# Spice

A portable data and AI engine, written in Rust, that adds a real-time analytics node to an operational database and serves SQL query, search and LLM inference to applications and agents from one binary.

Spice (Spice.ai Open Source) is a data and AI engine that brings a user's existing data to their application. It connects to databases, data warehouses, data lakes and APIs already in use and serves that data to applications and AI agents through standard APIs, with SQL query, search and LLM inference built in; accelerated datasets return in milliseconds because Spice keeps a local, continuously refreshed copy, while other datasets are queried in place with query push-down.

Datasets from more than 30 sources, including PostgreSQL, MySQL, Snowflake, Databricks, S3 and Iceberg, are declared in one YAML file; selected datasets are materialized in memory or on local disk and kept fresh through scheduled refreshes or real-time change-data capture. Applications then query over SQL through HTTP, Arrow Flight SQL, JDBC, ODBC or ADBC, search with vector, full-text or hybrid search, and call LLMs through OpenAI-compatible APIs and MCP, all over the same data. Version 2.0 adds a real-time analytics replica of an operational PostgreSQL, MySQL or MongoDB database through native change-data capture, without a separate ETL pipeline.

It is a single Rust binary, built on Apache DataFusion, Apache Arrow and Vortex, and this repository's runtime and CLI are released under the Apache 2.0 license.

Website: https://docs.spiceai.org