dspy-skills MCP — 使用场景, 安装 & 实时演示

为什么要用

核心特性

DSPy signatures / modules / predictors encoded as patterns
Teleprompter (BootstrapFewShot, MIPRO) usage guidance
Evaluation + metric design
RAG pipeline scaffolds with DSPy retrievers
Opinionated integration with common LLM providers

实时演示

实际使用效果

dspy-skill.replay ▶ 就绪

0/0

安装

选择你的客户端

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "dspy-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OmidZamani/dspy-skills",
        "~/.claude/skills/dspy-skills"
      ],
      "_inferred": true
    }
  }
}

打开 Claude Desktop → Settings → Developer → Edit Config。保存后重启应用。

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "dspy-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OmidZamani/dspy-skills",
        "~/.claude/skills/dspy-skills"
      ],
      "_inferred": true
    }
  }
}

Cursor 使用与 Claude Desktop 相同的 mcpServers 格式。项目级配置优先于全局。

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "dspy-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OmidZamani/dspy-skills",
        "~/.claude/skills/dspy-skills"
      ],
      "_inferred": true
    }
  }
}

点击 Cline 侧栏中的 MCP Servers 图标，然后选 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "dspy-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OmidZamani/dspy-skills",
        "~/.claude/skills/dspy-skills"
      ],
      "_inferred": true
    }
  }
}

格式与 Claude Desktop 相同。重启 Windsurf 生效。

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "dspy-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OmidZamani/dspy-skills",
        "~/.claude/skills/dspy-skills"
      ]
    }
  ]
}

Continue 使用服务器对象数组，而非映射。

~/.config/zed/settings.json

{
  "context_servers": {
    "dspy-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/OmidZamani/dspy-skills",
          "~/.claude/skills/dspy-skills"
        ]
      }
    }
  }
}

加入 context_servers。Zed 保存后热重载。

claude mcp add dspy-skill -- git clone https://github.com/OmidZamani/dspy-skills ~/.claude/skills/dspy-skills

一行命令搞定。用 claude mcp list 验证，claude mcp remove 卸载。

使用场景

实战用法： dspy-skills

How to build your first DSPy program and optimize it

👤 ML engineers and applied researchers ⏱ ~90 min advanced

何时使用： You have a task where prompt quality matters and want a systematic way to improve it.

前置条件

Python 3.10+ with dspy-ai installed — pip install dspy-ai
Skill cloned — git clone https://github.com/OmidZamani/dspy-skills ~/.claude/skills/dspy-skills

步骤

Define the signature

Task: classify support tickets into {billing, technical, account}. Give me a DSPy signature and a simple Predict module.✓ 已复制

→ Signature + module code
Write an eval

Add an evaluation set of 50 labeled examples and an accuracy metric.✓ 已复制

→ Eval harness with metric callable
Optimize

Run BootstrapFewShot to compile the module against the eval set.✓ 已复制

→ Compiled predictor + improved score

结果： A DSPy-optimized predictor that beats a hand-written prompt, with reproducible code.

注意事项

Eval too small — optimizer overfits — Minimum 100–200 examples; hold out a true test set
Metric doesn't capture what you care about — Spend on metric design before on model choice

搭配使用： filesystem

Build a RAG pipeline with DSPy

👤 Engineers building retrieval-augmented systems ⏱ ~120 min advanced

何时使用： You want modular, optimizable RAG rather than a hand-wired chain.

步骤

Define the modules

Create a DSPy RAG pipeline: RetrieveThenRead with ColBERTv2 or a local retriever.✓ 已复制

→ Modular pipeline with separate retrieval and generation
Optimize end-to-end

Write an eval on our QA set and run MIPRO to improve.✓ 已复制

→ Compiled pipeline with score delta

结果： A RAG pipeline you can improve by changing the eval, not by rewriting prompts.

注意事项

Retriever quality caps end-to-end quality — Evaluate retrieval separately (recall@k) before optimizing generation

搭配使用： local-rag

组合

与其他 MCP 搭配，撬动十倍杠杆

dspy-skill + local-rag

Plug a local retriever into DSPy's RAG modules

Swap the ColBERT retriever for a local-rag MCP as the retrieval source.✓ 已复制

dspy-skill + filesystem

Organize DSPy programs, evals, and artifacts in a repo

Lay out a DSPy project with programs/, evals/, and artifacts/ directories.✓ 已复制

工具

此 MCP 暴露的能力

工具	输入参数	何时调用	成本
signature-design	task spec	Start of any DSPy program	0
module-authoring	signatures + flow	After signatures	0
teleprompter-optimization	module + eval	After eval is ready	LLM tokens during optimization
evaluation-harness	task data	Before optimizing	0

成本与限制

运行它的成本

API 配额: LLM tokens dominate during optimization runs
每次调用 Token 数: Can be high — optimizations may invoke the LLM hundreds of times
费用: depends on provider
提示: Use cheap models during teleprompter runs, upgrade only for final evaluation

安全

权限、密钥、影响范围

凭据存储： LLM provider keys in env vars

数据出站： LLM provider endpoints

故障排查

常见错误与修复

Teleprompter seems to make things worse

Check metric correctness; use a held-out set; widen the example pool.

Optimization eats your budget

Cap max_bootstrapped_demos and use cheaper models during search.

替代方案

dspy-skills 对比其他方案

替代方案	何时用它替代	权衡
prompt-architect-skill	You want prompt-level craft, not DSPy's programmatic approach	Hand-crafted vs optimized

dspy-skills

为什么要用

核心特性

实时演示

实际使用效果

安装

选择你的客户端

使用场景

实战用法： dspy-skills

How to build your first DSPy program and optimize it

前置条件

步骤

注意事项

Build a RAG pipeline with DSPy

步骤

注意事项

组合

与其他 MCP 搭配，撬动十倍杠杆

工具

此 MCP 暴露的能力

成本与限制

运行它的成本

安全

权限、密钥、影响范围

故障排查

常见错误与修复

替代方案

dspy-skills 对比其他方案

更多

资源