[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6D32G9WF5EB210HT9WWSQYX":3,"skill-same:01M6D32G9WF5EB210HT9WWSQYX":103},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":26,"publishedAt":47,"linkOnly":48,"markdownContent":49,"markdownBody":50,"frontMatter":51,"refName":52,"commitSha":53,"skillDirectory":54,"forksCount":55,"repositoryHtmlUrl":56,"repositoryHomepageUrl":6,"ownerHtmlUrl":57,"ownerBlogUrl":58,"ownerDescription":59,"updatedAt":60,"install":61,"resources":74},"01M6D32G9WF5EB210HT9WWSQYX","dspy-ruby",null,"This skill should be used when working with DSPy.rb, a Ruby framework for building type-safe, composable LLM applications. Use this when implementing predictable AI features, creating LLM signatures and modules, configuring language model providers (OpenAI, Anthropic, Gemini, Ollama), building agent systems with tools, optimizing prompts, or testing LLM-powered functionality in Ruby applications.","hemory\u002Famp-legacy-do-not-use","hemory","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F21179881?v=4","01M6D31HFJ391DN8N0JP8AQGFE","Hemory Phifer","01M6D31HKXVTEN27AT6R4PP2ZJ",3,"MIT","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"dspy","ruby","dspy ruby","ollama","systems","signatures","language","building agent systems","Optimizing Prompts","modules","prompts","use","testing","model","applications","Tools","openai","composable","gemini","safe","2026-11-10 10:12:30",false,"---\nname: dspy-ruby\ndescription: This skill should be used when working with DSPy.rb, a Ruby framework for building type-safe, composable LLM applications. Use this when implementing predictable AI features, creating LLM signatures and modules, configuring language model providers (OpenAI, Anthropic, Gemini, Ollama), building agent systems with tools, optimizing prompts, or testing LLM-powered functionality in Ruby applications.\n---\n\n# DSPy.rb Expert\n\n## Overview\n\nDSPy.rb is a Ruby framework that enables developers to **program LLMs, not prompt them**. Instead of manually crafting prompts, define application requirements through type-safe, composable modules that can be tested, optimized, and version-controlled like regular code.\n\nThis skill provides comprehensive guidance on:\n- Creating type-safe signatures for LLM operations\n- Building composable modules and workflows\n- Configuring multiple LLM providers\n- Implementing agents with tools\n- Testing and optimizing LLM applications\n- Production deployment patterns\n\n## Core Capabilities\n\n### 1. Type-Safe Signatures\n\nCreate input\u002Foutput contracts for LLM operations with runtime type checking.\n\n**When to use**: Defining any LLM task, from simple classification to complex analysis.\n\n**Quick reference**:\n```ruby\nclass EmailClassificationSignature \u003C DSPy::Signature\n  description \"Classify customer support emails\"\n\n  input do\n    const :email_subject, String\n    const :email_body, String\n  end\n\n  output do\n    const :category, T.enum([\"Technical\", \"Billing\", \"General\"])\n    const :priority, T.enum([\"Low\", \"Medium\", \"High\"])\n  end\nend\n```\n\n**Templates**: See `assets\u002Fsignature-template.rb` for comprehensive examples including:\n- Basic signatures with multiple field types\n- Vision signatures for multimodal tasks\n- Sentiment analysis signatures\n- Code generation signatures\n\n**Best practices**:\n- Always provide clear, specific descriptions\n- Use enums for constrained outputs\n- Include field descriptions with `desc:` parameter\n- Prefer specific types over generic String when possible\n\n**Full documentation**: See `references\u002Fcore-concepts.md` sections on Signatures and Type Safety.\n\n### 2. Composable Modules\n\nBuild reusable, chainable modules that encapsulate LLM operations.\n\n**When to use**: Implementing any LLM-powered feature, especially complex multi-step workflows.\n\n**Quick reference**:\n```ruby\nclass EmailProcessor \u003C DSPy::Module\n  def initialize\n    super\n    @classifier = DSPy::Predict.new(EmailClassificationSignature)\n  end\n\n  def forward(email_subject:, email_body:)\n    @classifier.forward(\n      email_subject: email_subject,\n      email_body: email_body\n    )\n  end\nend\n```\n\n**Templates**: See `assets\u002Fmodule-template.rb` for comprehensive examples including:\n- Basic modules with single predictors\n- Multi-step pipelines that chain modules\n- Modules with conditional logic\n- Error handling and retry patterns\n- Stateful modules with history\n- Caching implementations\n\n**Module composition**: Chain modules together to create complex workflows:\n```ruby\nclass Pipeline \u003C DSPy::Module\n  def initialize\n    super\n    @step1 = Classifier.new\n    @step2 = Analyzer.new\n    @step3 = Responder.new\n  end\n\n  def forward(input)\n    result1 = @step1.forward(input)\n    result2 = @step2.forward(result1)\n    @step3.forward(result2)\n  end\nend\n```\n\n**Full documentation**: See `references\u002Fcore-concepts.md` sections on Modules and Module Composition.\n\n### 3. Multiple Predictor Types\n\nChoose the right predictor for your task:\n\n**Predict**: Basic LLM inference with type-safe inputs\u002Foutputs\n```ruby\npredictor = DSPy::Predict.new(TaskSignature)\nresult = predictor.forward(input: \"data\")\n```\n\n**ChainOfThought**: Adds automatic reasoning for improved accuracy\n```ruby\npredictor = DSPy::ChainOfThought.new(TaskSignature)\nresult = predictor.forward(input: \"data\")\n# Returns: { reasoning: \"...\", output: \"...\" }\n```\n\n**ReAct**: Tool-using agents with iterative reasoning\n```ruby\npredictor = DSPy::ReAct.new(\n  TaskSignature,\n  tools: [SearchTool.new, CalculatorTool.new],\n  max_iterations: 5\n)\n```\n\n**CodeAct**: Dynamic code generation (requires `dspy-code_act` gem)\n```ruby\npredictor = DSPy::CodeAct.new(TaskSignature)\nresult = predictor.forward(task: \"Calculate factorial of 5\")\n```\n\n**When to use each**:\n- **Predict**: Simple tasks, classification, extraction\n- **ChainOfThought**: Complex reasoning, analysis, multi-step thinking\n- **ReAct**: Tasks requiring external tools (search, calculation, API calls)\n- **CodeAct**: Tasks best solved with generated code\n\n**Full documentation**: See `references\u002Fcore-concepts.md` section on Predictors.\n\n### 4. LLM Provider Configuration\n\nSupport for OpenAI, Anthropic Claude, Google Gemini, Ollama, and OpenRouter.\n\n**Quick configuration examples**:\n```ruby\n# OpenAI\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n\n# Anthropic Claude\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('anthropic\u002Fclaude-3-5-sonnet-20241022',\n    api_key: ENV['ANTHROPIC_API_KEY'])\nend\n\n# Google Gemini\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('gemini\u002Fgemini-1.5-pro',\n    api_key: ENV['GOOGLE_API_KEY'])\nend\n\n# Local Ollama (free, private)\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('ollama\u002Fllama3.1')\nend\n```\n\n**Templates**: See `assets\u002Fconfig-template.rb` for comprehensive examples including:\n- Environment-based configuration\n- Multi-model setups for different tasks\n- Configuration with observability (OpenTelemetry, Langfuse)\n- Retry logic and fallback strategies\n- Budget tracking\n- Rails initializer patterns\n\n**Provider compatibility matrix**:\n\n| Feature | OpenAI | Anthropic | Gemini | Ollama |\n|---------|--------|-----------|--------|--------|\n| Structured Output | ✅ | ✅ | ✅ | ✅ |\n| Vision (Images) | ✅ | ✅ | ✅ | ⚠️ Limited |\n| Image URLs | ✅ | ❌ | ❌ | ❌ |\n| Tool Calling | ✅ | ✅ | ✅ | Varies |\n\n**Cost optimization strategy**:\n- Development: Ollama (free) or gpt-4o-mini (cheap)\n- Testing: gpt-4o-mini with temperature=0.0\n- Production simple tasks: gpt-4o-mini, claude-3-haiku, gemini-1.5-flash\n- Production complex tasks: gpt-4o, claude-3-5-sonnet, gemini-1.5-pro\n\n**Full documentation**: See `references\u002Fproviders.md` for all configuration options, provider-specific features, and troubleshooting.\n\n### 5. Multimodal & Vision Support\n\nProcess images alongside text using the unified `DSPy::Image` interface.\n\n**Quick reference**:\n```ruby\nclass VisionSignature \u003C DSPy::Signature\n  description \"Analyze image and answer questions\"\n\n  input do\n    const :image, DSPy::Image\n    const :question, String\n  end\n\n  output do\n    const :answer, String\n  end\nend\n\npredictor = DSPy::Predict.new(VisionSignature)\nresult = predictor.forward(\n  image: DSPy::Image.from_file(\"path\u002Fto\u002Fimage.jpg\"),\n  question: \"What objects are visible?\"\n)\n```\n\n**Image loading methods**:\n```ruby\n# From file\nDSPy::Image.from_file(\"path\u002Fto\u002Fimage.jpg\")\n\n# From URL (OpenAI only)\nDSPy::Image.from_url(\"https:\u002F\u002Fexample.com\u002Fimage.jpg\")\n\n# From base64\nDSPy::Image.from_base64(base64_data, mime_type: \"image\u002Fjpeg\")\n```\n\n**Provider support**:\n- OpenAI: Full support including URLs\n- Anthropic, Gemini: Base64 or file loading only\n- Ollama: Limited multimodal depending on model\n\n**Full documentation**: See `references\u002Fcore-concepts.md` section on Multimodal Support.\n\n### 6. Testing LLM Applications\n\nWrite standard RSpec tests for LLM logic.\n\n**Quick reference**:\n```ruby\nRSpec.describe EmailClassifier do\n  before do\n    DSPy.configure do |c|\n      c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n        api_key: ENV['OPENAI_API_KEY'])\n    end\n  end\n\n  it 'classifies technical emails correctly' do\n    classifier = EmailClassifier.new\n    result = classifier.forward(\n      email_subject: \"Can't log in\",\n      email_body: \"Unable to access account\"\n    )\n\n    expect(result[:category]).to eq('Technical')\n    expect(result[:priority]).to be_in(['High', 'Medium', 'Low'])\n  end\nend\n```\n\n**Testing patterns**:\n- Mock LLM responses for unit tests\n- Use VCR for deterministic API testing\n- Test type safety and validation\n- Test edge cases (empty inputs, special characters, long texts)\n- Integration test complete workflows\n\n**Full documentation**: See `references\u002Foptimization.md` section on Testing.\n\n### 7. Optimization & Improvement\n\nAutomatically improve prompts and modules using optimization techniques.\n\n**MIPROv2 optimization**:\n```ruby\nrequire 'dspy\u002Fmipro'\n\n# Define evaluation metric\ndef accuracy_metric(example, prediction)\n  example[:expected_output][:category] == prediction[:category] ? 1.0 : 0.0\nend\n\n# Prepare training data\ntraining_examples = [\n  {\n    input: { email_subject: \"...\", email_body: \"...\" },\n    expected_output: { category: 'Technical' }\n  },\n  # More examples...\n]\n\n# Run optimization\noptimizer = DSPy::MIPROv2.new(\n  metric: method(:accuracy_metric),\n  num_candidates: 10\n)\n\noptimized_module = optimizer.compile(\n  EmailClassifier.new,\n  trainset: training_examples\n)\n```\n\n**A\u002FB testing different approaches**:\n```ruby\n# Test ChainOfThought vs ReAct\napproach_a_score = evaluate_approach(ChainOfThoughtModule, test_set)\napproach_b_score = evaluate_approach(ReActModule, test_set)\n```\n\n**Full documentation**: See `references\u002Foptimization.md` section on Optimization.\n\n### 8. Observability & Monitoring\n\nTrack performance, token usage, and behavior in production.\n\n**OpenTelemetry integration**:\n```ruby\nrequire 'opentelemetry\u002Fsdk'\n\nOpenTelemetry::SDK.configure do |c|\n  c.service_name = 'my-dspy-app'\n  c.use_all\nend\n\n# DSPy automatically creates traces\n```\n\n**Langfuse tracing**:\n```ruby\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\n\n  c.langfuse = {\n    public_key: ENV['LANGFUSE_PUBLIC_KEY'],\n    secret_key: ENV['LANGFUSE_SECRET_KEY']\n  }\nend\n```\n\n**Custom monitoring**:\n- Token tracking\n- Performance monitoring\n- Error rate tracking\n- Custom logging\n\n**Full documentation**: See `references\u002Foptimization.md` section on Observability.\n\n## Quick Start Workflow\n\n### For New Projects\n\n1. **Install DSPy.rb and provider gems**:\n```bash\ngem install dspy dspy-openai  # or dspy-anthropic, dspy-gemini\n```\n\n2. **Configure LLM provider** (see `assets\u002Fconfig-template.rb`):\n```ruby\nrequire 'dspy'\n\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n```\n\n3. **Create a signature** (see `assets\u002Fsignature-template.rb`):\n```ruby\nclass MySignature \u003C DSPy::Signature\n  description \"Clear description of task\"\n\n  input do\n    const :input_field, String, desc: \"Description\"\n  end\n\n  output do\n    const :output_field, String, desc: \"Description\"\n  end\nend\n```\n\n4. **Create a module** (see `assets\u002Fmodule-template.rb`):\n```ruby\nclass MyModule \u003C DSPy::Module\n  def initialize\n    super\n    @predictor = DSPy::Predict.new(MySignature)\n  end\n\n  def forward(input_field:)\n    @predictor.forward(input_field: input_field)\n  end\nend\n```\n\n5. **Use the module**:\n```ruby\nmodule_instance = MyModule.new\nresult = module_instance.forward(input_field: \"test\")\nputs result[:output_field]\n```\n\n6. **Add tests** (see `references\u002Foptimization.md`):\n```ruby\nRSpec.describe MyModule do\n  it 'produces expected output' do\n    result = MyModule.new.forward(input_field: \"test\")\n    expect(result[:output_field]).to be_a(String)\n  end\nend\n```\n\n### For Rails Applications\n\n1. **Add to Gemfile**:\n```ruby\ngem 'dspy'\ngem 'dspy-openai'  # or other provider\n```\n\n2. **Create initializer** at `config\u002Finitializers\u002Fdspy.rb` (see `assets\u002Fconfig-template.rb` for full example):\n```ruby\nrequire 'dspy'\n\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n```\n\n3. **Create modules in** `app\u002Fllm\u002F` directory:\n```ruby\n# app\u002Fllm\u002Femail_classifier.rb\nclass EmailClassifier \u003C DSPy::Module\n  # Implementation here\nend\n```\n\n4. **Use in controllers\u002Fservices**:\n```ruby\nclass EmailsController \u003C ApplicationController\n  def classify\n    classifier = EmailClassifier.new\n    result = classifier.forward(\n      email_subject: params[:subject],\n      email_body: params[:body]\n    )\n    render json: result\n  end\nend\n```\n\n## Common Patterns\n\n### Pattern: Multi-Step Analysis Pipeline\n\n```ruby\nclass AnalysisPipeline \u003C DSPy::Module\n  def initialize\n    super\n    @extract = DSPy::Predict.new(ExtractSignature)\n    @analyze = DSPy::ChainOfThought.new(AnalyzeSignature)\n    @summarize = DSPy::Predict.new(SummarizeSignature)\n  end\n\n  def forward(text:)\n    extracted = @extract.forward(text: text)\n    analyzed = @analyze.forward(data: extracted[:data])\n    @summarize.forward(analysis: analyzed[:result])\n  end\nend\n```\n\n### Pattern: Agent with Tools\n\n```ruby\nclass ResearchAgent \u003C DSPy::Module\n  def initialize\n    super\n    @agent = DSPy::ReAct.new(\n      ResearchSignature,\n      tools: [\n        WebSearchTool.new,\n        DatabaseQueryTool.new,\n        SummarizerTool.new\n      ],\n      max_iterations: 10\n    )\n  end\n\n  def forward(question:)\n    @agent.forward(question: question)\n  end\nend\n\nclass WebSearchTool \u003C DSPy::Tool\n  def call(query:)\n    results = perform_search(query)\n    { results: results }\n  end\nend\n```\n\n### Pattern: Conditional Routing\n\n```ruby\nclass SmartRouter \u003C DSPy::Module\n  def initialize\n    super\n    @classifier = DSPy::Predict.new(ClassifySignature)\n    @simple_handler = SimpleModule.new\n    @complex_handler = ComplexModule.new\n  end\n\n  def forward(input:)\n    classification = @classifier.forward(text: input)\n\n    if classification[:complexity] == 'Simple'\n      @simple_handler.forward(input: input)\n    else\n      @complex_handler.forward(input: input)\n    end\n  end\nend\n```\n\n### Pattern: Retry with Fallback\n\n```ruby\nclass RobustModule \u003C DSPy::Module\n  MAX_RETRIES = 3\n\n  def forward(input, retry_count: 0)\n    begin\n      @predictor.forward(input)\n    rescue DSPy::ValidationError => e\n      if retry_count \u003C MAX_RETRIES\n        sleep(2 ** retry_count)\n        forward(input, retry_count: retry_count + 1)\n      else\n        # Fallback to default or raise\n        raise\n      end\n    end\n  end\nend\n```\n\n## Resources\n\nThis skill includes comprehensive reference materials and templates:\n\n### References (load as needed for detailed information)\n\n- [core-concepts.md](.\u002Freferences\u002Fcore-concepts.md): Complete guide to signatures, modules, predictors, multimodal support, and best practices\n- [providers.md](.\u002Freferences\u002Fproviders.md): All LLM provider configurations, compatibility matrix, cost optimization, and troubleshooting\n- [optimization.md](.\u002Freferences\u002Foptimization.md): Testing patterns, optimization techniques, observability setup, and monitoring\n\n### Assets (templates for quick starts)\n\n- [signature-template.rb](.\u002Fassets\u002Fsignature-template.rb): Examples of signatures including basic, vision, sentiment analysis, and code generation\n- [module-template.rb](.\u002Fassets\u002Fmodule-template.rb): Module patterns including pipelines, agents, error handling, caching, and state management\n- [config-template.rb](.\u002Fassets\u002Fconfig-template.rb): Configuration examples for all providers, environments, observability, and production patterns\n\n## When to Use This Skill\n\nTrigger this skill when:\n- Implementing LLM-powered features in Ruby applications\n- Creating type-safe interfaces for AI operations\n- Building agent systems with tool usage\n- Setting up or troubleshooting LLM providers\n- Optimizing prompts and improving accuracy\n- Testing LLM functionality\n- Adding observability to AI applications\n- Converting from manual prompt engineering to programmatic approach\n- Debugging DSPy.rb code or configuration issues\n","# DSPy.rb Expert\n\n## Overview\n\nDSPy.rb is a Ruby framework that enables developers to **program LLMs, not prompt them**. Instead of manually crafting prompts, define application requirements through type-safe, composable modules that can be tested, optimized, and version-controlled like regular code.\n\nThis skill provides comprehensive guidance on:\n- Creating type-safe signatures for LLM operations\n- Building composable modules and workflows\n- Configuring multiple LLM providers\n- Implementing agents with tools\n- Testing and optimizing LLM applications\n- Production deployment patterns\n\n## Core Capabilities\n\n### 1. Type-Safe Signatures\n\nCreate input\u002Foutput contracts for LLM operations with runtime type checking.\n\n**When to use**: Defining any LLM task, from simple classification to complex analysis.\n\n**Quick reference**:\n```ruby\nclass EmailClassificationSignature \u003C DSPy::Signature\n  description \"Classify customer support emails\"\n\n  input do\n    const :email_subject, String\n    const :email_body, String\n  end\n\n  output do\n    const :category, T.enum([\"Technical\", \"Billing\", \"General\"])\n    const :priority, T.enum([\"Low\", \"Medium\", \"High\"])\n  end\nend\n```\n\n**Templates**: See `assets\u002Fsignature-template.rb` for comprehensive examples including:\n- Basic signatures with multiple field types\n- Vision signatures for multimodal tasks\n- Sentiment analysis signatures\n- Code generation signatures\n\n**Best practices**:\n- Always provide clear, specific descriptions\n- Use enums for constrained outputs\n- Include field descriptions with `desc:` parameter\n- Prefer specific types over generic String when possible\n\n**Full documentation**: See `references\u002Fcore-concepts.md` sections on Signatures and Type Safety.\n\n### 2. Composable Modules\n\nBuild reusable, chainable modules that encapsulate LLM operations.\n\n**When to use**: Implementing any LLM-powered feature, especially complex multi-step workflows.\n\n**Quick reference**:\n```ruby\nclass EmailProcessor \u003C DSPy::Module\n  def initialize\n    super\n    @classifier = DSPy::Predict.new(EmailClassificationSignature)\n  end\n\n  def forward(email_subject:, email_body:)\n    @classifier.forward(\n      email_subject: email_subject,\n      email_body: email_body\n    )\n  end\nend\n```\n\n**Templates**: See `assets\u002Fmodule-template.rb` for comprehensive examples including:\n- Basic modules with single predictors\n- Multi-step pipelines that chain modules\n- Modules with conditional logic\n- Error handling and retry patterns\n- Stateful modules with history\n- Caching implementations\n\n**Module composition**: Chain modules together to create complex workflows:\n```ruby\nclass Pipeline \u003C DSPy::Module\n  def initialize\n    super\n    @step1 = Classifier.new\n    @step2 = Analyzer.new\n    @step3 = Responder.new\n  end\n\n  def forward(input)\n    result1 = @step1.forward(input)\n    result2 = @step2.forward(result1)\n    @step3.forward(result2)\n  end\nend\n```\n\n**Full documentation**: See `references\u002Fcore-concepts.md` sections on Modules and Module Composition.\n\n### 3. Multiple Predictor Types\n\nChoose the right predictor for your task:\n\n**Predict**: Basic LLM inference with type-safe inputs\u002Foutputs\n```ruby\npredictor = DSPy::Predict.new(TaskSignature)\nresult = predictor.forward(input: \"data\")\n```\n\n**ChainOfThought**: Adds automatic reasoning for improved accuracy\n```ruby\npredictor = DSPy::ChainOfThought.new(TaskSignature)\nresult = predictor.forward(input: \"data\")\n# Returns: { reasoning: \"...\", output: \"...\" }\n```\n\n**ReAct**: Tool-using agents with iterative reasoning\n```ruby\npredictor = DSPy::ReAct.new(\n  TaskSignature,\n  tools: [SearchTool.new, CalculatorTool.new],\n  max_iterations: 5\n)\n```\n\n**CodeAct**: Dynamic code generation (requires `dspy-code_act` gem)\n```ruby\npredictor = DSPy::CodeAct.new(TaskSignature)\nresult = predictor.forward(task: \"Calculate factorial of 5\")\n```\n\n**When to use each**:\n- **Predict**: Simple tasks, classification, extraction\n- **ChainOfThought**: Complex reasoning, analysis, multi-step thinking\n- **ReAct**: Tasks requiring external tools (search, calculation, API calls)\n- **CodeAct**: Tasks best solved with generated code\n\n**Full documentation**: See `references\u002Fcore-concepts.md` section on Predictors.\n\n### 4. LLM Provider Configuration\n\nSupport for OpenAI, Anthropic Claude, Google Gemini, Ollama, and OpenRouter.\n\n**Quick configuration examples**:\n```ruby\n# OpenAI\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n\n# Anthropic Claude\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('anthropic\u002Fclaude-3-5-sonnet-20241022',\n    api_key: ENV['ANTHROPIC_API_KEY'])\nend\n\n# Google Gemini\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('gemini\u002Fgemini-1.5-pro',\n    api_key: ENV['GOOGLE_API_KEY'])\nend\n\n# Local Ollama (free, private)\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('ollama\u002Fllama3.1')\nend\n```\n\n**Templates**: See `assets\u002Fconfig-template.rb` for comprehensive examples including:\n- Environment-based configuration\n- Multi-model setups for different tasks\n- Configuration with observability (OpenTelemetry, Langfuse)\n- Retry logic and fallback strategies\n- Budget tracking\n- Rails initializer patterns\n\n**Provider compatibility matrix**:\n\n| Feature | OpenAI | Anthropic | Gemini | Ollama |\n|---------|--------|-----------|--------|--------|\n| Structured Output | ✅ | ✅ | ✅ | ✅ |\n| Vision (Images) | ✅ | ✅ | ✅ | ⚠️ Limited |\n| Image URLs | ✅ | ❌ | ❌ | ❌ |\n| Tool Calling | ✅ | ✅ | ✅ | Varies |\n\n**Cost optimization strategy**:\n- Development: Ollama (free) or gpt-4o-mini (cheap)\n- Testing: gpt-4o-mini with temperature=0.0\n- Production simple tasks: gpt-4o-mini, claude-3-haiku, gemini-1.5-flash\n- Production complex tasks: gpt-4o, claude-3-5-sonnet, gemini-1.5-pro\n\n**Full documentation**: See `references\u002Fproviders.md` for all configuration options, provider-specific features, and troubleshooting.\n\n### 5. Multimodal & Vision Support\n\nProcess images alongside text using the unified `DSPy::Image` interface.\n\n**Quick reference**:\n```ruby\nclass VisionSignature \u003C DSPy::Signature\n  description \"Analyze image and answer questions\"\n\n  input do\n    const :image, DSPy::Image\n    const :question, String\n  end\n\n  output do\n    const :answer, String\n  end\nend\n\npredictor = DSPy::Predict.new(VisionSignature)\nresult = predictor.forward(\n  image: DSPy::Image.from_file(\"path\u002Fto\u002Fimage.jpg\"),\n  question: \"What objects are visible?\"\n)\n```\n\n**Image loading methods**:\n```ruby\n# From file\nDSPy::Image.from_file(\"path\u002Fto\u002Fimage.jpg\")\n\n# From URL (OpenAI only)\nDSPy::Image.from_url(\"https:\u002F\u002Fexample.com\u002Fimage.jpg\")\n\n# From base64\nDSPy::Image.from_base64(base64_data, mime_type: \"image\u002Fjpeg\")\n```\n\n**Provider support**:\n- OpenAI: Full support including URLs\n- Anthropic, Gemini: Base64 or file loading only\n- Ollama: Limited multimodal depending on model\n\n**Full documentation**: See `references\u002Fcore-concepts.md` section on Multimodal Support.\n\n### 6. Testing LLM Applications\n\nWrite standard RSpec tests for LLM logic.\n\n**Quick reference**:\n```ruby\nRSpec.describe EmailClassifier do\n  before do\n    DSPy.configure do |c|\n      c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n        api_key: ENV['OPENAI_API_KEY'])\n    end\n  end\n\n  it 'classifies technical emails correctly' do\n    classifier = EmailClassifier.new\n    result = classifier.forward(\n      email_subject: \"Can't log in\",\n      email_body: \"Unable to access account\"\n    )\n\n    expect(result[:category]).to eq('Technical')\n    expect(result[:priority]).to be_in(['High', 'Medium', 'Low'])\n  end\nend\n```\n\n**Testing patterns**:\n- Mock LLM responses for unit tests\n- Use VCR for deterministic API testing\n- Test type safety and validation\n- Test edge cases (empty inputs, special characters, long texts)\n- Integration test complete workflows\n\n**Full documentation**: See `references\u002Foptimization.md` section on Testing.\n\n### 7. Optimization & Improvement\n\nAutomatically improve prompts and modules using optimization techniques.\n\n**MIPROv2 optimization**:\n```ruby\nrequire 'dspy\u002Fmipro'\n\n# Define evaluation metric\ndef accuracy_metric(example, prediction)\n  example[:expected_output][:category] == prediction[:category] ? 1.0 : 0.0\nend\n\n# Prepare training data\ntraining_examples = [\n  {\n    input: { email_subject: \"...\", email_body: \"...\" },\n    expected_output: { category: 'Technical' }\n  },\n  # More examples...\n]\n\n# Run optimization\noptimizer = DSPy::MIPROv2.new(\n  metric: method(:accuracy_metric),\n  num_candidates: 10\n)\n\noptimized_module = optimizer.compile(\n  EmailClassifier.new,\n  trainset: training_examples\n)\n```\n\n**A\u002FB testing different approaches**:\n```ruby\n# Test ChainOfThought vs ReAct\napproach_a_score = evaluate_approach(ChainOfThoughtModule, test_set)\napproach_b_score = evaluate_approach(ReActModule, test_set)\n```\n\n**Full documentation**: See `references\u002Foptimization.md` section on Optimization.\n\n### 8. Observability & Monitoring\n\nTrack performance, token usage, and behavior in production.\n\n**OpenTelemetry integration**:\n```ruby\nrequire 'opentelemetry\u002Fsdk'\n\nOpenTelemetry::SDK.configure do |c|\n  c.service_name = 'my-dspy-app'\n  c.use_all\nend\n\n# DSPy automatically creates traces\n```\n\n**Langfuse tracing**:\n```ruby\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\n\n  c.langfuse = {\n    public_key: ENV['LANGFUSE_PUBLIC_KEY'],\n    secret_key: ENV['LANGFUSE_SECRET_KEY']\n  }\nend\n```\n\n**Custom monitoring**:\n- Token tracking\n- Performance monitoring\n- Error rate tracking\n- Custom logging\n\n**Full documentation**: See `references\u002Foptimization.md` section on Observability.\n\n## Quick Start Workflow\n\n### For New Projects\n\n1. **Install DSPy.rb and provider gems**:\n```bash\ngem install dspy dspy-openai  # or dspy-anthropic, dspy-gemini\n```\n\n2. **Configure LLM provider** (see `assets\u002Fconfig-template.rb`):\n```ruby\nrequire 'dspy'\n\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n```\n\n3. **Create a signature** (see `assets\u002Fsignature-template.rb`):\n```ruby\nclass MySignature \u003C DSPy::Signature\n  description \"Clear description of task\"\n\n  input do\n    const :input_field, String, desc: \"Description\"\n  end\n\n  output do\n    const :output_field, String, desc: \"Description\"\n  end\nend\n```\n\n4. **Create a module** (see `assets\u002Fmodule-template.rb`):\n```ruby\nclass MyModule \u003C DSPy::Module\n  def initialize\n    super\n    @predictor = DSPy::Predict.new(MySignature)\n  end\n\n  def forward(input_field:)\n    @predictor.forward(input_field: input_field)\n  end\nend\n```\n\n5. **Use the module**:\n```ruby\nmodule_instance = MyModule.new\nresult = module_instance.forward(input_field: \"test\")\nputs result[:output_field]\n```\n\n6. **Add tests** (see `references\u002Foptimization.md`):\n```ruby\nRSpec.describe MyModule do\n  it 'produces expected output' do\n    result = MyModule.new.forward(input_field: \"test\")\n    expect(result[:output_field]).to be_a(String)\n  end\nend\n```\n\n### For Rails Applications\n\n1. **Add to Gemfile**:\n```ruby\ngem 'dspy'\ngem 'dspy-openai'  # or other provider\n```\n\n2. **Create initializer** at `config\u002Finitializers\u002Fdspy.rb` (see `assets\u002Fconfig-template.rb` for full example):\n```ruby\nrequire 'dspy'\n\nDSPy.configure do |c|\n  c.lm = DSPy::LM.new('openai\u002Fgpt-4o-mini',\n    api_key: ENV['OPENAI_API_KEY'])\nend\n```\n\n3. **Create modules in** `app\u002Fllm\u002F` directory:\n```ruby\n# app\u002Fllm\u002Femail_classifier.rb\nclass EmailClassifier \u003C DSPy::Module\n  # Implementation here\nend\n```\n\n4. **Use in controllers\u002Fservices**:\n```ruby\nclass EmailsController \u003C ApplicationController\n  def classify\n    classifier = EmailClassifier.new\n    result = classifier.forward(\n      email_subject: params[:subject],\n      email_body: params[:body]\n    )\n    render json: result\n  end\nend\n```\n\n## Common Patterns\n\n### Pattern: Multi-Step Analysis Pipeline\n\n```ruby\nclass AnalysisPipeline \u003C DSPy::Module\n  def initialize\n    super\n    @extract = DSPy::Predict.new(ExtractSignature)\n    @analyze = DSPy::ChainOfThought.new(AnalyzeSignature)\n    @summarize = DSPy::Predict.new(SummarizeSignature)\n  end\n\n  def forward(text:)\n    extracted = @extract.forward(text: text)\n    analyzed = @analyze.forward(data: extracted[:data])\n    @summarize.forward(analysis: analyzed[:result])\n  end\nend\n```\n\n### Pattern: Agent with Tools\n\n```ruby\nclass ResearchAgent \u003C DSPy::Module\n  def initialize\n    super\n    @agent = DSPy::ReAct.new(\n      ResearchSignature,\n      tools: [\n        WebSearchTool.new,\n        DatabaseQueryTool.new,\n        SummarizerTool.new\n      ],\n      max_iterations: 10\n    )\n  end\n\n  def forward(question:)\n    @agent.forward(question: question)\n  end\nend\n\nclass WebSearchTool \u003C DSPy::Tool\n  def call(query:)\n    results = perform_search(query)\n    { results: results }\n  end\nend\n```\n\n### Pattern: Conditional Routing\n\n```ruby\nclass SmartRouter \u003C DSPy::Module\n  def initialize\n    super\n    @classifier = DSPy::Predict.new(ClassifySignature)\n    @simple_handler = SimpleModule.new\n    @complex_handler = ComplexModule.new\n  end\n\n  def forward(input:)\n    classification = @classifier.forward(text: input)\n\n    if classification[:complexity] == 'Simple'\n      @simple_handler.forward(input: input)\n    else\n      @complex_handler.forward(input: input)\n    end\n  end\nend\n```\n\n### Pattern: Retry with Fallback\n\n```ruby\nclass RobustModule \u003C DSPy::Module\n  MAX_RETRIES = 3\n\n  def forward(input, retry_count: 0)\n    begin\n      @predictor.forward(input)\n    rescue DSPy::ValidationError => e\n      if retry_count \u003C MAX_RETRIES\n        sleep(2 ** retry_count)\n        forward(input, retry_count: retry_count + 1)\n      else\n        # Fallback to default or raise\n        raise\n      end\n    end\n  end\nend\n```\n\n## Resources\n\nThis skill includes comprehensive reference materials and templates:\n\n### References (load as needed for detailed information)\n\n- [core-concepts.md](.\u002Freferences\u002Fcore-concepts.md): Complete guide to signatures, modules, predictors, multimodal support, and best practices\n- [providers.md](.\u002Freferences\u002Fproviders.md): All LLM provider configurations, compatibility matrix, cost optimization, and troubleshooting\n- [optimization.md](.\u002Freferences\u002Foptimization.md): Testing patterns, optimization techniques, observability setup, and monitoring\n\n### Assets (templates for quick starts)\n\n- [signature-template.rb](.\u002Fassets\u002Fsignature-template.rb): Examples of signatures including basic, vision, sentiment analysis, and code generation\n- [module-template.rb](.\u002Fassets\u002Fmodule-template.rb): Module patterns including pipelines, agents, error handling, caching, and state management\n- [config-template.rb](.\u002Fassets\u002Fconfig-template.rb): Configuration examples for all providers, environments, observability, and production patterns\n\n## When to Use This Skill\n\nTrigger this skill when:\n- Implementing LLM-powered features in Ruby applications\n- Creating type-safe interfaces for AI operations\n- Building agent systems with tool usage\n- Setting up or troubleshooting LLM providers\n- Optimizing prompts and improving accuracy\n- Testing LLM functionality\n- Adding observability to AI applications\n- Converting from manual prompt engineering to programmatic approach\n- Debugging DSPy.rb code or configuration issues\n",{"name":5,"description":7},"main","5417cad9215b767627c7c947dcb3a0bbe3ebe52d",".claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby",0,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use","https:\u002F\u002Fgithub.com\u002Fhemory","https:\u002F\u002Fwww.heminthehat.com\u002F","The People Developer | L&D Program Manager | Software Engineer","2026-06-04 23:46:30",{"source":8,"skillName":5,"commands":62,"prompt":72,"zipAvailable":73},[63,66,69],{"tool":64,"command":65},"npx","npx skills add hemory\u002Famp-legacy-do-not-use --skill dspy-ruby",{"tool":67,"command":68},"bunx","bunx skills add hemory\u002Famp-legacy-do-not-use --skill dspy-ruby",{"tool":70,"command":71},"pnpm","pnpm dlx skills add hemory\u002Famp-legacy-do-not-use --skill dspy-ruby","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002FSKILL.md",true,[75,81,85,89,95,99],{"path":76,"resourceType":77,"resourceTypeDesc":78,"sizeBytes":79,"sourceHtmlUrl":80},"assets\u002Fconfig-template.rb","SCRIPT","脚本",10321,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Fassets\u002Fconfig-template.rb",{"path":82,"resourceType":77,"resourceTypeDesc":78,"sizeBytes":83,"sourceHtmlUrl":84},"assets\u002Fmodule-template.rb",7664,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Fassets\u002Fmodule-template.rb",{"path":86,"resourceType":77,"resourceTypeDesc":78,"sizeBytes":87,"sourceHtmlUrl":88},"assets\u002Fsignature-template.rb",4838,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Fassets\u002Fsignature-template.rb",{"path":90,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":93,"sourceHtmlUrl":94},"references\u002Fcore-concepts.md","REFERENCE","参考文档",6301,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Freferences\u002Fcore-concepts.md",{"path":96,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":97,"sourceHtmlUrl":98},"references\u002Foptimization.md",14272,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Freferences\u002Foptimization.md",{"path":100,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":101,"sourceHtmlUrl":102},"references\u002Fproviders.md",8229,"https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fplugins\u002Fcompound-engineering\u002Fskills\u002Fdspy-ruby\u002Freferences\u002Fproviders.md",[104,137,146,160,175,187],{"skillNo":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":116,"publishedAt":47},"01M6D3DVEAHJTTK4W21B15ZZKK","zoom-setup","Connect Zoom to Amp for meeting recordings, scheduling, and transcript context","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fzoom-setup\u002FSKILL.md",[110,111],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":112,"categoryName":113,"taxonomyType":114,"taxonomyTypeDesc":115},"01JBSEEDCAT000000000000034","行政与办公人员","PROFESSION","适用职业",[117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136],"zoom setup","setup","schedule","MCP","review","connect","transcript","search","stats","oauth2","confirmation","context","meetings","process","recordings","surfaces","scheduling","granola","time","distribution",{"skillNo":138,"name":139,"nameZh":6,"description":140,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":141,"categories":142,"keywords":143,"publishedAt":47},"01M6D3DRAK8VWESYDZDF8JJDSH","xray","Understand what just happened under the hood - learn AI by seeing it in action","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fxray\u002FSKILL.md",[],[139,144,145],"learn","just",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":150,"categories":151,"keywords":153,"publishedAt":47},"01M6D3DN7BZQDY6143A2QSDP9A","week-review","Review week's progress with concrete accomplishments (not fake percentages), pattern detection, and goal tracking.","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fweek-review\u002FSKILL.md",[152],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[154,121,155,156,157,158,159],"week review","pattern detection","detection","tracking","goal","pattern",{"skillNo":161,"name":162,"nameZh":6,"description":163,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":164,"categories":165,"keywords":166,"publishedAt":47},"01M6D3DJ7F1TCFXBBSTQW82WE7","week-plan","Set weekly priorities and plan the week ahead with intelligent suggestions based on goals, calendar shape, and task effort.","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fweek-plan\u002FSKILL.md",[],[167,168,169,170,171,172,173,174],"week plan","plan","calendar","SET","shape","goals","task","priorities",{"skillNo":176,"name":177,"nameZh":6,"description":178,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":179,"categories":180,"keywords":181,"publishedAt":47},"01M6D3DF90D7RC8ZJE4AW2P9VC","triage","Strategically route orphaned files and extract scattered tasks","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Ftriage\u002FSKILL.md",[],[177,182,183,184,185,186],"tasks","fast","extract","files","route",{"skillNo":188,"name":189,"nameZh":6,"description":190,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":191,"categories":192,"keywords":197,"publishedAt":47},"01M6D3DC9BBHK5TASE34SQYXSQ","trello-setup","Connect Trello to Amp for visual Kanban task sync","https:\u002F\u002Fgithub.com\u002Fhemory\u002Famp-legacy-do-not-use\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Ftrello-setup\u002FSKILL.md",[193,196],{"categoryNo":194,"categoryName":195,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[198,199,118,200,122,120,201,173,202,203,204,205,206],"trello","trello setup","sync","token","kanban","bun","visual","api","server"]