AltimateAI/altimate-code

What it solves

General AI coding agents often struggle with data engineering because they lack a deep understanding of the specific data stack, leading to hallucinated SQL, incorrect schema assumptions, and missed PII. Altimate-code provides a deterministic intelligence layer that gives LLMs the precise tools needed for SQL analysis, lineage tracking, and warehouse management without relying on pattern matching.

How it works

It acts as a specialized harness that can be run standalone in a terminal or embedded under other AI agents (like Claude Code or Codex). It integrates directly with data warehouses and dbt projects to provide deterministic tools for parsing SQL, indexing schemas, and analyzing costs. It uses a model-agnostic approach, allowing users to connect their own LLM provider or run models locally via Ollama.

Who it’s for

Data engineers and analysts who need AI assistance for building SQL pipelines, managing dbt projects, and performing FinOps analysis while maintaining high precision and security.

Highlights

• Deterministic SQL Intelligence: Includes 19 rules for anti-pattern detection with 100% benchmark accuracy.
• Column-Level Lineage: Automatically extracts lineage from SQL across multiple dialects.
• dbt Native Integration: Supports manifest parsing, test generation, model scaffolding, and unit test generation for dbt 1.8+.
• Cross-Dialect Capabilities: Translates SQL between major warehouses (Snowflake, BigQuery, Databricks, etc.) and performs row-by-row data parity checks.
• FinOps & Safety: Detects expensive queries, suggests warehouse right-sizing, and scans for PII using 30+ regex patterns.
• Agent Modes: Offers scoped permissions via Builder, Analyst (read-only), and Plan modes to control SQL write access.