dspy-go

Resources

Resources

Comprehensive collection of tools, examples, and learning materials for dspy-go.

Official Resources

Documentation

Tools

Examples

Core Examples

Quick Start

Advanced Features

Multimodal

  • Multimodal Processing - Image analysis and vision Q&A
    • Image analysis with questions
    • Vision question answering
    • Multimodal chat
    • Streaming multimodal content
    • Multiple image comparison

Optimizers

Agent Patterns

  • Agent Examples - Various agent implementations
    • ReAct pattern
    • Orchestrator pattern
    • Memory management

Production Applications

Maestro - Code Review Agent

GitHub: XiaoConstantine/maestro

A production code review and question-answering agent built with dspy-go. Demonstrates:

  • RAG pipeline implementation
  • Tool integration (MCP)
  • Smart tool registry usage
  • Production deployment patterns

Key Features

  • 🔍 Automated code review
  • 💬 Natural language Q&A about codebases
  • 🔧 MCP tool integration
  • 📊 Performance optimization with MIPRO

Learning Materials

Video Tutorials

Coming soon! Check the GitHub repository for announcements.

Blog Posts & Articles

  • Introduction to DSPy - Original DSPy paper
  • Building LLM Applications with Go (coming soon)
  • Prompt Optimization Strategies (coming soon)

Community Examples

Check GitHub Discussions for community-contributed examples and patterns.

Datasets

Built-in Dataset Support

dspy-go includes automatic downloading and management for popular datasets:

GSM8K - Grade School Math

import "github.com/XiaoConstantine/dspy-go/pkg/datasets"

// Automatically downloads if not present
gsm8kPath, err := datasets.EnsureDataset("gsm8k")
dataset, err := datasets.LoadGSM8K(gsm8kPath)

Use for: Math reasoning, chain-of-thought, optimization

HotPotQA - Multi-hop Question Answering

hotpotPath, err := datasets.EnsureDataset("hotpotqa")
dataset, err := datasets.LoadHotPotQA(hotpotPath)

Use for: Multi-step reasoning, RAG pipelines, complex Q&A

Custom Datasets

// Create in-memory dataset
dataset := datasets.NewInMemoryDataset()
dataset.AddExample(map[string]interface{}{
    "question": "What is the capital of France?",
    "answer": "Paris",
})

LLM Provider Setup

Google Gemini

export GEMINI_API_KEY="your-api-key"
  • ✅ Multimodal support (images)
  • ✅ Fast responses
  • ✅ Good for prototyping

Get API Key →

OpenAI

export OPENAI_API_KEY="your-api-key"
  • ✅ GPT-4, GPT-3.5
  • ✅ Reliable performance
  • ✅ Well-documented

Get API Key →

Anthropic Claude

export ANTHROPIC_API_KEY="your-api-key"
  • ✅ Long context windows
  • ✅ Strong reasoning
  • ✅ Safety features

Get API Key →

Ollama (Local)

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama2

# Set base URL
export OLLAMA_BASE_URL="http://localhost:11434"
  • ✅ Free, local execution
  • ✅ Privacy (no data leaves your machine)
  • ✅ No API costs

Install Ollama →

Development Tools

IDE Extensions

Testing & Debugging

// Enable detailed tracing
ctx = core.WithExecutionState(context.Background())

// Execute your program
result, err := program.Execute(ctx, inputs)

// Inspect trace
executionState := core.GetExecutionState(ctx)
steps := executionState.GetSteps("moduleId")
for _, step := range steps {
    fmt.Printf("Duration: %s\n", step.Duration)
    fmt.Printf("Prompt: %s\n", step.Prompt)
    fmt.Printf("Response: %s\n", step.Response)
}

Monitoring

import "github.com/XiaoConstantine/dspy-go/pkg/logging"

// Configure logging
logger := logging.NewLogger(logging.Config{
    Severity: logging.DEBUG,
    Outputs:  []logging.Output{logging.NewConsoleOutput(true)},
})
logging.SetLogger(logger)

Community & Support

Get Help

Contributing

Stay Updated

MCP (Model Context Protocol)

DSPy Ecosystem

Benchmarks & Performance

Compatibility Results

dspy-go maintains compatibility with Python DSPy implementations. See:

Performance Metrics

  • Parallel processing can improve throughput by 3-4x
  • Smart Tool Registry adds < 50ms overhead
  • Optimization times vary by optimizer (see Optimizers Guide)

Tips & Best Practices

Getting Started

  1. ✅ Start with BootstrapFewShot optimizer
  2. ✅ Use clear, detailed field descriptions
  3. ✅ Test with small datasets first
  4. ✅ Enable logging during development

Production Readiness

  1. ✅ Use train/validation splits
  2. ✅ Monitor performance metrics
  3. ✅ Implement error handling
  4. ✅ Cache results where possible
  5. ✅ Use Parallel module for batches

Optimization

  1. ✅ Aim for 50+ training examples
  2. ✅ Balance dataset (don’t skew toward one class)
  3. ✅ Start simple, then optimize
  4. ✅ Validate on held-out data

Next Steps

Prev
LLM Providers