import os import gradio as gr from openai import OpenAI import json import tempfile OPEN_AI_KEY = os.getenv("OPEN_AI_KEY") OPEN_AI_CLIENT = OpenAI(api_key=OPEN_AI_KEY) def generate_topics(model, max_tokens, sys_content, scenario, eng_level, user_generate_topics_prompt): """ 根据系统提示和用户输入的情境及主题，调用OpenAI API生成相关的主题句。 """ user_content = f""" scenario is {scenario} english level is {eng_level} {user_generate_topics_prompt} """ messages = [ {"role": "system", "content": sys_content}, {"role": "user", "content": user_content} ] request_payload = { "model": model, "messages": messages, "max_tokens": max_tokens, } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content.strip() return content def generate_points(model, max_tokens, sys_content, scenario, eng_level, topic, user_generate_points_prompt): """ 根据系统提示和用户输入的情境、主题，调用OpenAI API生成相关的主题句。 """ user_content = f""" scenario is {scenario} english level is {eng_level} topic is {topic} {user_generate_points_prompt} """ messages = [ {"role": "system", "content": sys_content}, {"role": "user", "content": user_content} ] request_payload = { "model": model, "messages": messages, "max_tokens": max_tokens, } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content.strip() return content def generate_topic_sentences(model, max_tokens, sys_content, scenario, eng_level, topic, points, user_generate_topic_sentences_prompt): """ 根据系统提示和用户输入的情境及要点，调用OpenAI API生成相关的主题句及其合理性解释。 """ user_content = f""" scenario is {scenario} english level is {eng_level} topic is {topic} points is {points} {user_generate_topic_sentences_prompt} """ messages = [ {"role": "system", "content": sys_content}, {"role": "user", "content": user_content} ] response_format = { "type": "json_object" } request_payload = { "model": model, "messages": messages, "max_tokens": max_tokens, "response_format": response_format } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) json_content = response.choices[0].message.content content = parse_and_display_topic_sentences(json_content) return content def parse_and_display_topic_sentences(json_data): """ 解析JSON格式的主题句数据，并转换成易于阅读的格式。 """ # 将JSON字符串解析成Python字典 data = json.loads(json_data) # 初始化一个空字符串用于存放最终的格式化文本 formatted_text = "" # 遍历每个主题句及其评价 for key, value in data.items(): topic_sentence = value[0]['topic-sentence'] appropriate = "適當" if value[0]['appropriate'] == "Y" else "不適當" reason = value[0]['reason'] # 将每个主题句的信息添加到格式化文本中 formatted_text += f"主题句 {int(key)+1}: {topic_sentence}\n" formatted_text += f"是否適當: {appropriate}\n" formatted_text += f"原因: {reason}\n\n" return formatted_text def generate_supporting_sentences(model, max_tokens, sys_content, scenario, eng_level, topic, points, topic_sentence, user_generate_supporting_sentences_prompt): """ 根据系统提示和用户输入的情境、主题、要点、主题句，调用OpenAI API生成相关的支持句。 """ user_content = f""" scenario is {scenario} english level is {eng_level} topic is {topic} points is {points} topic sentence is {topic_sentence} {user_generate_supporting_sentences_prompt} """ messages = [ {"role": "system", "content": sys_content}, {"role": "user", "content": user_content} ] request_payload = { "model": model, "messages": messages, "max_tokens": max_tokens, } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content.strip() return content def generate_conclusion_sentences(model, max_tokens, sys_content, scenario, eng_level, topic, points, topic_sentence, user_generate_conclusion_sentence_prompt): """ 根据系统提示和用户输入的情境、主题、要点、主题句，调用OpenAI API生成相关的结论句。 """ user_content = f""" scenario is {scenario} english level is {eng_level} topic is {topic} points is {points} topic sentence is {topic_sentence} {user_generate_conclusion_sentence_prompt} """ messages = [ {"role": "system", "content": sys_content}, {"role": "user", "content": user_content} ] request_payload = { "model": model, "messages": messages, "max_tokens": max_tokens, } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content.strip() return content def generate_paragraph(topic_sentence, supporting_sentences, conclusion_sentence): """ 根据用户输入的主题句、支持句、结论句，生成完整的段落。 """ paragraph = f"{topic_sentence}\n{supporting_sentences}\n{conclusion_sentence}" return paragraph def generate_paragraph_evaluate(paragraph, user_generate_paragraph_evaluate_prompt): """ 根据用户输入的段落，调用OpenAI API生成相关的段落分析。 """ user_content = f""" paragraph is {paragraph} {user_generate_paragraph_evaluate_prompt} """ messages = [ {"role": "system", "content": paragraph}, {"role": "user", "content": user_content} ] response_format = { "type": "json_object" } request_payload = { "model": "gpt-3.5-turbo", "messages": messages, "max_tokens": 500, "response_format": response_format } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content return content def generate_correct_grammatical_spelling_errors(eng_level, paragraph, user_correct_grammatical_spelling_errors_prompt): """ 根据用户输入的段落，调用OpenAI API生成相关的文法和拼字错误修正。 """ user_content = f""" level is {eng_level} paragraph is {paragraph} {user_correct_grammatical_spelling_errors_prompt} """ messages = [ {"role": "system", "content": paragraph}, {"role": "user", "content": user_content} ] response_format = { "type": "json_object" } request_payload = { "model": "gpt-3.5-turbo", "messages": messages, "max_tokens": 500, "response_format": response_format } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content return content def generate_refine_paragraph(eng_level, paragraph, user_refine_paragraph_prompt): """ 根据用户输入的段落，调用OpenAI API生成相关的段落改善建议。 """ user_content = f""" eng_level is {eng_level} paragraph is {paragraph} {user_refine_paragraph_prompt} """ messages = [ {"role": "system", "content": paragraph}, {"role": "user", "content": user_content} ] response_format = { "type": "json_object" } request_payload = { "model": "gpt-3.5-turbo", "messages": messages, "max_tokens": 500, "response_format": response_format } response = OPEN_AI_CLIENT.chat.completions.create(**request_payload) content = response.choices[0].message.content return content def paragraph_save_and_tts(paragraph_text): """ Saves the paragraph text and generates an audio file using OpenAI's TTS. """ try: # Call OpenAI's TTS API to generate speech from text response = OPEN_AI_CLIENT.audio.speech.create( model="tts-1", voice="alloy", input=paragraph_text, ) with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file: temp_file.write(response.content) # Get the file path of the temp file audio_path = temp_file.name # Return the path to the audio file along with the text return paragraph_text, audio_path except Exception as e: print(f"An error occurred while generating TTS: {e}") # Handle the error appropriately (e.g., return an error message or a default audio path) return paragraph_text, None with gr.Blocks() as demo: with gr.Row(): with gr.Column(): # basic inputs gr.Markdown("## 1. Basic Inputs") model = gr.Radio(["gpt-4-1106-preview", "gpt-3.5-turbo"], label="Model", value="gpt-4-1106-preview") max_tokens = gr.Slider(minimum=50, maximum=4000, value=1000, label="Max Tokens") sys_content_input = gr.Textbox(label="System Prompt", value="You are an English teacher who is practicing with me to improve my English writing skill.") scenario_input = gr.Textbox(label="Scenario") eng_level_input = gr.Radio(["beginner", "intermediate", "advanced"], label="English Level", value="beginner") gr.Markdown("## 2. Generate Topic 主題") default_generate_topics_prompt = """ Give me 10 topics relevant to Scenario, for a paragraph. Just the topics, no explanation, use simple English language. Make sure the vocabulary you use is at english level. """ user_generate_topics_prompt = gr.Textbox(label="Topics Prompt", value=default_generate_topics_prompt) generate_topics_button = gr.Button("AI Generate Topic Sentences") topic_output = gr.Textbox(label="AI Generated Topic 主題") topic_input = gr.Textbox(label="Topic") gr.Markdown("## 3. Generate Points 要點") default_generate_points_prompt = """ Please provide main points to develop in a paragraph about topic in the context of scenario, use simple English language and make sure the vocabulary you use is at eng_level. No more explanation either no developing these points into a simple paragraph. """ user_generate_points_prompt = gr.Textbox(label="Points Prompt", value=default_generate_points_prompt) generate_points_button = gr.Button("AI Generate Points") points_output = gr.Textbox(label="AI Generated Points 要點") points_input = gr.Textbox(label="Points") gr.Markdown("## 4. Generate Topic Sentences 主題句") default_generate_topic_sentences_prompt = """ Please provide one appropriate topic sentence that aptly introduces the subject for the given scenario and topic. Additionally, provide two topic sentences that, while related to the topic, would be considered inappropriate or less effective for the specified context. Those sentences must include the three main points:". Use English language and each sentence should not be too long. For each sentence, explain the reason in Traditional Chinese, Taiwan, 繁體中文 zh-TW. Make sure the vocabulary you use is at level. Only return the result in JSON format starting as: {{ "0": [ {{ "topic-sentence": "#","appropriate": "Y/N", "reason": "#中文解釋" }} ], "1": [ {{ "topic-sentence": "#","appropriate": "Y/N", "reason": "#中文解釋" }} ], "2": [ {{ "topic-sentence": "#","appropriate": "Y/N", "reason": "#中文解釋" }} ] }} """ user_generate_topic_sentences_prompt = gr.Textbox(label="Topic Sentences Prompt", value=default_generate_topic_sentences_prompt) generate_topic_sentences_button = gr.Button("AI Generate Topic Sentences") topic_sentence_output = gr.Textbox(label="AI Generated Topic Sentences 主題句") topic_sentence_input = gr.Textbox(label="Topic Sentences") gr.Markdown("## 5. Generate Supporting Sentence 支持句") default_generate_supporting_sentences_prompt = """ I'm aiming to improve my writing. I have a topic sentence as topic_sentence_input. Please assist me by "Developing supporting detials" based on the keyword: points to write three sentences as an example. - Make sure any revised vocabulary aligns with the eng_level. - Guidelines for Length and Complexity: Please keep the example concise and straightforward, avoiding overly technical language. Total word-count is around 50. no more explanation either no more extra non-relation sentences. """ user_generate_supporting_sentences_prompt = gr.Textbox(label="Supporting Sentences Prompt", value=default_generate_supporting_sentences_prompt) generate_supporting_sentences_button = gr.Button("AI Generate Supporting Sentences") supporting_sentences_output = gr.Textbox(label="AI Generated Supporting Sentences 支持句", show_copy_button=True) supporting_sentences_input = gr.Textbox(label="Supporting Sentences") gr.Markdown("## 6. Conclusion sentence 結論句") default_generate_conclusion_sentence_prompt = """ I'm aiming to improve my writing. By the topic sentence, please assist me by "Developing conclusion sentences" based on keywords of points to finish a paragrpah as an example. - Make sure any revised vocabulary aligns with the correctly eng_level. - Guidelines for Length and Complexity: Please keep the example concise and straightforward, avoiding overly technical language. Total word-count is around 20. """ user_generate_conclusion_sentence_prompt = gr.Textbox(label="Conclusion Sentence Prompt", value=default_generate_conclusion_sentence_prompt) generate_conclusion_sentence_button = gr.Button("AI Generate Conclusion Sentence") conclusion_sentence_output = gr.Textbox(label="AI Generated Conclusion Sentence 結論句", show_copy_button=True) conclusion_sentence_input = gr.Textbox(label="Conclusion Sentence") gr.Markdown("## 7. Paragraph Integration and Revision 段落確認與修訂") generate_paragraph_button = gr.Button("Generate Paragraph") paragraph_output = gr.Textbox(label="Generated Paragraph 完整段落", show_copy_button=True) paragraph_input = gr.Textbox(label="Paragraph") gr.Markdown("## 8. Evaluate 分析") default_user_generate_paragraph_evaluate_prompt = """ Based on the final paragraph provided, evaluate the writing in terms of content, organization, grammar, and vocabulary. Provide feedback in simple and supportive language. -- 根據上述的文章，以「內容（content）」層面評分。 - 評分等級有三級：beginner, intermediate, advanced. - 以繁體中文解釋 評分結果以 JSON 格式輸出: content: { "content_level": "#beginner/intermediate/advanced", "content_explanation": "#中文解釋" } -- 根據上述的文章，以「組織（organization）」層面評分。 - 評分等級有三級：beginner, intermediate, advanced. - 以繁體中文解釋 評分結果以 JSON 格式輸出: organization: { "organization_level": "#beginner/intermediate/advanced", "organization_explanation": "#中文解釋" } -- 根據上述的文章，以「文法和用法（Grammar and usage）」層面評分。 - 評分等級有三級：beginner, intermediate, advanced. - 以繁體中文解釋 評分結果以 JSON 格式輸出: grammar_and_usage: { "GrammarAndUsage_level": "#beginner/intermediate/advanced", "GrammarAndUsage_explanation": "#中文解釋" } -- 根據上述的文章，以「詞彙（Vocabulary ）」層面評分。 - 評分等級有三級：beginner, intermediate, advanced. - 以繁體中文解釋 評分結果以 JSON 格式輸出: vocabulary: { "Vocabulary_level": "#beginner/intermediate/advanced", "Vocabulary_explanation": "#中文解釋" } -- 根據上述的文章，以「連貫性和連接詞（Coherence and Cohesion）」層面評分。 - 評分等級有三級：beginner, intermediate, advanced. - 以繁體中文解釋 評分結果以 JSON 格式輸出: coherence_and_cohesion: { "CoherenceAndCohesion_level": "#beginner/intermediate/advanced", "CoherenceAndCohesion_explanation": "#中文解釋" } 將上述的輸出為 JSON: {{ “content“: {content’s dict}, “organization“: {organization dict}, “grammar_and_usage“: {grammar_and_usage dict}, “vocabulary“: {vocabulary dict}, “coherence_and_cohesion“: {coherence_and_cohesion dict} }} """ user_generate_paragraph_evaluate_prompt = gr.Textbox(label="Paragraph evaluate Prompt", value=default_user_generate_paragraph_evaluate_prompt) generate_paragraph_evaluate_button = gr.Button("Save and Evaluate") paragraph_evaluate_output = gr.Textbox(label="Generated Paragraph evaluate 完整段落分析", show_copy_button=True) gr.Markdown("## 9. Correct Grammatical and Spelling Errors 修訂文法與拼字錯誤") default_user_correct_grammatical_spelling_errors_prompt = """ I'm aiming to improve my writing. Please assist me by "Correcting Grammatical and Spelling Errors" in the provided paragraph. For every correction you make, I'd like an "Explanation" to understand the reasoning behind it. - Paragraph for Correction: [paragraph split by punctuation mark] - The sentence to remain unchanged: [sentence_to_remain_unchanged] - When explaining, use Traditional Chinese (Taiwan, 繁體中文) for clarity. - But others(original, Correction, revised_paragraph) in English. - Make sure any revised vocabulary aligns with the eng_level. - Guidelines for Length and Complexity: Please keep explanations concise and straightforward, - Avoiding overly technical language. The response should strictly be in the below JSON format and nothing else: { "Corrections and Explanations": [ { "original": "# original_sentence1", "Correction": "#correction_1", "Explanation": "#explanation_1(in_traditional_chinese)" }, { "original": "# original_sentence2", "Correction": "#correction_2", "Explanation": "#explanation_2(in_traditional_chinese)" }, ... ], "Revised Paragraph": "#revised_paragraph" } """ user_correct_grammatical_spelling_errors_prompt = gr.Textbox(label="Correct Grammatical and Spelling Errors Prompt", value=default_user_correct_grammatical_spelling_errors_prompt) generate_correct_grammatical_spelling_errors_button = gr.Button("Correct Grammatical and Spelling Errors") correct_grammatical_spelling_errors_output = gr.Textbox(label="Correct Grammatical and Spelling Errors 修訂文法與拼字錯誤") paragraph_correct_grammatical_spelling_errors_input = gr.Textbox(label="Paragraph") gr.Markdown("## 10. Refine Paragraph 段落改善建議") default_user_refine_paragraph_prompt = """ I need assistance with revising a paragraph. Please "Refine" the "Revised Version 1" and immediately "Provide Explanations" for each suggestion you made. - Revised Version 1 (for correction): paragraph_ai_modification(split by 標點符號) - Do not modify the sentence: topicSentence" - Make sure any revised vocabulary aligns with the eng_level. - When explaining, use Traditional Chinese (Taiwan, 繁體中文) for clarity. - But others(Origin, Suggestion, revised_paragraph_v2) use English, that's very important. - Guidelines for Length and Complexity: Please keep explanations concise and straightforward, avoiding overly technical language. The response should strictly be in the below JSON format and nothing else: { "Suggestions and Explanations": [ { "Origin": "#original_text_1", "Suggestion": "#suggestion_1", "Explanation": "#explanation_1(in_traditional_chinese)" }, { "Origin": "#original_text_2", "Suggestion": "#suggestion_2", "Explanation": "#explanation_2(in_traditional_chinese)" }, ... ], "Revised Paragraph": "#revised_paragraph_v2" } """ user_refine_paragraph_prompt = gr.Textbox(label="Refine Paragraph Prompt", value=default_user_refine_paragraph_prompt) generate_refine_paragraph_button = gr.Button("Refine Paragraph") refine_output = gr.Textbox(label="Refine Paragraph 段落改善建議", show_copy_button=True) paragraph_refine_input = gr.Textbox(label="Paragraph 段落改善", show_copy_button=True) gr.Markdown("## 11. Save and Share") paragraph_save_button = gr.Button("Save and Share") paragraph_save_output = gr.Textbox(label="Save and Share") audio_output = gr.Audio(label="Generated Speech", type="filepath") generate_topics_button.click( fn=generate_topics, inputs=[ model, max_tokens, sys_content_input, scenario_input, eng_level_input, user_generate_topics_prompt ], outputs=topic_output ) generate_points_button.click( fn=generate_points, inputs=[ model, max_tokens, sys_content_input, scenario_input, eng_level_input, topic_input, user_generate_points_prompt ], outputs=points_output ) generate_topic_sentences_button.click( fn=generate_topic_sentences, inputs=[ model, max_tokens, sys_content_input, scenario_input, eng_level_input, topic_input, points_input, user_generate_topic_sentences_prompt ], outputs=topic_sentence_output ) generate_supporting_sentences_button.click( fn=generate_supporting_sentences, inputs=[ model, max_tokens, sys_content_input, scenario_input, eng_level_input, topic_input, points_input, topic_sentence_input, user_generate_supporting_sentences_prompt ], outputs=supporting_sentences_output ) generate_conclusion_sentence_button.click( fn=generate_conclusion_sentences, inputs=[ model, max_tokens, sys_content_input, scenario_input, eng_level_input, topic_input, points_input, topic_sentence_input, user_generate_conclusion_sentence_prompt ], outputs=conclusion_sentence_output ) generate_paragraph_button.click( fn=generate_paragraph, inputs=[ topic_sentence_input, supporting_sentences_input, conclusion_sentence_input ], outputs=paragraph_output ) generate_paragraph_evaluate_button.click( fn=generate_paragraph_evaluate, inputs=[ paragraph_input, user_generate_paragraph_evaluate_prompt ], outputs=paragraph_evaluate_output ) generate_correct_grammatical_spelling_errors_button.click( fn=generate_correct_grammatical_spelling_errors, inputs=[ eng_level_input, paragraph_input, user_correct_grammatical_spelling_errors_prompt ], outputs=correct_grammatical_spelling_errors_output ) generate_refine_paragraph_button.click( fn=generate_refine_paragraph, inputs=[ eng_level_input, paragraph_correct_grammatical_spelling_errors_input, user_refine_paragraph_prompt ], outputs=refine_output ) paragraph_save_button.click( fn=paragraph_save_and_tts, inputs=[ paragraph_refine_input ], outputs=[ paragraph_save_output, audio_output ] ) demo.launch()