Spiceai

Spice.ai OSS is an open source SQL query, search and large language model inference engine written in Rust, built for data driven applications and AI agents. It ships as a single portable binary of roughly 140 MB that is deployed alongside an application rather than as a separate hosted service. The runtime exposes four families of interfaces. SQL query and search APIs are available over HTTP, Arrow Flight, Arrow Flight SQL, ODBC, JDBC and ADBC, including vector search and full text search functions callable from SQL. OpenAI compatible HTTP APIs support local model serving with CUDA and Metal acceleration as well as a hosted model gateway. An Iceberg Catalog REST API provides a unified catalog interface, and MCP endpoints let agents reach external tools using the Model Context Protocol. The engine is built on Apache DataFusion, Apache Arrow, SQLite and DuckDB. Its main uses are data federation, so a single SQL query can join PostgreSQL, Snowflake, S3 and other sources without ETL pipelines; materializing and accelerating queries in memory or on disk with automatic refresh; keyword, vector and hybrid enterprise search; and grounding retrieval augmented generation and agents in real data through built in tools. It suits developers building data intensive AI applications who want one runtime in place of separate databases, caches and inference services. The project is free to use, and the site lists a price of $0 for the open source edition. The documentation covers the current 2.3 release line. No paid plans are published on the documentation site.