with caching for enhanced performance * Data Source Integration * Connectors for PostgreSQL, MySQL, BigQuery, Snowflake, MongoDB * Support for file-based sources (CSV, Excel, Google Sheets) * User Management & Authentication * Admin/dev authentication using FastAPI and fastapi-users * End user authentication via JWT, mapped to local user records * Authorization * Out-of-the-box RBAC with extensible hooks for dynamic, rule-based (ABAC-like) policies * Plugin interface for custom authorization logic * API-First Approach * All functionality exposed via FastAPI-based REST endpoints * Clear separation of admin and end-user API flows * Caching & Performance * Redis integration for caching query results and configuration data * Monitoring & Metrics * Event logging and metrics aggregation to track system health * Automated Configuration * Auto-detect and update semantic models based on underlying data source changes * Multi-Tenancy & Environment Isolation * Workspace and environment-level isolation to support multiple tenants

Architecture

The project follows a layered architecture within a monorepo, ensuring modularity, ease of maintenance, and independent evolution of key components.

Git hub link: https://github.com/TelescopeAI/cortex