nilsherzig/LLocalSearchArchived
LLocalSearch
LLocalSearch 是一个注重隐私的搜索引擎和代理框架，使用本地托管的大语言模型来搜索互联网并聚合答案。它作为一个检索增强生成（RAG）界面，所有查询和处理都保留在用户自己的硬件上，以确保数据隐私并消除对外部云 API 提供商的依赖。
该系统采用一系列自主代理，执行递归互联网搜索，多次调用搜索工具以收集和综合信息。它协调这些模型来推理复杂查询，提供实时日志流和来源引用以监控代理的推理过程。
该项目包括用于支持多轮对话和迭代优化的状态化对话管理功能。它集成了基于工具的信息检索和本地模型推理，以提供从初始搜索到最终答案生成的完整流水线。
Features
- Recursive Research Agents - Implements autonomous agents that perform recursive internet searches to gather and synthesize comprehensive information.
- Autonomous Research Agents - Automates the process of gathering information through recursive internet searches handled by autonomous local agents.
- Local Model Generation - Provides answer generation using locally hosted models to ensure data privacy and remove cloud dependencies.
- Local LLM Search - Uses a chain of local language models to find answers to questions without requiring external API keys.
- API-Free Local Inference - Executes foundation models on local hardware to eliminate dependency on external cloud API providers.
- Private Information Retrieval - Retrieves data and answers using locally hosted models to ensure all processing remains on the user's hardware.
- Search Query Recursion - Employs autonomous agents that iteratively generate new search queries to gather current external information.
- Local LLM Research Agents - Provides a framework of autonomous agents that recursively call search tools using local model inference.
- Tool-Augmented Retrieval - Equips language models with search tools to fetch external web data into the local prompt context.
- LLM Task Chains - Coordinates a sequence of LLM calls in a pipeline to process queries from initial search to final synthesis.
- Context-Aware Conversation Managers - Combines local language models with retrieved search data to maintain context in multi-turn dialogues.
- Conversation State Persistence - Maintains a persistent record of interactions to provide context for follow-up queries and iterative refinements.
- Conversational Search Engines - Allows users to refine search results and ask follow-up questions within a persistent, LLM-powered conversation.
- Retrieval-Augmented Generation Interfaces - Provides a specialized interface for monitoring the real-time reasoning and source attribution of local RAG.
- AI-Powered Search Aggregators - Coordinates multiple local models to search and synthesize information into comprehensive answers without external APIs.
- Privacy-Focused Search Engines - Functions as a self-hosted search tool that processes all queries locally to ensure complete data privacy.
- Task Progress Streams - Pushes real-time logs and source citations of the agent's reasoning process directly to the user interface.
- Agent Frameworks - Locally running search aggregator using LLM agents.
- Application Development - Local web search implementation using LLM chains.
