import os import gradio as gr from llmriddles.questions import QuestionExecutor from llmriddles.questions import list_ordered_questions _QUESTION_ID = -1 _QUESTIONS = list_ordered_questions() _LANG = os.environ.get('QUESTION_LANG', 'cn') _LLM = os.environ.get('QUESTION_LLM', 'chatgpt') def _need_api_key(): return _LLM == 'chatgpt' def _get_api_key_cfgs(api_key): if _LLM == 'chatgpt': return {'api_key': api_key} else: return {} if __name__ == '__main__': with gr.Blocks() as demo: with gr.Row(): with gr.Column(): gr_requirement = gr.TextArea(placeholder='Click \'Next\' to Start', label='Requirements') gr_question = gr.TextArea(placeholder='Your Question for LLM', label='Question') gr_answer = gr.TextArea(placeholder='Answer From LLM', label='Answer') gr_submit = gr.Button('Submit', interactive=False) with gr.Column(): gr_api_key = gr.Text(placeholder='Your API Key', label='API Key', type='password', visible=_need_api_key()) gr_predict = gr.Label(label='Correctness') gr_explanation = gr.TextArea(label='Explanation') gr_next = gr.Button('Next') def _next_question(): global _QUESTION_ID _QUESTION_ID += 1 if _QUESTION_ID >= len(_QUESTIONS): return 'Congratulations!', '', '', {}, '', \ gr.Button('Submit', interactive=False), \ gr.Button('Next', interactive=False) else: executor = QuestionExecutor(_QUESTIONS[_QUESTION_ID], _LANG) return executor.question_text, '', '', {}, '', \ gr.Button('Submit', interactive=True), \ gr.Button('Next', interactive=False) gr_next.click( fn=_next_question, inputs=[], outputs=[gr_requirement, gr_question, gr_answer, gr_predict, gr_explanation, gr_submit, gr_next], ) def _submit_answer(qs_text: str, api_key: str): if _need_api_key() and not api_key: return '---', {}, 'Please Enter API Key Before Submitting Question.', \ gr.Button('Next', interactive=False) executor = QuestionExecutor( _QUESTIONS[_QUESTION_ID], _LANG, llm=_LLM, llm_cfgs=_get_api_key_cfgs(api_key) if _need_api_key() else {} ) answer_text, correctness, explanation = executor.check(qs_text) labels = {'Correct': 1.0} if correctness else {'Wrong': 1.0} if correctness: return answer_text, labels, explanation, gr.Button('Next', interactive=True) else: return answer_text, labels, explanation, gr.Button('Next', interactive=False) gr_submit.click( _submit_answer, inputs=[gr_question, gr_api_key], outputs=[gr_answer, gr_predict, gr_explanation, gr_next], ) demo.launch()