dspy-development

name	dspy-development
description	This skill should be used when the user asks to "build with DSPy", "create a DSPy module", "optimize prompts", "build a RAG system", "create an AI agent with DSPy", "use declarative LM programming", "build an LM pipeline", "optimize few-shot examples", "use teleprompters", "compile a DSPy program", "fine-tune prompts", "create a DSPy signature", or mentions any of: DSPy, dspy, import dspy, dspy.LM, dspy.configure, dspy.Predict, dspy.ChainOfThought, dspy.ReAct, dspy.Module, dspy.Signature, dspy.InputField, dspy.OutputField, dspy.Retrieve, dspy.TypedPredictor, dspy.ProgramOfThought, dspy.Refine, dspy.BestofN, dspy.asyncify, dspy.streamify, dspy.MIPROv2, BootstrapFewShot, BootstrapFinetune, MIPRO, COPRO, dspy.Evaluate, dspy.Example, dspy.Prediction, dspy.RLM, dspy.RLM(, recursive language model, RLM module, long context processing, programmatic context exploration, REPL-based reasoning, llm_query, context decomposition, dspy.CodeAct, dspy.GEPA, dspy.SIMBA, ArborGRPO, dspy.Reasoning, dspy.Image, dspy.Audio, ChatAdapter, JSONAdapter, XMLAdapter, BAMLAdapter, Module.batch, dspy.adapter, dspy.teleprompt, dspy.evaluate, prompt optimization, few-shot learning, language model pipeline, LM programming framework, Stanford NLP DSPy, declarative prompting, or automated prompt engineering. Also trigger when code contains `import dspy` or `from dspy`.
version	3.0.0
title	Dspy Development
status	active

DSPy: Declarative Language Model Programming

Framework for programming — not prompting — language models. Build modular AI systems with automatic prompt optimization, RL training, and reflective prompt evolution.

GitHub: 22,000+ stars | By: Stanford NLP | Current: DSPy 3.1+ (Feb 2026)

Installation

pip install dspy                    # Stable release (3.1.3)
pip install dspy[all]               # All LM providers
pip install git+https://github.com/stanfordnlp/dspy.git  # Latest dev

For RLM/ProgramOfThought/CodeAct (sandboxed code execution):

# Install Deno runtime (required for WASM sandbox)
curl -fsSL https://deno.land/install.sh | sh

Python: 3.10+ required (3.9 dropped in 3.0)

Quick Start

import dspy

# Configure LM (unified API — works with any provider)
lm = dspy.LM("anthropic/claude-sonnet-4-5-20250929", max_tokens=1000)
dspy.configure(lm=lm)

# Define a signature (input -> output contract)
class QA(dspy.Signature):
    """Answer questions with short factual answers."""
    question = dspy.InputField()
    answer = dspy.OutputField(desc="often between 1 and 5 words")

# Create and use a module
qa = dspy.Predict(QA)
result = qa(question="What is the capital of France?")
print(result.answer)  # "Paris"

Chain of Thought

cot = dspy.ChainOfThought("question -> answer")
result = cot(question="If John has 5 apples and gives 2 to Mary, how many remain?")
print(result.rationale)  # Step-by-step reasoning
print(result.answer)     # "3"

RLM — Recursive Language Model (3.1+)

Process documents far beyond context window limits via programmatic exploration: