dxww123/dsh-alphasolve

AlphaSolve helps you work on difficult mathematics problems in DeepSeek Harness: it develops candidate proofs, checks them independently, and revises them. It follows the workflow of AlphaSolve.

catalog descriptioncatalog 简介 / catalog description:Session-scoped AlphaSolve workflow for DeepSeek Harness

Project Overview项目介绍

AlphaSolve for DSH is a native DeepSeek Harness plugin that integrates the external AlphaSolve mathematical workflow into DSH 0.2.0-rc.1 sessions. The bundle is installed through the dsh plugin --profile web add and dsh plugin --profile headless add commands for the Web and Headless profiles, or via a pnpm pack tarball loaded through DeepSeek Harness Desktop 0.2.0-rc.1's Plugin Manager. The shipped controller stays dormant by default and only activates the solver tools for a session when the user explicitly mentions AlphaSolve and asks it to solve the selected workspace's problem.md. The plugin is a DSH-only product: there is no support for Claude Code, Codex, Cursor, Gemini CLI, Copilot, OpenCode, or Cline.

A typical workflow begins with a problem.md in the current workspace. A Curator keeps one durable Session across restarts and generations, reusing a remembered wiki map and reading only relevant notes. The Main Session dispatches two helper roles, compute and numerical_experiment, each running a separate persistent Python process with its own SymPy namespace that survives ordinary exceptions but is reset on timeout, cancellation, or interpreter failure. Workers join the same preset generation as their main session, narrowed by AlphaSolve's role-specific restrictions and stripped of shell, run_code, Web, and subagent access. The target users are researchers and engineers performing symbolic derivation, numerical experimentation, and mathematical modeling inside DSH.

Dependencies require DSH 0.2.0-rc.1, Cordis 4.0.4, and a Python 3.10+ virtual environment created by node scripts/setup-python.mjs, which installs SymPy 1.14.0 and mpmath 1.3.0 under ${DSH_HOME:-~/.dsh}/runtimes/alphasolve-python with system site packages disabled. The license is MIT. Compute helpers run under Harness's read-only process policy plus an extra Python policy that blocks writes, cross-worker reads, subprocess spawning, and network use; on Windows the read-only backend is only partially enforced and the Python audit/syntax policies are not a strong sandbox against hostile code. First-run caveats include pinning the full commit SHA during gh-based installation, verifying dsh-alphasolve appears in dsh --dump-config for every custom profile, and restarting any running dsh web process after installation or upgrade because a browser refresh does not load new code.

AlphaSolve for DSH 是面向 DeepSeek Harness 的原生插件,将外部 AlphaSolve 数学求解工作流整合进 DSH 0.2.0-rc.1 会话环境。插件以 bundle 形式通过 dsh plugin --profile web add 或 dsh plugin --profile headless add 命令安装到 Web 与 Headless 配置,也可通过 pnpm pack 生成的本地 tarball 在 DeepSeek Harness Desktop 0.2.0-rc.1 的插件管理器内加载。安装后控制器默认休眠,仅当用户在当前工作区的 problem.md 中显式提及 AlphaSolve 并请求求解时,对应会话才加载求解工具。

典型使用流程是用户在工作区编写 problem.md,由 Curator 维护跨会话的持久会话与 wiki 映射,Main Session 调度 compute 与 numerical_experiment 两个 helper 角色调用 Python/SymPy 完成符号计算,并将结果回传主会话整理。每个 helper 拥有独立的 SymPy 命名空间与持久进程,异常、超时或取消都会重置该进程。该插件面向需要在 DSH 内做数学建模、符号推导与数值实验的研究者与工程师。

依赖方面要求 DSH 0.2.0-rc.1、Cordis 4.0.4,并需通过 node scripts/setup-python.mjs 创建 Python 3.10+ 虚拟环境,安装 SymPy 1.14.0 与 mpmath 1.3.0。License 为 MIT。计算 helper 受 Harness 只读进程策略及 Python 策略约束,无法写项目文件、读其他 worker 草稿、启动子进程或访问网络;Windows 下的只读后端仅提供部分强制执行,Python 审计策略并非针对恶意代码的强沙箱。首次运行请先 dsh --version 确认版本,再用 dsh --dump-config 校验 dsh-alphasolve 条目已注入目标配置,并在升级后重启 dsh web 进程。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 5 stars - very few users, little community feedback星标只有 5,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add github:dxww123/dsh-alphasolve

把 dxww123/dsh-alphasolve 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

AlphaSolve for DSH

中文使用说明

AlphaSolve helps you work on difficult mathematics problems in DeepSeek Harness: it develops candidate proofs, checks them independently, and revises them. It follows the workflow of AlphaSolve.

Install

  1. Install DeepSeek Harness Desktop 0.2.0-rc.2 and configure a working model connection.
  2. Open the plugin manager, choose Add plugin, enter dsh-alphasolve, and install. Choose Enable now and restart if prompted.
  3. Complete the one-time Python/SymPy setup. Desktop includes Python, but installing the plugin does not install SymPy.

No GitHub account or source build is needed. If you have the older @dsh-external/dsh-alphasolve package, replace it before installing this one.

Solve a problem

  1. Create a folder and put your question in problem.md. You may add hint.md for ideas, references, or approaches to try.

  2. Open that folder as a workspace in DSH and start a conversation in standard mode.

  3. Send:

    Use AlphaSolve to solve the problem in problem.md.

Click AlphaSolve at the top of the conversation to see progress and open each participant's working transcript. When the workflow accepts a solution, it writes solution.md in your folder.

Start with a small problem. Repeated checking can take a long time and incur substantial model charges; the default is two parallel solving tasks. Automated checks do not guarantee a correct proof—review the mathematics yourself.

npm package · Advanced usage and development · Compatibility · Report a problem

← 上一个 Prev dsh-login-gateway 下一个 Next dsh-version-update →