Originally published on andrew.ooo — visit the original for any updates, code snippets that aged out, or follow-up posts.

TL;DR

CocoIndex is an open-source Python framework (with a Rust core) that solves the most underrated problem in production AI: your agent's RAG index goes stale the moment the data changes . Instead of rebuilding the whole vector store every hour, CocoIndex tracks per-row provenance and only reprocesses the delta when a source file, a chunking function, or an embedding model changes. It's trending hard on GitHub right now — +1,798 stars this week , ~9,700 total — and the framework has been pitched as "React for data engineering" because you declare the target state and the engine keeps it in sync forever.

Key facts:

Incremental by design — change one file in a 10,000-document corpus and only that file's chunks re-embed; the other 99.9% stay cached

Rust core + Python API — production-grade ingestion under the hood, but you write your pipeline in 20 lines of Python

Connectors — local…