EDDA - Intelligent Data Dictionary Agent (Elephantidae)

What it is:
EDDA is a software-only agent that connects to enterprise databases (PostgreSQL / SQL Server / Snowflake), extracts complete schema metadata, profiles data quality on the fly, and generates business-friendly data dictionary documentation.

Why we built it:
In real teams, database documentation is either missing or outdated. Analysts waste time guessing what columns mean, joining tables incorrectly, or trusting fields that are incomplete/stale. EDDA makes schema + data health visible and explains it in plain language.

How it works (high level):

1. Connect DB using read-only credentials

2. Extract metadata: tables, columns, PK/FK, constraints, relationships

3. Sample rows per table (fast, configurable)

4. Compute metrics: completeness (null%), freshness (last timestamp), key health (uniqueness), FK orphan rate + quick stats (min/max, top values, p50/p95)

5. Generate documentation (Markdown + JSON) grounded in the extracted metadata and measured signals

6. Provide a chat interface for questions like “How do I join orders to payments?” with optional safe SELECT SQL suggestion