v0.7.0 · Apache 2.0 · New: Pipeline Manifests + Hermes MCP

Business docs in.
Root causes out.

Point Aegis at your business docs — policies, schema definitions, SLAs, anything that describes your data. It generates executable rules, runs them across any warehouse, and diagnoses every failure with LLM-powered root cause analysis — automatically.

Get started → Hermes integration GitHub

pip install aegis-dq

How it works

From business docs to diagnosed failures in minutes.

Point Aegis at your docs and warehouse

Give it your business docs — policies, schema definitions, SLAs, anything that describes your data. Aegis reads the context and generates data quality rules automatically. No hand-writing rules.

aegis generate orders --kb policy.md --kb schema.md

↓

Run checks across any warehouse

Aegis executes the rules against your warehouse. Every failure is classified by severity and diagnosed by an LLM — what broke, why it broke, and what regulation or SLA it violates.

DuckDBBigQueryAthena DatabricksPostgres

↓

Get root cause and a fix for every failure

Every failing rule comes back with a plain-English diagnosis, root cause, severity tier, and remediation SQL — ready to act on. Every LLM decision is logged with cost and latency.

Root cause Severity Remediation SQL Full audit trail

✦ Prefer plain English?

Connect Hermes and just ask.

Hermes connects to Aegis via MCP. Load a pipeline manifest and ask in plain English — Hermes runs the checks, diagnoses failures, remembers past runs, and can schedule recurring validations.

Set up Hermes →

Load the pipeline at demo/fraud/pipeline.yaml and run it.

✓ Loaded fraud-aml · 6 tables · 55 rules

✓ Validated against /tmp/fraud.duckdb

⚠ 11 violations — 5 CRITICAL (CTR, OFAC, PEP)

→ Remediation SQL ready · Total cost: $0.01

Three ways to use Aegis

Generate

Rules from business docs

Point Aegis at your business docs and schema. It introspects the database, reads the context, and writes not_null, accepted_values, and complex custom_sql rules — including CTEs, window functions, and multi-table JOINs.

aegis generate transactions --kb policy.md --kb schema.md

Run

Pipeline manifests

A pipeline.yaml captures your database, rules, docs, LLM config, and goal in one file. Define it once. Run it from the CLI, Airflow, GitHub Actions, or any MCP client — same config everywhere.

aegis pipeline run my-project/pipeline.yaml

Talk

Conversational via Hermes

Connect Hermes via MCP. Ask it to load your pipeline, run validation, and diagnose failures in plain English. Hermes remembers past runs, can schedule recurring checks, and delivers results wherever you work.

Load the pipeline at demo/fraud/pipeline.yaml and run it.