Tag
A personal Android notes app built with Hark Pro using Kotlin and Jetpack Compose, featuring an AI thinking partner with a connection engine, idea development modes, knowledge graph, and meaning-based search over notes.
The Last Museum 是一个可对全球主要博物馆的 580 万件（更新数据约 630 万件）艺术品进行语义搜索的工具，让公众通过自然语言检索海量艺术藏品。
介绍一种基于768维嵌入空间上的吸收马尔可夫链（Absorbing Markov Chain）的语义分流引擎，通过在60节点领域流形上做随机游走来替代朴素余弦相似度路由，实现<2ms的零token意图分类，从而在唤醒LLM之前完成路由。文中给出JevBench基准测试结果：四层级联系统（Gevva0）在240个条目上达到100%召回、89.17%精度，平均p50延迟46ms，并配备否定/语法、流形邻近OOD和熵与边界三重防护网。
Obsidian Mind turns an Obsidian Vault into a long-term memory bank for AI coding assistants, using QMD semantic retrieval and Markdown notes to let Claude Code, Codex CLI, and Gemini CLI retain project context, technical decisions, and accumulated experience across sessions.
Sergio Pineiro 惊叹某开源搜索工具能在 30 毫秒内检索 3200 万条记录，性能极其惊人，并且设置简单、支持模糊搜索、向量搜索和语义搜索。
Abdur-Rahman Butler (Isaacus) 制作了一个交互式教程加 Wordle 风格游戏 SemanticGuessr，通过在 43,339 个英语单词的二维语义地图上「围捕」目标词来讲解向量嵌入、余弦相似度、语义搜索、聚类等概念，并通过降维量化把模型压缩到 12.3MB 实现纯前端运行。
互动式指南讲解向量嵌入的历史、原理与实践应用，通过可视化展示语义空间、向量运算与语义相似性，并演示语义搜索、聚类、零样本分类等真实场景。文中使用 Kanon 2 Embedder 的真实嵌入构建了一个交互式词语地图游戏。
Maxime Rivest launched ask.sciencemadereadable.com, a free tool that searches ~200 million scientific study titles and abstracts from OpenAlex by meaning (or exact keywords) and returns plain-text results with no API key required. It offers URL-based queries, filtering options (year, type, open access), a JSON API, and links to a plain-language 'map of science' version of each paper.
VectorDash 是一款面向 Windows 11 的全功能搜索替代工具，灵感来自 macOS 的 DigUp，借助 EmbeddingGemma 2 实现跨图片、音频、视频、PDF 和 Office 文档的语义与精确搜索，并支持快速文件索引、全局启动器、MCP 服务器及本地加密存储。
The author argues that EmbeddingGemma 2 and embedding models in general are being underestimated, highlighting the open-source DigUp project — a Mac search app built on Google's EmbeddingGemma 2 that outperforms Spotlight with content, image, and audio search — and suggests embeddings could also improve RAG knowledge base search.
一位网友在 GitHub 上发现了一款基于 Google EmbeddingGemma2 小模型的 Mac 搜索应用，称其准确度远超系统自带 Spotlight，支持文件名、文件内容、图片和音频搜索。
Maxime Rivest 发布了 ScienceMadeReadable：一个免费、开源的科学文献语义搜索 API，覆盖约 2 亿篇论文，可直接被 AI 代理调用，无需密钥或注册，嵌入由 LightOn 的 DenseOn 模型驱动。
Recall is an open-source, local semantic file search tool for Linux built in Rust with GTK 4, using EmbeddingGemma 2 via llama.cpp to index PDFs, Office files, media and more for content-based queries with no network required after model download.
DigUp 是一款免费开源的 Mac 应用，本地运行 Google DeepMind 的 EmbeddingGemma 2 模型，通过文本、图像、音频和视频的统一嵌入空间实现语义搜索你的本地文件。
Maxime Rivest 介绍了一个可直接供 AI agent 调用的语义搜索 API，覆盖 2008 年至今的科学文献，只需发出 GET 请求即可获得检索结果。目前仍是 demo 版本，呼吁开发者试用并反馈改进建议。
作者提出一种新思路：构建大量用自然语言描述的任务专用 agent，通过语义搜索为每个请求匹配并执行正确的 agent，以此规避 Claude Code 等 agent 框架工具集过大、上下文腐化、不可预测和成本膨胀的问题。文中还对比了 Agent/MCP 网关、子 agent 和工具搜索等替代方案。
The author demonstrates how frontier models like Opus 5.5 can be used for historical research, using embedding-based semantic search over the GLOBALISE Dutch East India Company archive to surface a previously unnoticed 1615 Dutch eyewitness account of hunting dodos, alongside Carter Church's use of GPT-6 Astra to crack a Napoleonic-era cipher.
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.
A local, file-based long-term memory layer for personal AI that stores raw conversation logs with timestamps, enabling time-first search and maintaining full context without summarization.
FIND is an open-source Chrome extension that provides AI-powered semantic search for near real-time, meaning-based finding on the web.