LearningCircuit / local-deep-research

Local Deep Research is an AI-powered research assistant designed for deep, agentic research. It performs comprehensive, agent-driven research using multiple LLMs and search engines, complete with proper citations. Local Deep Research 是一款由 AI 驱动的研究助手，专为深度代理式研究而设计。它利用多个大语言模型（LLM）和搜索引擎进行深度、代理式的研究，并提供准确的引用。

▶️ Watch Review by The Art Of The Terminal ▶️ 观看 The Art Of The Terminal 的评测

🚀 What is Local Deep Research? 🚀 什么是 Local Deep Research？

An AI research assistant you control. Run locally for privacy, use any LLM, and build your own searchable knowledge base. You own your data and see exactly how it works. 这是一个由你掌控的 AI 研究助手。在本地运行以确保隐私，支持使用任何 LLM，并构建你自己的可搜索知识库。你拥有自己的数据，并能清楚地了解其运作方式。

⚡ Quick Start ⚡ 快速开始

Option 1: Docker Run (Linux) 选项 1：Docker 运行 (Linux)

Step 1: Pull and run Ollama

第一步：拉取并运行 Ollama

docker run -d -p 11434:11434 —name ollama ollama/ollama docker exec ollama ollama pull gpt-oss:20b

Step 2: Pull and run SearXNG for optimal search results

第二步：拉取并运行 SearXNG 以获得最佳搜索结果

docker run -d -p 8080:8080 —name searxng searxng/searxng

Step 3: Pull and run Local Deep Research

第三步：拉取并运行 Local Deep Research

docker run -d -p 5000:5000 —network host
—name local-deep-research
—volume “deep-research:/data”
-e LDR_DATA_DIR=/data
localdeepresearch/local-deep-research

Option 2: Docker Compose 选项 2：Docker Compose

CPU-only (all platforms): 仅 CPU (所有平台): curl -O https://raw.githubusercontent.com/LearningCircuit/local-deep-research/main/docker-compose.yml && docker compose up -d

With NVIDIA GPU (Linux): 使用 NVIDIA GPU (Linux): curl -O https://raw.githubusercontent.com/LearningCircuit/local-deep-research/main/docker-compose.yml && \ curl -O https://raw.githubusercontent.com/LearningCircuit/local-deep-research/main/docker-compose.gpu.override.yml && \ docker compose -f docker-compose.yml -f docker-compose.gpu.override.yml up -d

Open http://localhost:5000 after ~30 seconds. For GPU setup, environment variables, and more, see the Docker Compose Guide. 约 30 秒后访问 http://localhost:5000。有关 GPU 设置、环境变量及更多信息，请参阅 Docker Compose 指南。

Option 3: pip install 选项 3：pip 安装

pip install local-deep-research

Works on Windows, macOS, and Linux. SQLCipher encryption is included via pre-built wheels — no compilation needed. PDF export on Windows requires Pango (setup guide). If you encounter issues with encryption, set export LDR_BOOTSTRAP_ALLOW_UNENCRYPTED=true to use standard SQLite instead. More install options → 适用于 Windows、macOS 和 Linux。通过预构建的 wheel 包内置了 SQLCipher 加密，无需编译。Windows 上的 PDF 导出功能需要 Pango（查看设置指南）。如果遇到加密问题，可设置 export LDR_BOOTSTRAP_ALLOW_UNENCRYPTED=true 以使用标准 SQLite。更多安装选项 →

🏗️ How It Works 🏗️ 工作原理

Research 研究

You ask a complex question.
LDR: Does the research for you automatically.
Searches across web, academic papers, and your own documents.
Synthesizes everything into a report with proper citations.
你提出一个复杂的问题。
LDR：自动为你进行研究。
在网络、学术论文和你自己的文档中进行搜索。
将所有内容综合成一份带有准确引用的报告。

Choose from 20+ research strategies for quick facts, deep analysis, or academic research. 你可以从 20 多种研究策略中进行选择，以获取快速事实、深度分析或进行学术研究。

New: LangGraph Agent Strategy — An autonomous agentic research mode where the LLM decides what to search, which specialized engines to use (arXiv, PubMed, Semantic Scholar, etc.), and when to synthesize. Early results are promising — it adaptively switches between search engines based on what it finds and collects significantly more sources than pipeline-based strategies. Select langgraph-agent in Settings to try it. 新功能：LangGraph 代理策略 — 一种自主代理研究模式，LLM 会决定搜索什么、使用哪些专业引擎（arXiv、PubMed、Semantic Scholar 等）以及何时进行综合。初步结果非常理想——它能根据发现的内容自适应地切换搜索引擎，并比流水线式策略收集到更多的来源。在设置中选择 langgraph-agent 即可尝试。

Build Your Knowledge Base 构建你的知识库

Every research session finds valuable sources. Download them directly into your encrypted library—academic papers from ArXiv, PubMed articles, web pages. LDR extracts text, indexes everything, and makes it searchable. Next time you research, ask questions across your own documents and the live web together. Your knowledge compounds over time. 每次研究都会发现有价值的来源。你可以将它们直接下载到加密库中——包括来自 ArXiv 的学术论文、PubMed 文章和网页。LDR 会提取文本、索引所有内容并使其可搜索。下次研究时，你可以同时针对自己的文档和实时网络进行提问。你的知识会随着时间的推移而积累。

🛡️ Security 🛡️ 安全性

Your data stays yours. Each user gets their own isolated SQLCipher database encrypted with AES-256 (Signal-level security). No password recovery means true zero-knowledge—even server admins can’t read your data. Run fully local with Ollama + SearXNG and nothing ever leaves your machine. 你的数据永远属于你。每个用户都拥有自己独立的 SQLCipher 数据库，并使用 AES-256（Signal 级别安全性）进行加密。没有密码恢复功能意味着真正的零知识——即使是服务器管理员也无法读取你的数据。通过 Ollama + SearXNG 在本地完全运行，没有任何数据会离开你的机器。

In-memory credentials: Like all applications that use secrets at runtime — including password managers, browsers, and API clients — credentials are held in plain text in process memory during active sessions. This is an industry-wide accepted reality, not specific to LDR: if an attacker can read process memory, they can also read any in-process decryption key. We mitigate this with session-scoped credential lifetimes and core dump exclusion. 内存中的凭据： 与所有在运行时使用密钥的应用程序（包括密码管理器、浏览器和 API 客户端）一样，凭据在活动会话期间以明文形式保存在进程内存中。这是行业普遍接受的现实，并非 LDR 特有：如果攻击者能够读取进程内存，他们也能读取任何进程内的解密密钥。我们通过会话范围的凭据生命周期和核心转储排除来缓解此风险。

Supply Chain Security: Docker images are signed with Cosign, include SLSA provenance attestations, and attach SBOMs. 供应链安全： Docker 镜像均使用 Cosign 签名，包含 SLSA 溯源证明，并附带 SBOM（软件物料清单）。

🔒 Privacy & Data 🔒 隐私与数据

Local Deep Research contains no telemetry, no analytics, and no tracking. We do not collect, transmit, or store any data about you or your usage. No analytics SDKs, no phone-home calls, no crash reporting, no external scripts. Usage metrics stay in your local encrypted database. Local Deep Research 不包含任何遥测、分析或跟踪功能。我们不会收集、传输或存储关于你或你使用情况的任何数据。没有分析 SDK，没有“回传”调用，没有崩溃报告，没有外部脚本。使用指标仅保留在你本地的加密数据库中。

📊 Performance 📊 性能

~95% accuracy on SimpleQA benchmark (preliminary results).
Tested with GPT-4.1-mini + SearXNG + focused-iteration strategy.
Comparable to state-of-the-art AI research systems.
Local models can achieve similar performance with proper configuration.
在 SimpleQA 基准测试中准确率约为 95%（初步结果）。
使用 GPT-4.1-mini + SearXNG + 聚焦迭代策略进行测试。
可媲美最先进的 AI 研究系统。
通过适当的配置，本地模型也能达到类似的性能。

🧭 Picking a model? Use the community benchmarks 🧭 如何选择模型？使用社区基准测试

Not sure which local model to run with LDR? The community-maintained LDR Benchmarks dataset on Hugging Face tracks accuracy across models, search engines, and research strategies — it’s the fastest way to see which Ollama / LM Studio / llama.cpp models actually work. 不确定该在 LDR 中运行哪个本地模型？Hugging Face 上由社区维护的 LDR 基准测试数据集跟踪了不同模型、搜索引擎和研究策略的准确率——这是了解哪些 Ollama / LM Studio / llama.cpp 模型真正有效的最快途径。