Skip to main content

​
Ibex Analytics Service

The ibex-analytics-service is the high-performance query engine that powers all analytics and reporting across the Ibex platform. It leverages an embedded DuckDB instance to perform federated queries across multiple heterogeneous data sources in a single SQL statement.

​
Architectural Role

This Python/FastAPI service operates internally on port 8000 within the Docker ajna network. It is not directly exposed to the internet; all requests originate from either the ibex-bi-backend (for user-generated reports) or the ibex-ai-service/ibex-agent-engine (for AI-generated SQL execution).

​
High-Level Working

​
Federated Catalog Integration

Upon startup, the service automatically contacts the ibex-data-platform to fetch the registry of active data sources. It then dynamically attaches these sources as virtual DuckDB catalogs:
  • MySQL/PostgreSQL: Attached natively via DuckDB scanner extensions.
  • Iceberg: Connected via the Iceberg REST catalog, querying underlying Parquet files physically stored in MinIO.
  • S3 (Raw): Scans adhoc uploaded CSV and Parquet files.

​
Cross-Source Queries

Because of the catalog architecture, users and AI agents can execute fully-qualified hybrid SQL queries: 
SELECT orders.total, users.email 
FROM mysql_business.orders AS orders 
JOIN postgres_metadata.public.users AS users 
ON orders.user_id = users.id

​
Protection & Security

It relies entirely on network isolation and an internal API key (ANALYTICS_SERVICE_API_KEY) for protection, trusting the upstream ibex-bi-backend to handle user-level row security and authentication.