$ show pathwaycom/llm-app
This repository is a collection of deployable templates and educational cookbooks for building real-time LLM applications using the Pathway framework. It provides ready-to-deploy RAG (Retrieval-Augmented Generation) solutions with live data synchronization from various sources (files, Google Drive, SharePoint, etc.), built-in vector and hybrid search, and integration with LangGraph for agentic workflows. The templates are production-oriented with Docker support, while the cookbooks demonstrate advanced patterns like Self-RAG with self-reflection and query rewriting. The codebase is well-structured but exhibits some duplication across templates, and the core logic relies heavily on the external Pathway library.
The LLM identified 3 files worth highlighting for reuse:
A Jupyter notebook demonstrating a complete Self-RAG agent built with Pathway and LangGraph. It implements a retrieval-augmented generation pipeline with self-reflection steps: document relevance grading, hallucination detection, answer usefulness checking, and query rewriting. The notebook is self-contained, uses Pydantic for structured LLM output, and can be adapted to other document sources with minimal changes. It showcases reusable patterns for building agentic RAG systems with real-time indexing.
A companion notebook that extends the Self-RAG agent by deploying it as an HTTP endpoint using Pathway's serve_callable mechanism. It demonstrates how to wrap a LangGraph agent as a Pathway REST endpoint, enabling production-like serving. The pattern of wrapping any Python function as a Pathway endpoint is reusable for deploying custom agent workflows.
A custom module implementing DeckRetrieverWithFileSave, a specialized retriever for slide decks that saves parsed content to disk. It demonstrates how to extend Pathway's DocumentStore with custom retrieval logic and file persistence. The pattern is reusable for any document type requiring custom parsing and storage.