dspy icon indicating copy to clipboard operation
dspy copied to clipboard

DSPy: The framework for programming—not prompting—language models

Results 691 dspy issues
Sort by recently updated
recently updated
newest added

Hello I have this code ```python import dspy backend_llm = dspy.OllamaLocal( model='llama3', # The model we're going to use. ) dspy.configure(lm=backend_llm) class MySig(dspy.Signature): """ You are a binary classifier You'll...

Given that Llama Index has some of the most wide and robust support for various retrieval abstractions and methods, adding an interface from LlamaIndex into DSPy would allow many to...

When the kwargs of the model get altered for the body, we don't want the new kwargs saved to the original llm object. This fixes a bug caused by the...

Hello, I just wanted to check when the next PyPI release is planned for? There are some more recent changes (e.g., GoogleVertexAI module) that it would be great to have...

## Description This PR adds an implementation of `BaseSentenceVectorizer` using [FastEmbed](https://qdrant.github.io/fastembed/) - A lightweight, CPU-first library to generate embeddings locally. This PR will be followed by https://github.com/Anush008/dspy/pull/1, which updates `QdrantRM`...

Here are the changes proposed here 1. Colbert as a local retriever 2. Colbert as a local reranker 3. Not treating reranking as first-class citizen, rather added RetrieveThenRerank apart from...

DSPy misses newlines, especially with longer context and smaller models. So is there any possibility to enforce newline character treatment in the few-shot in-context learning examples.

Add MyScaleRM to support retrieval for MyScaleDB: - MyScaleDB: https://github.com/myscale/myscaledb

I'd like to request some additional information on how to write custom tools for the ReAct module in DSPy. I've seen a few examples, but I've been struggling to wrap...