Tag
作者分享对 GraphRAG 的兴趣,并尝试构建一个轻量本体论、无需向量搜索的 Jev-based GraphRAG 变体。
The author introduces Synapse, an MIT-licensed local Markdown memory vault for AI agents that imports ChatGPT/Claude exports and is queryable by OpenClaw over a stdio MCP server, keeping data on-disk and offline by default.
This paper introduces LLM-Guided Graph Pruning (LGP), a framework that uses LLM reasoning to refine graph-based ANN indices (e.g., HNSW, DiskANN) by pruning low-value neighbors, addressing the geometry–semantic mismatch in how these indices are built versus evaluated, and improving end-to-end retrieval over vanilla greedy search and LLM reranking.
This article critiques Retrieval-Augmented Generation (RAG) systems, demonstrating through experiments that embeddings fail to capture contextual details like contradictions, leading to hallucinations, and emphasizes the essential role of strong underlying models for accurate AI responses.
This post discusses the limitations of traditional databases for modern applications and promotes converged databases as a unified solution to handle diverse data types like relational, JSON, and vector data in a single system.
Manticore Search introduces automatic document chunking for vector search, improving recall for long documents by splitting them into chunks and embedding each chunk.
The article explains how agent memory systems can leak data due to post-filter tenant scoping in vector search and recommends scoping at write time to prevent cross-tenant data exposure.
Antfly has rewritten their search-and-inference database from Go to pure Zig for improved performance and zero dependencies. The article details their technical decisions, first principles, and implementations of vector search algorithms.
Alpine Agentic Search is an AI community meetup in Zurich on October 7, featuring technical talks on vector search, reranking, and agents for search workloads, with speakers from Nebius, NVIDIA, and Tavily.
An individual built a local long-term memory system for AI agents using markdown files and a local index, enabling persistent memory across sessions and including tests for false memories, with plans to potentially productize it.
The article discusses the challenge of memory staleness in long-running AI agents, where context becomes outdated and contradictory, and seeks practical solutions for maintaining reliable memory over time.
This article presents vector-bench, a tool designed to benchmark vector indexes across different databases with consistent conditions to accurately measure approximate nearest neighbor search performance.
This article describes a reference RAG agent implementation using Mastra and Elasticsearch vector store, demonstrated on a corpus of 500 sci-fi movies, with the agent built in approximately 60 lines of TypeScript.
Bonsai discovered a pervasive issue where embedding vectors are unnecessarily cast from float32 to float64, causing 'Float Bloat' that wastes storage and bandwidth, with an estimated global impact of over 20 Petabytes of overhead.
The author built a retrieval engine from scratch to benchmark HNSW against FAISS, finding that brute force search is faster for small document sets and that encoder latency dominates retrieval time, while RRF fusion of BM25 and dense retrieval improves quality significantly.
The article introduces Polign DB, a typed, stateless database designed for agent memory, optimized for edge deployment and cost efficiency. It demonstrates a demo showing how it can manage structured memory for AI agents with minimal resource usage.
A 30-question breakdown explaining how embeddings, vector search, and retrieval systems work, covering similarity metrics, training methods, indexing techniques, and evaluation.
This article provides a deep-dive into Retrieval-Augmented Generation (RAG) and vector search, with measured benchmarks on 100,000 documents showing the trade-offs between exact search and IVF index for speed and recall.
LatticeDB is an embedded property-graph database that integrates vector and full-text indexing into a single-file format, enabling graph traversal, similarity search, and BM25 search in one query layer for local applications.
This article explains embedding models and their role in LLM inference, covering architectures, traffic profiles, and optimization techniques like quantization.