How eriktromp/ragmeup works
Project type and purpose
• RAGMeUp is a full-stack, modular RAG (Retrieval-Augmented Generation) framework.
• The Python “server” directory implements your RAG logic (chunkers, retrievers, LAG orchestration, LLM calls, provenance).
• The “ui/scala” directory is a Play Framework front-end delivering a web app and calling the Python API.
• A PostgreSQL container provides the vector-enabled storage and user data persistence.
File structure
• Top-level:
• postgres/ – Dockerfile for a Postgres instance (likely with pgvector extension)
• server/ – Python RAG server with Dockerfile, .env template, and modules:
– LLMHelper.py, RAGHelper.py, ParagraphChunker.py, PostgresHybridRetriever.py, Reranker.py, provenance.py, server.py
• ui/scala/ – Play Framework app:
– app/controllers (HomeController.scala)
– app/models/daos (ChatDAO, FeedbackDAO)
– app/models/SQLTables.scala
– conf/routes, application.conf, evolutions (DB migrations)
– public assets (JS/CSS), views (.scala.html)
– build.sbt, plugins.sbt, build.properties
• Patterns:
• Two separate services (Scala frontend, Python backend) communicating over HTTP
• Modular “server” code adopting plug-in pattern for chunkers/retrievers/LLM
• Play’s MVC for UI, DAOs for data access, migrations under conf/evolutions
README for additional insights
• Modularity (“Use your own chunkers, vectorstores or retrievers”)
• Quickstart: spin up Python server and SBT UI separately
• Docs mention API references and architecture guides