| # + tags=["hide_inp"] | |
| desc = """ | |
| # Math | |
| Notebook to answer a math problem with code. | |
| Adapted from Dust [maths-generate-code](https://dust.tt/spolu/a/d12ac33169) | |
| """ | |
| # - | |
| import minichain | |
| # Prompt that asks LLM for code from math. | |
| class MathPrompt(minichain.TemplatePrompt[str]): | |
| template_file = "math.pmpt.tpl" | |
| # Ask a question and run it as python code. | |
| with minichain.start_chain("math") as backend: | |
| math_prompt = MathPrompt(backend.OpenAI()) | |
| code_prompt = minichain.SimplePrompt(backend.Python()) | |
| prompt = math_prompt.chain(code_prompt) | |
| # result = prompt({"question": question}) | |
| # print(result) | |
| gradio = prompt.to_gradio(fields =["question"], | |
| examples=["What is the sum of the powers of 3 (3^i) that are smaller than 100?"], | |
| out_type="markdown", | |
| description=desc | |
| ) | |
| if __name__ == "__main__": | |
| gradio.launch() | |
| # View the prompt | |
| # + tags=["hide_inp"] | |
| # MathPrompt().show({"question": "What is 10 + 12?"}, "10 + 12") | |
| # # - | |
| # # View the log | |
| # minichain.show_log("math.log") | |