Unknown · siet Oktober 2023 (Paper/erste DSPy-Version); ursprünglicher Vorläufer DSP Dezember 2022 · 2× · tolest 27. Sept. 2026
84
Momentum
DSPy (Declarative Self-improving Python) is an open-source framework developed by Stanford NLP for programming language models declaratively instead of manual prompt engineering. Developers define signatures (input/output specifications) and modules in Python; optimizers automatically tune prompts and few-shot examples against a chosen metric. It supports virtually any LLM provider via its LiteLLM integration and is typically installed via pip for self-hosted or custom pipeline use. The foundational paper was published in October 2023, and the project continues to be actively developed under the MIT license.
Momentum-Verloop
29.06.27.09.
Features
| License | MIT License |
| Platform | Python framework/library (pip package 'dspy') |
| Price | Free, open source (no license fee) |
| Protocol Compatibility | OpenAI- and Anthropic-compatible HTTP endpoints via LiteLLM, plus MCP compatibility |
| Release Date | October 2023 (foundational paper 'DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines') |
| Supported Models/Providers | Any LLM provider via LiteLLM (incl. OpenAI, Anthropic, Google, Azure, AWS Bedrock, Ollama, local models) |