FIND-20260402-015 · 2026-04-02 · Innovation Veille

Trending: pathwaycom/pathway — Python streaming ETL with Rust engine for Zero-ETL + LLM RAG (63k stars)

trending-repo MEDIUM
Pathway is a Python ETL framework for stream processing and real-time analytics, powered by a Rust differential dataflow engine. Handles 60k messages/second with sub-second latency. The same code runs in batch and streaming modes, bridges real-time Kafka/Redpanda streams with LLM pipelines and vector RAG. Provides Kafka-ETL examples directly compatible with Redpanda. 63k GitHub stars, 31 open issues, last commit 2026-04-01. License is BSL 1.1 (source-available, not OSI-approved — commercial use requires review after 4-year delay window).

Source

https://github.com/pathwaycom/pathway

ODS Impact

ODS Zero-ETL architecture uses Redpanda -> Debezium -> ClickHouse. Pathway could complement or replace Debezium for Python-native stream transformation use cases. The LLM/RAG integration is relevant for AI-enhanced document processing in DocStore and PDF Engine. BSL license requires legal review before production use.

Security Review

License: BSL-1.1 (non-OSI, source-available) | Maintenance: ACTIVE | Risk: MEDIUM | Recommendation: USE_WITH_CAUTION

Tags

etl streaming rust python zero-etl rag llm kafka redpanda