Create app.py

app.py

@@ -0,0 +1,185 @@
+import gradio as gr
+from openai import OpenAI
+import pandas as pd
+import io
+import tempfile
+import shutil
+import google.generativeai as genai
+import os
+
+api_key = os.getenv("OPENAI_API_KEY")
+gemni_api_key = os.getenv("GEMNI_API_KEY")
+
+
+supported_languages = ['English', 'Brazilian Portuguese', 'Latin American Spanish', 'French', 'European Portuguese', 'Castilian Spanish', 'German', 'Italian', 'Czech', 'Danish', 'Dutch', 'Finnish', 'Norwegian', 'Swedish', 'Hungarian', 'Greek', 'Romanian', 'Polish', 'Arabic', 'Urdu']
+
+
+
+
+# OPENAI
+client = OpenAI(api_key  = api_key)
+
+def translate_text_openai(source_language, target_language, TEXT, max_characters):
+
+    response = client.chat.completions.create(
+      model="gpt-3.5-turbo-0125",
+      temperature = 0.1,
+      messages=[
+        {"role": "system", "content": "You are a multilingual translator for movies subtitles."},
+        {"role": "system", "content": "The number of input characters and output characters should be the same despite the change in language."},
+         {"role": "system", "content": f"In response, maximum characters allowed are {max_characters}"},
+        {"role": "system", "content": "You SHOULD NOT SKIP ANY LINE OR ANY INFORMATION"},
+        {"role": "system", "content": "The Tranlation should be error proof"},
+
+
+        {"role": "user", "content": f"""Translate the text from {source_language} language to {target_language} language.:
+        \nTEXT: {TEXT}
+        NOTE: THE OUTPUT SHOULD BE IN  {target_language} language.
+        \nREMEMBER: MAXIMUM output chaeracters should be {max_characters}
+
+        END NOTE: THE OUTPUT SHOULD BE IN  {target_language} language.
+        REMEMBER: THE OUTPUT SHOULD BE IN  {target_language} language.
+        Hey on some instances you give response in languages other than {target_language}, which is wrong
+        """},
+      ]
+    )
+    return response.choices[0].message.content
+
+def translate_text_correct_openai(source_language, target_language, TEXT, max_characters):
+
+    response = client.chat.completions.create(
+      model="gpt-3.5-turbo-0125",
+      temperature = 0.1,
+      messages=[
+        {"role": "system", "content": "You reduce the size of the sentences."},
+        {"role": "system", "content": f"The maximuim output should not be more than {max_characters} characters."},
+        {"role": "user", "content": f"""
+        DO NOT CHANGE THE LANGUAGE
+         Reduce the size of the text to less than {max_characters} characters even if there is a change in meaning.
+         REMEMBER LLM, THE INPUT AND THE OUTPUT SHOULD BE IN THE SAME LANGUAGE
+         \nWrite the sentence in shortest possible manner
+        \nTEXT: {TEXT}\
+        REMEMBER LLM, THE INPUT AND THE OUTPUT SHOULD BE IN THE SAME LANGUAGE
+     
+        """},
+      ]
+    )
+    return response.choices[0].message.content
+
+# GEMNI
+
+
+genai.configure(api_key=gemni_api_key)
+
+model = genai.GenerativeModel('gemini-1.5-flash')
+
+
+def translate_text_gemni(source_language, target_language, TEXT, max_characters):
+
+    response = model.generate_content(f'''You are a multilingual translator for movies subtitles.
+        The number of input characters and output characters should be the same despite the change in language.
+        In response, maximum characters allowed are {max_characters}.
+        You SHOULD NOT SKIP ANY LINE OR ANY INFORMATION.
+        The Tranlation should be error proof.
+        Translate the text from {source_language} language to {target_language} language.:
+        \nTEXT: {TEXT}
+        NOTE: THE OUTPUT SHOULD BE IN  {target_language} language.
+        \nREMEMBER: MAXIMUM output chaeracters should be {max_characters}
+
+        END NOTE: THE OUTPUT SHOULD BE IN  {target_language} language.
+        REMEMBER: THE OUTPUT SHOULD BE IN  {target_language} language.
+        Hey on some instances you give response in languages other than {target_language}, which is wrong
+        '''
+
+    )
+    return response.text
+
+def translate_text_correct_gemni(source_language, target_language, TEXT, max_characters):
+
+    response = model.generate_content(f"""
+        You reduce the size of the sentences.
+        The maximuim output should not be more than {max_characters} characters.
+        DO NOT CHANGE THE LANGUAGE
+        Reduce the size of the text to less than {max_characters} characters even if there is a change in meaning.
+        REMEMBER LLM, THE INPUT AND THE OUTPUT SHOULD BE IN THE SAME LANGUAGE
+         \nWrite the sentence in shortest possible manner
+        \nTEXT: {TEXT}\
+        REMEMBER LLM, THE INPUT AND THE OUTPUT SHOULD BE IN THE SAME LANGUAGE
+        """
+    )
+    return response.text
+
+
+
+
+
+
+
+
+
+
+def check_conditon_openai(source_language, target_language, response, max_characters):
+  length = len(response)
+  
+  if length > int(max_characters):
+    response = translate_text_correct_openai(source_language, target_language, response, max_characters)
+    return check_conditon_openai(source_language, target_language, response, max_characters)
+  return response
+
+
+def check_conditon_gemni(source_language, target_language, response, max_characters):
+  length = len(response)
+  
+  if length > int(max_characters):
+    response = translate_text_correct_gemni(source_language, target_language, response, max_characters)
+    return check_conditon_gemni(source_language, target_language, response, max_characters)
+  return response
+
+
+
+
+def get_translation(source_language, target_language, TEXT, max_characters):
+    
+    response_openai = translate_text_openai(source_language, target_language, TEXT, max_characters)
+    response_openai = check_conditon_openai(source_language, target_language, response_openai, max_characters)
+
+    response_gemni = translate_text_gemni(source_language, target_language, TEXT, max_characters)
+    response_gemni = check_conditon_gemni(source_language, target_language, response_gemni, max_characters)
+
+
+    excel_data_path = create_excel(TEXT, response_openai, response_gemni)
+
+    
+    return excel_data_path
+
+
+def create_excel(TEXT, response_openai, response_gemni):
+    # Create a DataFrame from the input data
+    df = pd.DataFrame({"Original Text": TEXT, "OpenAI Translated": response_openai, "Gemni Translated": response_gemni}, index = [1])
+    
+    # Create a temporary file to store the Excel data
+    with tempfile.NamedTemporaryFile(delete=False, suffix='.xlsx') as temp_file:
+        # Write the DataFrame to the temporary file as an Excel file
+        with pd.ExcelWriter(temp_file, engine='xlsxwriter') as writer:
+            df.to_excel(writer, index=False, sheet_name='Sheet1')
+        
+        # Return the path to the temporary file
+        temp_file_path = temp_file.name
+    
+    return temp_file_path
+
+
+iface = gr.Interface(
+    fn=get_translation,
+    inputs=[
+        gr.Dropdown(choices= supported_languages, label="Source Language"),  # Add more languages as needed
+        gr.Dropdown(choices= supported_languages, label="Target Language"),
+        gr.Textbox(lines=2, label="Input Text"),
+        gr.Textbox(lines=1, label="Difine number of output characters"),
+    ],
+    outputs=gr.File(label="Download Excel File"),
+    title="MVP Multilingual Translation",
+    description="MVP Multilingual Translation by Farhan",
+)
+
+iface.launch(share=True, debug=True)