DSPy Advanced

Advanced DSPy modules for agents, tools, and output refinement.

Quick Start

ReAct Agent with Tools

import dspy

# Define tools
def get_weather(city: str) -> str:
    return f"The weather in {city} is sunny."

def search_web(query: str) -> str:
    return f"Search results for '{query}'"

# Create ReAct agent
react = dspy.ReAct(
    signature="question -> answer",
    tools=[get_weather, search_web],
    max_iters=5
)

# Use agent
result = react(question="What's the weather in Tokyo?")
print(result.answer)
print("Tool calls:", result.trajectory)

BestOfN for Multiple Attempts

def one_word_answer(args, pred: dspy.Prediction) -> float:
    return 1.0 if len(pred.answer.split()) == 1 else 0.0

best_of_3 = dspy.BestOfN(
    module=dspy.ChainOfThought("question -> answer"),
    N=3,
    reward_fn=one_word_answer,
    threshold=1.0
)

result = best_of_3(question="What is the capital of Belgium?")
print(result.answer)  # Brussels

Refine for Iterative Improvement

refine = dspy.Refine(
    module=dspy.ChainOfThought("question -> answer"),
    N=3,
    reward_fn=one_word_answer,
    threshold=1.0
)

result = refine(question="What is the capital of Belgium?")
print(result.answer)  # Brussels

When to Use This Skill

Use this skill when:

• •Building ReAct agents with tool calling
• •Implementing output refinement strategies
• •Integrating external tools and APIs
• •Using custom adapters for LM integration
• •Implementing multi-step reasoning with feedback loops

Core Concepts

ReAct Agents

ReAct combines reasoning and action, enabling agents to use tools to answer questions.

Key features:

• •Automatic reasoning before tool calls
• •Tool trajectory tracking
• •Multi-iteration tool usage
• •Dynamic tool selection

See: references/react-tools.md for:

• •ReAct agent patterns
• •Tool definition best practices
• •Multi-tool coordination

Output Refinement

Improve output quality through multiple attempts or iterative refinement.

Strategies:

• •BestOfN: Generate N candidates and select best
• •Refine: Iteratively improve with feedback loop

See: references/output-refinement.md for:

• •BestOfN vs Refine comparison
• •Reward function patterns
• •Use cases and examples

Adapters

Adapters bridge DSPy modules with Language Models, handling prompt formatting and output parsing.

Adapter types:

• •ChatAdapter: For chat-based LMs
• •JSONAdapter: For structured output
• •Custom adapters: For specialized use cases

See: references/adapters.md for:

• •Adapter architecture
• •Custom adapter implementation
• •Multi-provider configuration

Progressive Disclosure

This skill uses progressive disclosure:

1. •SKILL.md (this file): Quick reference and navigation
2. •references/: Detailed technical docs loaded as needed

Load reference files only when you need detailed information on a specific topic.

Related Skills

• •dspy-basics: Signature design, basic modules, program composition
• •dspy-optimization: Teleprompters, metrics, optimization workflows
• •dspy-configuration: LM setup, caching, and version management