Puchaczov/Musoq

What it solves

Musoq provides a way to perform ad-hoc SQL queries against non-database data sources—such as log files, git history, binary dumps, and system processes—without requiring data ingestion or intermediate storage. It replaces fragile chains of Bash commands or custom scripts with declarative SQL queries for quick data investigation and debugging.

How it works

Musoq is a modular SQL engine that streams and transforms data directly from its source. It extends standard SQL with specialized primitives for handling unstructured data:

• Inline Binary Decoding: Users can declare binary structures (structs) directly in the query to parse raw bytes from files.
• Declarative Text Parsing: Users can describe log formats inline to extract structured fields from text files.
• Strong Typing: The couple statement allows binding untyped data (e.g., from LLM extractions) to strict SQL tables.
• Data Sources: It uses a plugin architecture to connect to various sources like the OS filesystem, Git repositories, Docker, Kubernetes, and LLM APIs (OpenAI/Ollama).

Who it’s for

Developers and system administrators who need to perform complex investigations on local files or system state without the overhead of setting up a database or writing one-off scripts.

Highlights

• Zero Ingestion: Queries data where it resides without importing it into a database.
• Advanced SQL Support: Includes recursive CTEs and correlated subqueries.
• ** uma stdin Piping**: Native support for querying streamed stdin data using SQL.
• AI Agent Integration: Built-in Model Context Protocol (MCP) server to expose saved queries as tools for AI assistants like Claude or Cursor.
• Flexible Parsing: Ability to define binary and text schemas inline within the SQL script.