Subject
24 entries
Rag
Bookmarks
MOL: pipeline language with auto-tracing for AI/RAG
MOL is a pipeline programming language with auto-tracing built into the pipe operator, first-class AI types (Thought, Memory, Embedding, Chunk), and RAG as a one-liner. Transpiles to Python and JavaScript.
txtai: All-in-One Embeddings Database
txtai is an all-in-one open-source embeddings database combining semantic search, LLM orchestration, and language model workflows. Positions itself as the engine underneath an AI application rather than a standalone vector database.
Dense-X-Retrieval: Proposition-Level RAG
Dense-X-Retrieval is a LlamaIndex pack implementing proposition-level retrieval — splitting documents into atomic factual propositions rather than chunks, then retrieving at proposition granularity. Improves precision by matching query semantics at a finer level than paragraph chunks.
Comparing RAG Copilots: OpenAI, Anthropic, Perplexity, and More
Graphlit's comparison of RAG-powered copilots from OpenAI, Anthropic, Perplexity, and others against unstructured data retrieval tasks. Useful late-2023 benchmark of the RAG quality landscape across providers before the tooling matured.
Danswer: Natural Language Q&A Over Private Sources
Danswer (now Onyx) is an open-source enterprise Q&A system that connects to Slack, GitHub, Confluence, and other internal tools to answer natural language questions over private knowledge. Self-hosted with strong access-control enforcement — a production-ready alternative to building RAG from scratch.
ChunkViz: Visualize Text Splitting for RAG Pipelines
ChunkViz is a visual tool for comparing different text splitting strategies — see exactly how your documents get chunked with different chunk sizes, overlap settings, and splitting methods. Essential for debugging RAG pipelines where chunking quality directly affects retrieval quality.
SuperDuperDB: Bring AI to Your Database
SuperDuperDB integrates AI models and APIs directly with existing databases — train, manage, and query models where your data already lives rather than moving data to a separate vector database. A database-native alternative to building a separate AI data pipeline.
LLM Course: Roadmaps and Notebooks for Large Language Models
A comprehensive GitHub repository by Maxime Labonne with roadmaps, notebooks, and articles for learning about LLMs — from fundamentals through fine-tuning and deployment. One of the most starred LLM learning resources, covering the full practitioner pipeline.
The Architecture of Today's LLM Applications
GitHub's overview of LLM application architecture patterns as of late 2023 — the standard components (inference, context, orchestration, state, tools) and how they fit together. A useful snapshot of the consensus architecture before it fragmented into frameworks.
RAG at Planet Scale
Arcus describes their multi-tiered RAG approach for handling massive external data corpora at planet scale — one of the largest RAG deployments of 2023. The key innovation is tiered retrieval that narrows the candidate pool progressively rather than searching the full index directly.
LlamaIndex Retrieval and Chunk Evaluation Notebook
A LlamaIndex Google Colab notebook for evaluating retrieval quality and chunk size in RAG pipelines — demonstrating how to measure retrieval hit rate and MRR across different chunk sizes. Practical tooling for the underappreciated problem of RAG evaluation.
Contextually Splitting Documents — Neum AI
Neum AI introduces context-aware document splitting that improves RAG retrieval quality on structured documents like SEC filings and contracts — by splitting along semantic boundaries rather than fixed character counts. A practical improvement to the chunking step most RAG tutorials treat as an afterthought.
RAG Is More Than Just Embedding Search
Jason Liu's influential post arguing that RAG systems need more than vector similarity search — covering query understanding, routing, reranking, and structured extraction as the layers that separate working RAG from production-grade RAG. Written for the Instructor library blog.
PDFTriage: Question Answering over Long, Structured Documents
PDFTriage is a method for question answering over long, structured PDF documents that uses the document's structure (pages, sections, tables of contents) as a triage layer before retrieval — letting the LLM navigate the document intelligently rather than relying on flat embedding search.
Optimal Chunk Size for Large Document Summarization
Vectify AI introduces a method to automatically determine the optimal chunk size for large document summarization with LLMs — rather than fixed-size chunking, the approach finds the chunk granularity that maximizes summary quality for a given document type.
Patterns for Building LLM-based Systems & Products
Eugene Yan's comprehensive guide to patterns for building LLM-based systems — covering evals, RAG, fine-tuning, caching, guardrails, defensive UX, and user feedback collection. One of the most referenced practical engineering posts of 2023.
Build a Support Bot with Supabase, OpenAI, and Windmill
Windmill's tutorial on building a Discord/Slack support bot using Supabase for vector storage and OpenAI for embeddings and generation — a worked RAG example for team documentation Q&A. Shows how to wire up the standard stack (embed, store, retrieve, generate) for a specific product use case.
AnythingLLM: Documents to Chatbot
AnythingLLM is a full-stack open-source application for turning any document collection into a chatbot — with a polished UI, workspace management, and multi-model support. The self-hosted alternative to ChatGPT Enterprise for teams wanting document Q&A over their own files.
OpenChat: LLM Custom Chatbot Console
OpenChat is an open-source LLM chatbot console with document ingestion and custom bot creation — an early self-hosted alternative to ChatGPT for teams wanting to deploy domain-specific chatbots over their own data.
ChatGPT + Google Drive with LangChain
A tutorial for connecting ChatGPT to Google Drive using LangChain and Python in 30 lines of code — an early canonical example of the document Q&A pattern using LangChain's document loaders and retrieval chain.
LlamaIndex: Composable Indices and Query Decomposition
A LlamaIndex notebook demonstrating composable indices with query decomposition on city data — showing how to break complex queries into sub-queries across multiple document indices and synthesize the results. An early tutorial on the multi-hop retrieval patterns LlamaIndex specialized in.
Generate Rather Than Retrieve: Large Language Models Are Strong Context Generators
Yu et al. (2022) show that prompting an LLM to generate its own background context before answering a question (GenRead) outperforms retrieval-based approaches on several knowledge-intensive NLP benchmarks. The result challenges the assumption that retrieval is necessary for grounding LLM outputs.
Llama Hub: LlamaIndex Data Connector Marketplace
Llama Hub is the LlamaIndex community marketplace for data loaders — connectors that pull data from Notion, Slack, GitHub, databases, APIs, and more into LlamaIndex for RAG pipelines. The npm registry equivalent for LLM data connectors.
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Lewis et al. (Facebook AI, 2020) introduce Retrieval-Augmented Generation, a hybrid architecture that combines dense passage retrieval with seq2seq generation to ground language model outputs in a non-parametric knowledge store. RAG defined the template that most production knowledge-grounded LLM systems follow today.