0.7.0

Aegis DQ

The open-source agentic data quality framework. Validate data contracts, diagnose failures with LLM root-cause analysis, and auto-generate SQL remediation — all in a single CI step or Python call.

• 31 rule types — completeness, uniqueness, validity, referential integrity, statistical, ML anomaly detection
• 6 warehouse adapters — DuckDB, Postgres/Redshift, BigQuery, Databricks, AWS Athena, Snowflake
• Pluggable LLMs — Anthropic Claude, OpenAI, Ollama (local), AWS Bedrock
• Agentic pipeline — plan → parallel validation → LLM diagnose → RCA → SQL remediate → report

---

GitHub Actions — Quick Start

Add a data quality gate to any workflow in under 2 minutes:

# .github/workflows/data-quality.yml
name: Data Quality

on: [push, pull_request]

jobs:
  data-quality:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Validate data quality
        uses: aegis-dq/aegis-dq@v0.7.0
        with:
          rules-file: rules.yaml
          db: data/warehouse.duckdb
          anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}

The step fails the job automatically when any rules fail, blocking broken data from reaching production. Set fail-on-failure: 'false' to report without blocking.

Offline mode (no API key required):

      - name: Validate data quality (offline)
        uses: aegis-dq/aegis-dq@v0.7.0
        with:
          rules-file: rules.yaml
          db: data/warehouse.duckdb
          no-llm: 'true'

Action inputs

Action outputs

Using outputs in downstream steps:

      - name: Validate data quality
        id: dq
        uses: aegis-dq/aegis-dq@v0.7.0
        with:
          rules-file: rules.yaml

      - name: Post summary
        run: echo "Pass rate: ${{ steps.dq.outputs.pass-rate }}%"

---

Demo

╭──────────────────────────────────────────────────────╮
│ Aegis DQ  —  RetailCo E-commerce Demo                │
│ LLM: amazon.nova-pro-v1:0 via AWS Bedrock            │
╰──────────────────────────────────────────────────────╯

✓ Pipeline complete in 7.1s · 12 rules · $0.0056 LLM cost

╭──────────────── Validation Summary ─────────────────╮
│  Rules checked  │  12                               │
│  Passed         │  1   │  Failed  │  11             │
│  Pass rate      │  8%  │  Cost    │  $0.005576      │
╰─────────────────────────────────────────────────────╯

LLM Diagnoses
  orders_customer_fk  →  Order placed with customer_id=99 that does not exist.
                         Likely cause: customer deleted or test record not cleaned up.

  products_sku_unique →  Duplicate SKU-001 — two products share the same identifier.
                         Likely cause: duplicate import from supplier feed.

Remediation SQL (LLM-generated)
  orders_status_valid          UPDATE orders SET status = 'SHIPPED' WHERE status = 'DISPATCHED';
  products_price_positive      UPDATE products SET price = ABS(price) WHERE price < 0;
  products_stock_non_negative  UPDATE products SET stock_quantity = 0 WHERE stock_quantity < 0;

---

Why Aegis?

---

Install

pip install aegis-dq

---

5-minute quickstart

Seed a demo DuckDB database:

import duckdb

con = duckdb.connect("demo.db")
con.execute("""
    CREATE TABLE orders AS
    SELECT i AS order_id, 'placed' AS status, i * 9.99 AS revenue
    FROM range(1, 10001) t(i)
""")
# introduce some bad data
con.execute("UPDATE orders SET order_id = NULL WHERE order_id % 200 = 0")
con.execute("UPDATE orders SET revenue = -5.00 WHERE order_id % 500 = 0")
con.close()

Generate a starter rules file and run:

aegis init

export ANTHROPIC_API_KEY=sk-ant-...
aegis run rules.yaml --db demo.db

Run without an API key (validation only, no LLM diagnosis):

aegis run rules.yaml --db demo.db --no-llm

---

Pipeline

Every aegis run passes your data through a LangGraph pipeline:

rules (Python / YAML)
    │
    ▼
  plan ──► parallel_table ──► reconcile ──► remediate ──► report
                 │
         ┌──────────────────┐
         │  per table:      │
         │  execute         │
         │  classify        │
         │  diagnose        │  ← concurrent across all tables
         │  rca             │
         └──────────────────┘

• plan — parse and validate rules, build an execution graph
• parallel_table — concurrently fans out per table: execute all rules, classify failures by severity, diagnose with LLM, and trace root causes
• reconcile — compare results against expected thresholds
• remediate — LLM proposes a targeted SQL fix for each diagnosed failure
• report — structured JSON + optional Slack notification

---

Rule types (31 total)

Example rule:

rules:
  - apiVersion: aegis.dev/v1
    kind: DataQualityRule
    metadata:
      id: orders_revenue_non_negative
      severity: critical
      owner: revenue-team
      tags: [revenue, validity]
    scope:
      warehouse: duckdb
      table: orders
    logic:
      type: sql_expression
      expression: "revenue >= 0"

---

Generate rules with the LLM

Instead of writing rules by hand, let Aegis introspect your table schema and generate a draft rules file:

# Schema-aware structural rules (not_null, between, unique, accepted_values...)
aegis generate orders --db warehouse.duckdb --output orders_rules.yaml

Add a --kb document — any plain text or markdown file describing your business logic — and the LLM generates business validation rules alongside structural ones:

aegis generate orders \
  --db warehouse.duckdb \
  --kb docs/orders_policy.md \
  --output orders_rules.yaml

What goes in a KB file? Anything your team knows about the data:

# orders_policy.md
- status must be one of: placed, confirmed, shipped, delivered, cancelled
- amount must be greater than 0; refunds are handled in a separate table
- customer_id must reference a valid customer (no test accounts: id > 1000)
- order_date must not be in the future
- discount_pct must be between 0 and 0.5 (max 50% discount)

The LLM turns these into accepted_values, sql_expression, between, and foreign_key rules automatically. Generated rules are stamped status: draft — review, promote to active, and commit.

All aegis generate options:

---

Warehouse adapters

---

LLM providers

Switch providers at the CLI:

aegis run rules.yaml --llm openai --llm-model gpt-4o
aegis run rules.yaml --llm ollama --llm-model llama3.2
aegis run rules.yaml --llm bedrock --llm-model amazon.nova-pro-v1:0

---

Integrations

---

CLI reference

aegis run flags:

---

Roadmap

Full issue tracker: github.com/aegis-dq/aegis-dq/issues

---

Contributing

Contributions are welcome. See CONTRIBUTING.md to get started.

Good first issues: label:good first issue

License

Apache 2.0