Eval driven development

Build LLM apps with evals up front. 100% open source.

samples.py

my_task.py

@task
def my_agent():
  return Task(
    solver=[
      prompt(),
      chain_of_thought(),
      fewshot_examples(),
      generate()
    ],
    dataset=samples(),
    scorer=match()
  )

@task
def my_agent():
    return Task(
        solver=[
            prompt(),
            chain_of_thought(),
            fewshot_examples(),
            generate()
        ],
        dataset=samples(),
        scorer=match()
    )