# DSPy — Program LLMs Instead of Prompting

> DSPy is a Python framework for programming language models instead of prompting them. 33.3K+ GitHub stars. Build modular AI systems — classifiers, RAG pipelines, agent loops — and let DSPy optimize pr

## Install

Save as a script file and run:

## Quick Use

```bash
# Install
pip install dspy

# Quick start
python -c "
import dspy
lm = dspy.LM('openai/gpt-4o-mini')
dspy.configure(lm=lm)
print(dspy.ChainOfThought('question -> answer')(question='What is DSPy?'))
"
```

---

## Intro

DSPy is a Python framework by Stanford NLP for programming language models instead of writing brittle prompts. With 33,300+ GitHub stars and MIT license, DSPy lets you build modular AI systems — classifiers, RAG pipelines, agent loops — using compositional Python code. Instead of manually crafting prompts, DSPy compiles your program and algorithmically optimizes the prompts and weights for high-quality outputs. It supports self-improving pipelines, in-context learning with demonstrations, and works with any LLM provider.

**Best for**: AI engineers building production LLM pipelines who want reliable, optimized outputs without prompt engineering
**Works with**: Claude Code, OpenAI Codex, Cursor, Gemini CLI, Windsurf
**Providers**: OpenAI, Anthropic, Google, local models via Ollama

---

## Key Features

- **Programming over prompting**: Write Python modules instead of fragile prompt strings
- **Automatic optimization**: DSPy compiles and optimizes prompts and weights algorithmically
- **Modular design**: Compose classifiers, retrieval, reasoning, and agent loops as reusable modules
- **Self-improving pipelines**: Compilation learns from examples to improve output quality
- **Provider agnostic**: Works with OpenAI, Anthropic, Google, local models, and any LLM API
- **Built-in modules**: ChainOfThought, ReAct, RAG, multi-hop reasoning out of the box

---

## Example

```python
import dspy

# Configure LLM
lm = dspy.LM('openai/gpt-4o-mini')
dspy.configure(lm=lm)

# Define a module
classify = dspy.ChainOfThought('text -> sentiment: positive, negative, neutral')

# Use it
result = classify(text="DSPy is amazing for building reliable AI systems")
print(result.sentiment)  # positive

# Optimize with examples
optimizer = dspy.MIPROv2(metric=my_metric, auto="medium")
optimized = optimizer.compile(classify, trainset=examples)
```

---

### FAQ

**Q: What is DSPy?**
A: DSPy is a Stanford NLP framework with 33.3K+ stars for programming language models instead of prompting them. You write Python modules and DSPy automatically optimizes the prompts and weights for reliable, high-quality outputs. MIT licensed.

**Q: How do I install DSPy?**
A: Run `pip install dspy`. Then configure your LLM provider with `dspy.configure(lm=dspy.LM('openai/gpt-4o-mini'))` and start building modules.

**Q: How is DSPy different from LangChain?**
A: LangChain chains together prompts and tools manually. DSPy compiles and optimizes prompts algorithmically — you write what you want, DSPy figures out the best prompt automatically. DSPy is more focused on reliability and optimization.

---

## Source & Thanks

> Created by [Stanford NLP](https://github.com/stanfordnlp). Licensed under MIT.
> [stanfordnlp/dspy](https://github.com/stanfordnlp/dspy) — 33,300+ GitHub stars

---
Source: https://tokrepo.com/en/workflows/88a2e60f-54d1-48cb-9173-23d56a4f3f20
Author: Script Depot