DSPy

Program — not prompt — language models

What is DSPy?

DSPy from Stanford is a framework for programming LLMs with composable modules and optimizers that automatically tune prompts instead of hand-crafting them.

Why people choose DSPy

• Replaces prompt-hacking with optimization
• Composable, reusable modules
• Strong research backing

DSPy at a glance

Open-source alternatives to DSPy

Other open-source llm / rag framework tools worth comparing:

LangChainCompose chains, tools and agentsLlamaIndexThe data framework for RAGHaystackProduction pipelines for search and RAGtxtaiAll-in-one embeddings databaseRAGFlowDeep-document-understanding RAGSemantic KernelMicrosoft's enterprise agent frameworkGraphRAGRAG that builds a knowledge graph firstLiteLLMOne API for 100+ LLM providersInstructorReliable structured outputs from LLMsOutlinesGuarantee valid JSON and grammarsGuidanceInterleave control and generationLLMWareEnterprise RAG with small specialised modelsFlashRankUltra-light reranking for better RAGSentence TransformersThe standard way to make embeddingsRagasMeasure whether your RAG is any goodPromptfooUnit tests for your promptsLangfuseSee what your LLM app actually didPhoenixTrace, evaluate and debug LLM apps

DSPy head-to-head

FAQ