# VizPy Optimizers - llms.txt
# https://llmstxt.org

# VizPy is an LLM prompt optimization library built on top of DSPy.
# It provides two optimizers: PromptGrad and ContraPrompt.

## About
VizPy Optimizers is a drop-in replacement for DSPy optimizers that reduces
prompt failure rates by turning errors into executable rules. It outperforms
GEPA on every benchmark including BBH, HotPotQA, GPQA Diamond, and GDPR-Bench.

## Products
- PromptGrad: Epoch-based failure analysis using textual gradients. Best for generation tasks and continuous metrics.
- ContraPrompt: Contrastive pair mining to extract rules from wrong vs correct attempts. Best for classification tasks.

## Pricing
- Free: 10 optimization runs
- Pro: 200 optimization runs/month ($20/month)
- Enterprise: 1,000 optimization runs/month ($200/month)

## Links
- Homepage: https://vizpy.vizops.ai
- Documentation: https://vizpy-docs.vizops.ai
- Company: https://vizops.ai
- PromptGrad blog: https://vizops.ai/blog/promptgrad-gradient-descent-for-prompts/
- ContraPrompt blog: https://vizops.ai/blog/contraprompt-learning-from-mistakes/

## Usage
VizPy works as a drop-in replacement for DSPy optimizers:

  import dspy
  import vizpy

  dspy.configure(lm=dspy.LM("your_provider/model"))

  # For classification tasks
  optimizer = vizpy.ContraPromptOptimizer(metric=my_metric)

  # For generation tasks
  optimizer = vizpy.PromptGradOptimizer(metric=my_metric)

  optimized = optimizer.optimize(module, train_examples)