Research OS · Professional AI Research Assistant

Professional AI research assistant for literature review, study design, scholarly writing, and reproducible research operations.

Visual tour · Install · Quickstart · Live GitHub signals · README media guide

Research OS is a local-first AI research assistant that helps researchers turn questions into structured, evidence-backed deliverables instead of one-off chat answers.

Research OS 是一个面向 专业研究工作 的本地优先 AI 科研助手系统，目标不是只给你一段对话，而是帮助你交付：

文献综述包
研究设计与实验计划
学术写作骨架
可追溯、可审计、可复现的研究工作区

它适合这样的场景：

你要把一个研究问题快速整理成 evidence-backed review brief
你要把已有材料收敛成 claim / MVP / acceptance checks / experiment plan
你要把结构化结果写成 abstract / outline / rebuttal notes
你希望研究项目不是散落在聊天记录里，而是沉淀成 workspace + deliverables

当前状态：Early public release。仓库已经可以直接公开使用，但仍建议把它看成一个持续演进中的专业科研助手原型。

Visual tour

1) Workspace + deliverables at a glance

2) Workflow + evidence traceability

Research workflow from question to audit

Claim to evidence to checks to deliverables traceability view

What makes this different from a generic AI chat tool

Research OS emphasizes four things:

Evidence first — back claims with sources before raising confidence
Structured research flow — scan → design → execute → write → audit
Deliverable-oriented outputs — research brief, evidence matrix, deliverable index, audit report
Reproducibility — runs, results, artifacts, and checks stay inside a reusable workspace

如果你只想随手问一个问题，普通聊天工具就够了。
如果你想把研究工作真正沉淀下来，Research OS 更适合。

Four core capabilities

1) Literature Review

research question 拆解
搜索策略与 evidence registry
baseline / novelty / reviewer objection 梳理
gap analysis

2) Study Design

claim graph
acceptance checks
MVP 收敛
experiment plan

3) Scholarly Writing

title / abstract
outline
related work brief
rebuttal notes

4) Reproducible Research Ops

run registry
provenance-aware results
audit report
exportable showcase package

Install

Option A — clone this repository

git clone https://github.com/TYSJY/human-research-scanning-system.git
cd human-research-scanning-system
python -m pip install .

Option B — install directly from GitHub

python -m pip install "git+https://github.com/TYSJY/human-research-scanning-system.git"

Option C — contributor mode

python -m pip install -e .

面向普通使用者时，优先使用 python -m pip install .。-e . 更适合开发者和贡献者。

Quickstart

A. 直接体验官方 demo

ros quickstart --launch-ui

B. 复制官方 demo 到你自己的目录

ros demo --root projects --name my-demo --launch-ui

C. 从空白项目开始

ros init --root projects --name my-research --title "My Research Project"
ros ui projects/my-research

Export deliverables

除了 ros audit 之外，这一版额外强调 成果物导出。

ros showcase <项目路径>

它会生成：

reports/research_brief.md
reports/evidence_matrix.csv
reports/deliverable_index.md

也就是说，项目不仅能跑，还能把当前研究状态整理成适合协作、汇报和交接的结果包。

Live GitHub signals

当前仓库已经公开，所以 README 不再使用“上线前占位”文案，而是直接接入真实仓库信号。

Repository structure

research_os/              Python package
research_os/_resources/   打包到安装版里的资源镜像（由同步脚本维护）
configs/                  示例 provider / executor 配置
control_plane/            agents、prompts、schemas、workflows
docs/                     产品文档、参考资料、维护文档、历史归档
examples/                 面向读者的成果物示例包
projects/                 官方工作区样例
scripts/                  辅助脚本（含 bundled resource 同步脚本）
templates/                空项目模板与 run 模板
tests/                    自动化测试
.github/                  CI、Issue 模板、PR 模板、发布工作流

Read these docs first

docs/product_positioning.md
docs/scenarios.md
docs/deliverables.md
docs/agent_roles.md
docs/evaluation_framework.md
docs/maintainers/public_release_checklist.md
docs/maintainers/open_source_audit.md

Official workspace samples and deliverable packs

官方工作区样例

projects/sample_joint_tri_runtime_v4_2/
projects/sample_joint_tri_compress_v4_1/

成果物示例包

examples/literature_review_pack/
examples/study_design_pack/
examples/writing_pack/

前者展示 工作区如何推进，后者展示 这个系统应该交付什么。

Windows

已装 Python

双击 install_windows.bat
双击 start_windows.bat
浏览器默认打开 http://127.0.0.1:8765/

没装 Python

直接双击：

bootstrap_python_and_start.bat

它会尝试检测并安装 Python，然后自动安装依赖并启动工作台。

Cite this project

如果你在论文、研究报告或演示中使用了本项目，请使用根目录下的 CITATION.cff。GitHub 会自动为它显示 Cite this repository 入口。

Development and release

python -m pip install -e .
pytest -q
python -m compileall research_os tests
python scripts/sync_bundled_resources.py --check

Tagging v* now triggers the release workflow in .github/workflows/release.yml, which builds distributions and attaches them to a GitHub Release.

Security

漏洞报告请参考 SECURITY.md
上线前建议按 docs/maintainers/public_release_checklist.md 打开 GitHub 的 topics、social preview、secret scanning、push protection、private vulnerability reporting

License

MIT License.

Name		Name	Last commit message	Last commit date
.github		.github
configs		configs
control_plane		control_plane
docs		docs
examples		examples
projects		projects
research_os		research_os
scripts		scripts
templates		templates
tests		tests
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CITATION.cff		CITATION.cff
CODEOWNERS		CODEOWNERS
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
MANIFEST.in		MANIFEST.in
QUICKSTART.md		QUICKSTART.md
README.md		README.md
SECURITY.md		SECURITY.md
bootstrap_python_and_start.bat		bootstrap_python_and_start.bat
install_windows.bat		install_windows.bat
pyproject.toml		pyproject.toml
requirements-optional.txt		requirements-optional.txt
start_windows.bat		start_windows.bat
start_windows.ps1		start_windows.ps1

Latest commit

History

Folders and files

Repository files navigation