Nyraxis AI

DSPy

Wrap DSPy language model calls with Nyraxis to enforce policies on all predictions.

Setup

pip install dspy-ai nyraxis
export NYRAXIS_API_KEY="your-api-key"
export NYRAXIS_PROJECT_ID="your-project-id"

Code Example

import dspy
from nyraxis import Nyraxis

nx = Nyraxis()

lm = dspy.LM("openai/gpt-4")
dspy.configure(lm=lm)

class GuardedQA(dspy.Module):
    def __init__(self):
        self.generate = dspy.ChainOfThought("question -> answer")

    def forward(self, question):
        result = nx.evaluate(question)
        if result.blocked:
            return dspy.Prediction(answer="Request blocked by policy.")

        prediction = self.generate(question=question)

        output_result = nx.evaluate(prediction.answer, direction="output")
        if output_result.blocked:
            return dspy.Prediction(answer="Response blocked by policy.")

        return prediction

qa = GuardedQA()
response = qa(question="What is machine learning?")
print(response.answer)

What Gets Protected

  • All input questions are evaluated before being sent to the language model
  • Generated outputs are checked against policies before being returned
  • Works with any DSPy module (ChainOfThought, ReAct, Predict)
  • Policies apply consistently across compiled and uncompiled programs

On this page