stanfordnlp/dspy

This system is a Python-based framework for building and managing Large Language Model (LLM) applications, specifically focusing on programmatic control of model behavior. It provides a unified API for defining, executing, and evaluating LLM-based programs, including features for structured output formatting, tool calling, and multi-step reasoning. The codebase has been significantly refactored to remove legacy components, introducing a modern, flat public API and a robust testing infrastructure to ensure reliability across different model providers.

This period focused on modernizing the project's development workflow and significantly refactoring the public API. Legacy modules, caching mechanisms, and internal primitives were removed in favor of a flattened, unified interface with improved adapter support.

10 changes

This period focused on establishing a robust testing infrastructure for the DSPy codebase, introducing extensive unit and integration tests across all major modules. The work covered core primitives, prediction modules, evaluation metrics, and client integrations, while also adding reliability testing frameworks and test utilities for external dependencies like LiteLLM.

12 changes

This period focused on significantly expanding the test suite across the codebase, with a particular emphasis on new types, adapters, and core components. The work included adding comprehensive tests for language model types, retrievers, and proposers to ensure robustness and correct behavior.

5 changes

stanfordnlp/dspy was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point the surveyor at a repository you know and see whether you agree with it.