用于复现并校验 ml-quant-trading protocol v1 CPU 基准测试,并生成可提交 issue 报告的 DSH 工具。需要本地 clone initial-d/ml-quant-trading,并将其设为 workspace 或 repoPath,同时自行准备 Python 和 PyTorch;本插件不会获取仓库或安装依赖。
安装
# Release 预构建包
dsh plugin --profile web add "https://github.com/initial-d/dsh-plugin-mlquant-benchmark/releases/download/v0.1.0/dsh-plugin-mlquant-benchmark-0.1.0.tgz"
# GitHub 源码(首次需按提示配置 allowBuilds 构建授权后重试)
dsh plugin --profile web add github:initial-d/dsh-plugin-mlquant-benchmark
装任何插件都等于在你的机器上跑第三方代码,权限和你本人一样大——能读你的文件、用你的凭据、访问网络,工具审批管不到它。GitHub 来源的插件还会在安装时执行构建脚本——pnpm 默认拦截,所以安装可能停在 ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED 或 ERR_PNPM_IGNORED_BUILDS;dsh 会打印出需要添加的确切键名,把它加进该 profile 的 pnpm-workspace.yaml 的 allowBuilds 下,重跑一次即可装上。放行构建本身就是一次信任判断:请只安装可信来源,并尽量锁定 commit(github:owner/repo#sha)。
README
该插件的 README 只有英文版本。
DeepSeek Harness tools for reproducing the
initial-d/ml-quant-trading
protocol v1 CPU benchmark.
The point is narrow: make a DSH agent able to run the existing benchmark, read the machine-readable artifact, validate it against the benchmark protocol, and draft an issue-ready report. This plugin does not add a trading agent, does not call market data APIs, and does not configure any model provider.
Why this exists
ml-quant-trading is a good reproducibility target for agent harnesses:
- deterministic synthetic benchmark input;
- fixed protocol v1 command, seed, panel size, repetitions, and thread counts;
- JSON artifact suitable for automated checking;
- public issue template for DeepSeek Harness benchmark reports;
- explicit boundary that benchmark throughput is not trading performance.
Challenge: can DeepSeek Harness reproduce a quant benchmark end to end, preserve the evidence bundle, and avoid turning runtime numbers into alpha claims?
Listed in
awesome-dsh-plugin
via PR #2573.
Run-To-Report Path
- Install the plugin from GitHub.
- Open an
initial-d/ml-quant-tradingcheckout in DSH. - Ask DSH to run, validate, summarize, and draft a benchmark report.
- Submit the drafted report through the dedicated issue template.
That path is intentionally small: the plugin turns DSH attention into a reproducible benchmark report, not an investment or leaderboard claim.
Tools
This package registers four DSH tools:
| Tool | Purpose |
|---|---|
mlquant_benchmark_v1_cpu |
Run the fixed protocol v1 CPU benchmark and write artifacts/benchmark-v1.json. |
mlquant_read_benchmark_json |
Read the JSON artifact and render a compact Markdown result table. |
mlquant_validate_benchmark_json |
Check protocol v1 fields, expected cases, fixed parameters, and variance warnings. |
mlquant_draft_github_issue |
Draft a DeepSeek Harness benchmark issue body from the JSON artifact. It does not post to GitHub. |
Install
Install the package in a DeepSeek Harness profile or preset environment:
dsh plugin --profile web add github:initial-d/dsh-plugin-mlquant-benchmark
The package declares a dsh.bundle manifest that inserts:
- id: mlquant-benchmark
name: dsh-plugin-mlquant-benchmark
If you use a local checkout while developing, add the same row manually:
- id: mlquant-benchmark
name: file:/path/to/dsh-plugin-mlquant-benchmark
This package is intentionally not published to npm yet. GitHub distribution is enough for the first DSH-facing benchmark reports; npm can come later if there is real usage.
Suggested DSH prompt
Read AGENTS.md, docs/benchmarking.md, and docs/reality_check.md.
Use the mlquant benchmark tools to run the protocol v1 CPU benchmark, validate
and read the JSON artifact, and draft a DeepSeek Harness benchmark report. Keep
the result as an engineering reproducibility benchmark, not a trading-performance
claim.
Public report path
Post the drafted report through the main repository's dedicated template:
https://github.com/initial-d/ml-quant-trading/issues/new?template=deepseek_harness_benchmark.yml
Seed example:
https://github.com/initial-d/ml-quant-trading/issues/61
Independent DSH runs are tracked in the plugin challenge issue:
https://github.com/initial-d/dsh-plugin-mlquant-benchmark/issues/1
Post there if the plugin failed before a valid benchmark-v1.json artifact was
created. Post successful or caveated benchmark reports through the main
repository template above.
For context and agent-facing guardrails, read the main repository's
DeepSeek Harness Recipe
and
Quant Agent Reproducibility Target.
Development
npm install
npm test
The test loads the plugin with a mock ctx.tools.register, verifies that the
four tools register, reads and validates sample artifacts, and drafts an issue
body.
Non-goals
- No investment advice.
- No backtest-performance claim.
- No hidden model provider configuration.
- No posting to GitHub from the tool.
- No private data or API keys in artifacts.
链接
同类插件
Tencent/WeKnora#dsh-weknora★ 37918
把 WeKnora 知识库接入 dsh 的四个只读工具:列出知识库、混合检索原文片段、按顺序还原单篇文档,以及直接取用 WeKnora 自己带引用的 RAG 或 ReAct agent 回答(含可续聊的 session id)。
DayuanJiang/next-ai-draw-io#dsh-plugin★ 37097
dsh 的 draw.io 绘图插件:通过 next-ai-drawio MCP 服务器作图,点回复里的链接在右侧栏打开实时预览,可以手动修改图,AI 的每次修改都是可以恢复的版本。需 Node.js 18+,首次使用由 npx 下载 MCP 服务器。
superdesigndev/treg★ 8650
给 Agent 的工具目录:按「要做的事」检索约 2,600 个外部接口(SEO 与 SERP、外链、社交、人物与公司信息补全、广告库、抓取),查看参数与单次调用价格后直接调用,凭据由服务端注入。附带技能,MCP 行在未设置 TREG_TOKEN 前保持禁用。
T-Auto/dsh-ops★ 5964
为 Windows 的 DSH 提供 bash、PowerShell 7 及 Rust 高性能工具,减少 token 消耗。
sixtysevenlf/dsh-blender-plugin★ 2216
让 DSH agent 通过直连 TCP 通道驱动本地 Blender,提供 15 个工具:视口帧、自定义视角渲染、Python 执行、无头与作业卸载、事务回滚,以及 34 个 family / 228 个 op 的规划层。
helloHub1212/dsh-windows-mcp★ 2002
通过官方 MCP 客户端桥接层把 CursorTouch/Windows-MCP 桌面自动化服务器接入 DSH,把截屏、UI 树快照、鼠标键盘输入、剪贴板、进程与注册表等工具注册成 mcp__windows__* 供模型调用,并附带 windows_mcp_doctor 诊断工具;Computer Use 默认关闭,需在插件面板里打开。
社区评论
评论公开保存在 GitHub Discussions。加载评论会连接 GitHub 和 Giscus;发表内容需要 GitHub 账号。