Bell: Building Modular and Scalable RAG Systems with Hybrid Batch/Incremental Processing
Our short summary of Zenml’s engineering write-up. The full story, with all the detail, is on their blog.
Read the full write-up on zenml.io (opens in a new tab)Problem
Documentation changed constantly, making it hard to keep the RAG system fast and up‑to‑date.
Solution
They built a modular hybrid RAG pipeline that runs batch jobs for static docs and incremental jobs for new docs, using Apache Beam, Cloud Composer and GCP services.
Build details 2/5
What the original post describes.
- Architecturedescribed.
Shows the system's parts and how data moves between them.
- Stackdescribed.
Names the models, libraries, databases or services they used.
- Evaluationnot described.
Explains how they measured whether it works.
- Failure modesnot described.
Covers what went wrong, the limits, or lessons learned.
- Numbersnot described.
Reports real figures: latency, cost, accuracy, scale or impact.
Suggested by Darshan Chougule
Want the architecture, code and numbers? They’re in the original post.
Read the full write-up on zenml.io (opens in a new tab)
Elastic
JetBrains
Qdrant