// the find
pathwaycom/llm-app
Ready-to-run cloud templates for RAG, AI pipelines, and enterprise search with live data. 🐳Docker-friendly.⚡Always in sync with Sharepoint, Google Drive, S3, Kafka, PostgreSQL, real-time data APIs, and more.
A collection of Docker-ready RAG application templates built on Pathway, a Python/Rust streaming framework that keeps document indexes in sync with live data sources. The real pitch is that you skip standing up separate vector DB, cache, and API layers — Pathway handles all three in-process. Aimed at teams who want a working RAG pipeline against Google Drive, SharePoint, S3, or Kafka without stitching together a half-dozen services.
The live-sync angle is genuinely useful and not just marketing: Pathway's incremental computation model means your index updates when a document changes, not on a cron. The built-in hybrid search (usearch for vectors, Tantivy for full-text) in a single runtime is a real architectural simplification over the typical Pinecone + Elasticsearch combo. The Adaptive RAG template claims 4x token reduction while maintaining accuracy, and it's based on a published technique rather than a vague prompt tweak. Docker + HTTP API as the default deployment target means you can drop any of these into an existing stack without buying into a proprietary platform.
The primary language of the repo is listed as Jupyter Notebook, which tells you a lot: these are demos you're expected to productionize yourself, not production-ready libraries. Pathway is an external dependency you're essentially betting on — it's not a widely adopted framework, and if it diverges or goes quiet, your live-sync story breaks. Error handling, observability, and retry logic in the templates are minimal or absent; the Streamlit UIs included are clearly for demos only and would need full replacement. There's no multi-tenancy support in any template — isolation between users or data sources is your problem to solve.