DSPy

DSPy 3.1.0 officially releases the beta, adding new adapters, file handling, reasoning support, Python 3.14 compatibility, and several optimizer and security enhancements, while fixing numerous bugs and renaming the settings API.

DSPy 3.1.0b1 introduces several new features such as a File type, Reasoning capture, and Python 3.14 support, along with numerous bug fixes and security hardening, while renaming configuration helpers.

DSPy 3.0.4 adds extensive GEPA enhancements, new Anthropic citation support, tool execution improvements, and numerous bug fixes while deprecating Image.from_* helpers.

This release adds new tooling APIs, improves networking headers, and introduces several bug fixes and documentation updates.

DSPy 3.0.4b1 adds extensive GEPA enhancements, multimodal support, and new documentation, while introducing breaking changes to DummyLM, EvaluationResult location, and deprecating Image.from_* helpers.

dspy 3.0.3 adds rollout_id support, improves GEPA optimizer, updates caching behavior, and includes several bug fixes and maintenance updates.

This 3.0.2 release adds support for custom chunk types, OpenAI response API, and gpt-5‑nano reasoning models, while fixing mlflow‑GEPA integration, BAMLAdapter issues, and an error‑message typo. It also introduces deprecation warnings for requires_permission_to_run and adds type hints for Module call methods.

Version 3.0.1 introduces ConfigDict-based config, lazy gepa imports, improved error handling, flexible ToolCalls parsing, and several bug fixes.

DSPy 3.0 introduces powerful new optimizers, extensible adapters and types, async execution, and native MLflow 3.0 observability, while removing community retrievers, dropping Python 3.9 support, and cleaning up legacy APIs.

This release adds several optimizer and adapter improvements, introduces new features like a custom BAMLAdapter and a gpt-5 reasoning model flag, and includes numerous bug fixes and maintenance updates.

Version 3.0.0b3 restores the `datasets` dependency, adds several new features such as `dspy.Code`, token streaming for XMLAdapter, and `dspy.syncify`, and includes numerous bug fixes and API refinements.

DSPy 3.0.0b2 introduces new LM providers, async adapters, and extensive cleanup, but also removes several legacy APIs and drops Python 3.9 support, requiring code updates.

This release introduces several bug fixes, style standardization, and new features like a global max_errors setting. The main breaking change is the removal of unmaintained retriever integrations.

Version 2.6.27 fixes the BaseType annotation parsing issue.

This patch fixes the BaseType annotation parsing issue.

This release adds support for dspy.Tool and dspy.ToolCall field types and includes bug fixes for inspect_history and linting issues.

DSPy 2.6.25 introduces several new features such as a custom Signature base class, Audio handling, CodeAct module, and Python 3.13 support, along with numerous bug fixes and improvements.

DSPy 2.6.24 introduces the GRPO optimizer, new AdapterParseError, sync streaming improvements, and several bug fixes and API refinements.

Version 2.6.23 adds async streaming support, token streaming with JSON, and MIPROv2 optimizer updates while fixing usage tracking, argument errors, and litellm retries.

Version 2.6.22 adds async support for ReAct and mcp tools, introduces caching for async LM calls, improves adapter handling, and fixes several thread‑safety and validation bugs.

The release relaxes the input handling of Adapters.parse_value and Cache.request_cache, making them more permissive.

Version 2.6.20 adds native async support for callbacks and the dspy.Tool, introduces dspy.Tool.from_mcp_tool, and includes several bug fixes and dependency updates.

DSPy 2.6.19 adds async support, a fanout cache, and several enhancements to ReAct and the Tool API, while fixing Pydantic warnings and migrating the build system to UV.

Version 2.6.18 adds a TwoStepAdapter, default args for tools, streaming support, and several bug fixes, while moving key settings into a global configuration and relaxing the OpenAI version constraint.

DSPy 2.6.17 introduces extensible adapter refactoring, MIPROv2 optimizer defaults, and exponential backoff retries, while fixing several adapter and optimizer bugs.

dspy 2.6.16 restores Python 3.9 support, fixes the numpy dependency, adds usage tracking and trial logs, and includes several bug fixes.

Version 2.6.15 adds the experimental dspy.SIMBA optimizer, makes dspy.ChainOfThought more customizable, and allows dspy.ProgramOfThought to handle multiple output fields with added reliability improvements.

This release adds context protocol support and refinements to `PythonInterpreter`, introduces a new `construct_result_table` method, and includes a fix to avoid crashes during cache initialization, while temporarily limiting the OpenAI version and updating LiteLLM for security.

Version 2.6.13 introduces several enhancements such as extensible BaseLM, improved adapters with Pydantic support, and new features for BestOfN and Teleprompter, along with a bug fix for pickle deserialization.

dspy 2.6.12 improves error handling, simplifies the LM interface, adds callback metadata to evaluation, and updates dependencies.

dspy 2.6.11 introduces Python 3.13 support, adds timeout-based straggler resubmission to ParallelExecutor, reduces import memory usage, and includes several bug fixes.

Version 2.6.10 adds non-image MIME support for image URLs, refactors ProgramOfThought, switches LM checks to BaseLM, improves adapter type hints, and fixes a GitHub Actions variable issue.

This release adds multiturn history support, an evaluate callback, and the ability to pass `provide_traceback` in mipro, while fixing several bugs related to signature ordering, model saving, finetune naming, packaging, and a removed dependency.

Version 2.6.8 fixes the bundle files packaged with the dspy installation.

Version 2.6.7 resolves a Windows uvloop RuntimeError and bundles necessary files with the dspy installation.

Version 2.6.6 adds Refine and BestOfN modules, fixes thread safety for ChainOfThoughtWithHint, and moves packaging to pyproject.toml.

Version 2.6.5 introduces Weaviate multitenancy support, a signature overhaul, and fixes finetune bugs.

Version 2.6.4 improves the Adapters parse_value functionality and resolves several bugs and typos.

Version 2.6.3 fixes a tagging mistake and adds several enhancements including status streaming, context truncation for react, and a refactor of streamify to use context.

Version 2.6.2 adds the InferRules optimizer, supports arbitrary DSPy.Images examples, and introduces a pytest option to skip reliability tests by default.

dspy 2.6.1 adds several optimizer and caching enhancements, expands model support, and includes bug fixes for ReAct tools and cache handling.

This release removes deprecated functional and dsp clients, adds streaming, a sandboxed interpreter, retry support, and several feature enhancements while fixing literal handling bugs.

This release adds trusted library imports for the PythonInterpreter sandbox, enhances argument parsing for dspy.ReAct, improves literal handling in adapters, and includes several bug fixes and documentation updates.