Build declarative language-model programs and optimize prompts for RAG pipelines.
npx skills add https://github.com/afel6/scal-ai-pipeline --skill dspy-afel6
Please help me install this Agent Skill. Skill: dspy Source: https://github.com/afel6/scal-ai-pipeline/tree/main/hermes_skills_library/mlops/research/dspy Command: npx skills add https://github.com/afel6/scal-ai-pipeline --skill dspy-afel6
dspy removes brittle manual prompt engineering and ad-hoc LM calls by providing a declarative, modular programming model for language models that makes pipelines reproducible, testable, and optimizable.
Use dspy to declare a QA signature, wire a retriever and a ChainOfThought generator, and run a question through the pipeline to get a concise, validated answer.
High-intent search queries and answers about installing and using this skill.
To build RAG and agent pipelines, define typed input/output signatures and compose ChainOfThought or ReAct modules. This declarative approach replaces brittle manual prompt engineering with reproducible, testable language model programs for multi-stage question answering.
Automatic prompt optimization improves language model program performance using built-in optimizers like BootstrapFewShot, MIPRO, and BootstrapFinetune. These optimizers apply few-shot and instruction tuning against custom metrics, removing the need for ad-hoc manual prompt adjustments.
Multi-hop RAG systems are created by wiring a retriever and a ChainOfThought generator into a declarative pipeline. The system generates search queries, retrieves supporting documents, reranks passages, and produces concise, validated answers for research or support.
Yes, production ML pipelines are supported with multi-provider language model integration. Optimized declarative modules can be exported for production use, ensuring retrieval-augmented generation and agentic tool use workflows remain reproducible and scalable.
The best way to replace manual prompt engineering is using declarative signatures and modular programming for language models. This makes pipelines reproducible and testable, applying automatic optimizers like MIPRO and BootstrapFewShot instead of ad-hoc prompt tuning.
Declarative programming for language models may not suit projects requiring single, isolated ad-hoc LM calls without optimization. It is designed for modular pipelines, structured signatures, and multi-stage workflows needing automatic few-shot or instruction tuning.
ruiyangruiyiCompile declarative DSPy program signatures into optimized prompt and few-shot pipelines.
VYRE-StudiosBuild declarative AI pipelines for language model tasks using Python and dspy.
Dianshu-LiaoBuild retrieval-augmented generation systems with citation-ready outputs from external documents.
KarlinskySCompose declarative LM pipelines with optimizers like BootstrapFewShot and MIPRO.
