搜索结果：生成式引擎优化

VAC Memory System is an open-source conversational memory framework for LLM agents that achieved 80.1% accuracy on the LoCoMo 2025 benchmark through a hybrid retrieval architecture combining MCA gating, FAISS semantic search, BM25 lexical search, and cross-encoder reranking. 原文翻译： VAC记忆系统是一个开源的大语言模型智能体对话记忆框架，通过结合MCA门控、FAISS语义搜索、BM25词法搜索和交叉编码器重排序的混合检索架构，在LoCoMo 2025基准测试中实现了80.1%的准确率。

trawl is an LLM-powered web scraping tool that extracts structured data from websites using semantic queries instead of CSS selectors. It automatically adapts to site redesigns and runs efficiently in Go after initial strategy derivation. 原文翻译： trawl是一款基于LLM的网页爬取工具，通过语义查询而非CSS选择器从网站提取结构化数据。它能自动适应网站改版，并在初始策略推导后以Go语言高效运行。

Llama-github is a powerful Agentic RAG tool that retrieves relevant code snippets, issues, and repository information from GitHub to enhance LLM chatbots, AI agents, and Auto-dev agents for solving complex coding tasks. 原文翻译： Llama-github 是一款强大的 Agentic RAG 工具，可从 GitHub 检索相关代码片段、问题和仓库信息，赋能 LLM 聊天机器人、AI 智能体和自动开发智能体，以解决复杂的编程任务。

Stardog's benchmark demonstrates a 1 trillion-edge knowledge graph with sub-second query performance using distributed virtual graphs across hybrid cloud environments, achieving 98% cost savings compared to traditional approaches. 原文翻译： Stardog的基准测试展示了一个包含1万亿边的知识图谱，通过跨混合云环境的分布式虚拟图实现亚秒级查询性能，相比传统方法节省98%的成本。

l1m is a lightweight API that simplifies structured data extraction from unstructured text and images using LLMs, eliminating the need for prompt engineering through a schema-first approach with JSON Schema. 原文翻译： l1m 是一个轻量级 API，通过 JSON Schema 优先的方法，简化了使用 LLM 从非结构化文本和图像中提取结构化数据的过程，无需进行提示工程。

STDM (Self-Thinking Data Manifest) enables data artifacts to embed structured instructions that guide Large Language Models in processing, analyzing, and presenting data, creating interactive, self-di

EchOS is a self-hosted, AI-powered personal knowledge management system that captures, organizes, and retrieves information through natural conversation interfaces like Telegram, storing everything as plain Markdown files compatible with Obsidian. 原文翻译： EchOS 是一个自托管的、AI驱动的个人知识管理系统，通过Telegram等自然对话界面捕获、组织和检索信息，将所有内容存储为与Obsidian兼容的纯Markdown文件。

Cognee is an open-source framework for building AI systems with persistent, structured long-term memory through a hybrid architecture combining graphs, vectors, and structured data. 原文翻译： Cognee是一个开源框架，通过结合图、向量和结构化数据的混合架构，为构建具备持久化、结构化长期记忆的AI系统提供支持。

Cognee is an open-source framework for building sophisticated AI memory applications with hybrid architecture combining graphs, vectors, and structured data, enabling persistent, structured memory for AI agents. 原文翻译： Cognee 是一个开源框架，用于构建复杂的 AI 记忆应用程序，采用结合图、向量和结构化数据的混合架构，为 AI 智能体提供持久化、结构化的记忆能力。