Data & Analytics
RAG.md · 63 documents
Project Memory
Documents a production-ready RAG system for Paris cultural events with hybrid search, LLM generation, and Streamlit UI.
Metrics
Documents evaluation metrics for RAG systems, covering retrieval and generation quality, plus guidance on building golden datasets.
Code Span Semantic Chunking Executive Summary
Recommends an AST-driven chunking strategy for code retrieval, preserving function/class boundaries to improve precision and reduce token waste.
Text Splitting with LangChain
Explains LangChain's RecursiveCharacterTextSplitter for dividing documents into chunks that fit within LLM context windows.
语义块切分 Semantic Chunking
Implements semantic chunking for RAG by splitting text at points where sentence similarity drops below a threshold.
rag-vector-chunking
Explains embedding-based retrieval, cosine similarity, RAG evaluation metrics, chunking strategies, and retrieval optimization techniques.
语义块切分 Semantic Chunking
Implements three semantic chunking methods (percentile, standard deviation, IQR) for RAG pipelines, splitting text by sentence similarity.
Chunking Strategies
Explains the trade-off between chunk size for entity extraction and retrieval precision, with code examples for Neo4j's FixedSizeSplitter.
Semantic Chunking
Defines a quote-aware chunking and indexing strategy for semantic search over mailing-list corpora, with block parsing rules and retrieval flow.
Chunking Fundamentals
Teaches chunking strategies for RAG pipelines, covering fixed-size, recursive, and semantic splitting with LangChain code examples.
Late Chunking Strategy
Explains late chunking for RAG: embed entire documents first, then split into context-aware chunks to preserve cross-references.
ADR-0006: Paragraph-Boundary Chunking
Splits documents into chunks at paragraph boundaries for embedding and retrieval, with configurable token limits and overlap.
embedding and chunking
Compares late chunking against traditional chunking for RAG embedding, with Jina API code and similarity benchmarks.
Text Chunking - Anton (ChromaDB)
Summarises a technical session on text chunking strategies for RAG, covering heuristic and semantic approaches, evaluation metrics, and practical advice.
With Jina Embeddings v3
Explains late chunking, a technique that embeds full documents before splitting into chunks to preserve context, improving retrieval recall by 3-10%.
Text Chunking - Anton (ChromaDB)
Summarises a talk on text chunking for RAG, covering heuristic vs semantic approaches, evaluation metrics, and practical advice.
Chunking & Embeddings
Documents a chunk preview API and embedding pipeline for building a GPT chat knowledge base with FAISS indices.
Search and Retrieval
Documents a 6-stage RAG search pipeline with hybrid retrieval, reranking, Chain of Thought reasoning, and source attribution.
Semantic Chunking
Explains and implements three semantic chunking methods for RAG, with code examples and evaluation on two datasets.
Long Context Chunking
Splits long documents into overlapping chunks and averages their embeddings to avoid context limit errors.
Large-scale Document Retrieval with ElasticSearch
Walks through building an Elasticsearch index with dense vectors and comparing ANN algorithms (LSH, IVFPQ, HNSW, FlatL2) for movie retrieval.
Introduction ot Information_Retrieval
Summarises key concepts from the IIR textbook chapters on Boolean retrieval, indexing, scoring, evaluation, and classification.
Information Retrieval - Topic Modelling Relevent Part
Serves as a structured reference for information retrieval concepts, covering models, term weighting, evaluation measures, and link analysis.
lib-ai-app-community-rag
Collects community insights, leaderboard links, and production war stories for building RAG systems, especially with multi-modal and large-scale documents.