Update app.py

app.py

@@ -1,289 +1,289 @@
-import gradio as gr
-import pandas as pd
-from openai import AzureOpenAI
-import faiss
-import numpy as np
-import json
-import time
-import re
-import tiktoken
-import os
-from IPython.display import HTML
-
-def arabic_print(text, colour="blue"):
-    """
-    Displays Arabic text with proper RTL and right alignment in Jupyter Notebook.
-    
-    Parameters:
-    text (str): The Arabic text to display.
-    colour (str): The color of the text. Default is "blue".
-    """
-    text=text.replace("\n","<br>")
-    html_content = f"""
-    <div style="direction: rtl; text-align: right; font-size: 16px; line-height: 1.5; color: {colour};">
-        {text}
-    </div>
-    """
-    return HTML(html_content)
-
-# Example usage
-my_arabic_text = """
-باسم صاحب السمو الشيخ خليفة بن زايد آل نهيان رئيس دولة الإمارات العربية المتحدة / حاكم إمارة أبو ظبي
-بالجلسة المنعقدة بـ محكمة ابوظبي العمالية-ابتدائي بتاريخ 2 جمادى الآخرة 1441 هـ الموافق 27/01/2020 م
-برئاسة القاضي: إبراهيم ناصر الاحبابي وعضوية القاضي: مرتضى الصديق الحسن وعضوية القاضي: خليفة سليم
-"""
-
-# Display the text using the function
-arabic_print(my_arabic_text, colour="green")
-
-from openai import AzureOpenAI
-
-AZURE_OPENAI_PREVIEW_API_VERSION = os.getenv("AZURE_OPENAI_PREVIEW_API_VERSION")
-AZURE_OPENAI_ENDPOINT = os.getenv("AZURE_OPENAI_ENDPOINT")
-AZURE_OPENAI_KEY = os.getenv("AZURE_OPENAI_KEY")
-
-client = AzureOpenAI(
-  azure_endpoint = AZURE_OPENAI_ENDPOINT, 
-  api_key=AZURE_OPENAI_KEY,  
-  api_version= AZURE_OPENAI_PREVIEW_API_VERSION
-)
-
-def call_gpt_azure_message(message_text):
-    completion = client.chat.completions.create(
-    #model="GPT4Turbo",
-    model="gpt-4o",      
-    messages = message_text,
-    temperature=0.0,
-    max_tokens=1000,
-    top_p=0.95,
-    frequency_penalty=0,
-    presence_penalty=0,
-    stop=None,
-    )
-    return completion.choices[0].message.content
-
-
-def call_gpt_azure_message_stream(message_text):
-    completion = client.chat.completions.create(
-    #model="GPT4Turbo",
-    model="gpt-4o",      
-    messages = message_text,
-    temperature=0.0,
-    max_tokens=2000,
-    top_p=0.95,
-    frequency_penalty=0,
-    presence_penalty=0,
-    stop=None,
-    stream=True
-    )
-    return completion
-
-
-def call_gpt_azure(SYS_PROMPT,USER_PROMPT,MODEL="gpt-4o"):
-    message_text=[
-      {
-        "role": "system",
-        "content": SYS_PROMPT
-      },
-      {
-        "role": "user",
-        "content": USER_PROMPT
-      },
-
-    ]
-      
-    completion = client.chat.completions.create(
-    model=MODEL,   
-    messages = message_text,
-    temperature=0.0,
-    max_tokens=1000,
-    top_p=0.95,
-    frequency_penalty=0,
-    presence_penalty=0,
-    stop=None
-    )
-    return completion.choices[0].message.content
-
-
-
-# This is the main embedding function , that takes as input text and generate 1536 floats using ada3_small
-def generate_embeddings(text, model="ada3_small"): # model = "deployment_name"
-    return client.embeddings.create(input = [text], model=model).data[0].embedding    
-
-
-import tiktoken
-enc = tiktoken.get_encoding("o200k_base")
-assert enc.decode(enc.encode("hello world")) == "hello world"
-
-# To get the tokeniser corresponding to a specific model in the OpenAI API:
-enc = tiktoken.encoding_for_model("gpt-4o")
-
-import tiktoken
-
-def count_tokens_ada3(text, model_name="text-embedding-3-small"):
-    # Automatically get the correct encoding for the given model
-    encoding = tiktoken.encoding_for_model(model_name)
-    # Encode the text and count the tokens
-    return len(encoding.encode(text))
-
-
-def Get_nearest_cases_Json(case,K):
-    vquery=np.array(generate_embeddings(case))
-    vquery=vquery.reshape(1,-1)
-    
-    D, I = index_law.search(vquery, K)  # search
-    cxt_cases=""
-    cxt_list=[]
-    Locs=I[0]
-    for L in Locs:
-        cxt_cases+=str(json_cases[L])
-        
-        
-    return cxt_cases
-    
-
-
-def count_tokens(text):
-    return len(enc.encode(text))
-
-vec_embs=np.load("Data/All_cases_embedded.npy")
-
-index_law=faiss.IndexFlatIP(vec_embs.shape[1])
-index_law.add(vec_embs)
-
-#case=Emb_text_list[10]
-
-
-
-# File path
-output_file = 'Data/KSA_Legal_cases.json'
-
-# Read the JSON file
-with open(output_file, 'r', encoding='utf-8') as file:
-    json_cases = json.load(file)
-
-
-
-def GPT_AI_Judge_Json(case , cxt):
-    SYS_PROMPT="""
-    
-
-**System Role**:  
-You are an **Arabic Legal Judge Assistant**, specialized in analyzing legal cases and extracting insights from related legal precedences.
-
-
-### **Input Details**:  
-You will be given:
-1. A **legal case** (primary input).  
-2. A **context** containing multiple legal precedences in json format.
-
-
-
-### **Your Tasks**:  
-1. **Analyze the Input Case**:  
-   - Focus on the **description of the case** (توصيف القضية) and its key aspects.  
-
-2. **Identify Relevant legal Precedences**:  
-   - Search the provided context for precedences only **closely related** to the input case.  
-
-3. **Create a Comparative Analysis**:  
-   - Present a **contrastive table** comparing the relevant precedences with columns containing metadata in context
-4. **Discussion of Key Points**:  
-   - Highlight **commonalities and differences** between the input case and the relevant precedences.  
-
-5. **Suggest a Ruling Decision**:  
-   - Provide a **recommendation** for the Judge, based on the rulings of the similar precedences.  
-
----
-
-
-
-
-### **If No Relevant Precedences**:  
-- Clearly state that no related precedences were identified from the context.  
-- Apologize and note that a ruling recommendation cannot be provided.  
-
----
-
-### **Response Format**:  
-1. **Comparative Table**:  
-   - compare  relevant precedences in a table.  
-
-2. **RTL Formatting**:  
-   - Use  **right-to-left (RTL)** direction and **right alignment**.  
-   - Ensure all headers, lists, and paragraphs include `dir="rtl"` and `text-align: right`.  
-
-3. **Clear Structure**:  
-   - Provide a well-organized response for proper Arabic rendering.
-
----
-
-"""
-
-
-    
-    User_Prompt=f"Input Legal Case {case} \n Legal precedences context : {cxt}"
-    message_text=[
-      {
-        "role": "system",
-        "content": SYS_PROMPT
-      },
-      {
-        "role": "user",
-        "content": User_Prompt
-      },
-
-    ]
-    completion = client.chat.completions.create(
-    #model="gpt-35-turbo-16k",  
-    model="gpt-4o",   
-    messages = message_text,
-    temperature=0.0,
-    max_tokens=3500,
-    top_p=0.95,
-    frequency_penalty=0,
-    presence_penalty=0,
-    stop=None
-    )
-    return completion.choices[0].message.content
-
-
-import gradio as gr
-
-# Define the processing function
-def gpt_judge(case,history):
-    cxt=Get_nearest_cases_Json(case,5)
-    print("Tokens ==>",count_tokens(cxt))
-    # Example: Generate a markdown response
-    response = GPT_AI_Judge_Json(case,cxt)
-    # Save the response to a Markdown file
-    # Convert Markdown to HTML
-    return response
-welcome_message="اذكر احداث ووقائع وملابسات القضية وسأقوم بتحليلها و اقتراح الحكم بناءا عى سوابق قضائية مشابهة "
-
-chatbot=gr.Chatbot(value=[(None,welcome_message)],height=800,rtl=True)
-
-tit_html='\n<div style="text-align: right;">\n<p>اذكر الوقائع الخاصة بالقضية وتوصيفها للحصول على الاستشارة القانونية المناسبة.</p>\n</div>\n'
-tit_img = """
-<div style="text-align: right;">
-<img src="https://i.postimg.cc/rytvLcdm/ksa-leg.png" alt="Stars Logo" width="200" height="200">
-</div>
-"""
-with gr.Blocks() as demo:
-    
-    
-    gr.ChatInterface(
-        gpt_judge,
-        chatbot=chatbot,
-        title=tit_img,
-        description=tit_html,
-        theme="soft",
-        retry_btn=None,
-        undo_btn="Delete Previous",
-        clear_btn="Clear"
-    )
-    #btn = gr.Button("توعية قانونية")
-    #btn.click(fn=greet)
-    
-
-demo.launch()
+import gradio as gr
+import pandas as pd
+from openai import AzureOpenAI
+import faiss
+import numpy as np
+import json
+import time
+import re
+import tiktoken
+import os
+from IPython.display import HTML
+
+def arabic_print(text, colour="blue"):
+    """
+    Displays Arabic text with proper RTL and right alignment in Jupyter Notebook.
+    
+    Parameters:
+    text (str): The Arabic text to display.
+    colour (str): The color of the text. Default is "blue".
+    """
+    text=text.replace("\n","<br>")
+    html_content = f"""
+    <div style="direction: rtl; text-align: right; font-size: 16px; line-height: 1.5; color: {colour};">
+        {text}
+    </div>
+    """
+    return HTML(html_content)
+
+# Example usage
+my_arabic_text = """
+باسم صاحب السمو الشيخ خليفة بن زايد آل نهيان رئيس دولة الإمارات العربية المتحدة / حاكم إمارة أبو ظبي
+بالجلسة المنعقدة بـ محكمة ابوظبي العمالية-ابتدائي بتاريخ 2 جمادى الآخرة 1441 هـ الموافق 27/01/2020 م
+برئاسة القاضي: إبراهيم ناصر الاحبابي وعضوية القاضي: مرتضى الصديق الحسن وعضوية القاضي: خليفة سليم
+"""
+
+# Display the text using the function
+arabic_print(my_arabic_text, colour="green")
+
+from openai import AzureOpenAI
+
+AZURE_OPENAI_PREVIEW_API_VERSION = os.getenv("AZURE_OPENAI_PREVIEW_API_VERSION")
+AZURE_OPENAI_ENDPOINT = os.getenv("AZURE_OPENAI_ENDPOINT")
+AZURE_OPENAI_KEY = os.getenv("AZURE_OPENAI_KEY")
+
+client = AzureOpenAI(
+  azure_endpoint = AZURE_OPENAI_ENDPOINT, 
+  api_key=AZURE_OPENAI_KEY,  
+  api_version= AZURE_OPENAI_PREVIEW_API_VERSION
+)
+
+def call_gpt_azure_message(message_text):
+    completion = client.chat.completions.create(
+    #model="GPT4Turbo",
+    model="gpt-4o",      
+    messages = message_text,
+    temperature=0.0,
+    max_tokens=1000,
+    top_p=0.95,
+    frequency_penalty=0,
+    presence_penalty=0,
+    stop=None,
+    )
+    return completion.choices[0].message.content
+
+
+def call_gpt_azure_message_stream(message_text):
+    completion = client.chat.completions.create(
+    #model="GPT4Turbo",
+    model="gpt-4o",      
+    messages = message_text,
+    temperature=0.0,
+    max_tokens=2000,
+    top_p=0.95,
+    frequency_penalty=0,
+    presence_penalty=0,
+    stop=None,
+    stream=True
+    )
+    return completion
+
+
+def call_gpt_azure(SYS_PROMPT,USER_PROMPT,MODEL="gpt-4o"):
+    message_text=[
+      {
+        "role": "system",
+        "content": SYS_PROMPT
+      },
+      {
+        "role": "user",
+        "content": USER_PROMPT
+      },
+
+    ]
+      
+    completion = client.chat.completions.create(
+    model=MODEL,   
+    messages = message_text,
+    temperature=0.0,
+    max_tokens=1000,
+    top_p=0.95,
+    frequency_penalty=0,
+    presence_penalty=0,
+    stop=None
+    )
+    return completion.choices[0].message.content
+
+
+
+# This is the main embedding function , that takes as input text and generate 1536 floats using ada3_small
+def generate_embeddings(text, model="ada3_small"): # model = "deployment_name"
+    return client.embeddings.create(input = [text], model=model).data[0].embedding    
+
+
+import tiktoken
+enc = tiktoken.get_encoding("o200k_base")
+assert enc.decode(enc.encode("hello world")) == "hello world"
+
+# To get the tokeniser corresponding to a specific model in the OpenAI API:
+enc = tiktoken.encoding_for_model("gpt-4o")
+
+import tiktoken
+
+def count_tokens_ada3(text, model_name="text-embedding-3-small"):
+    # Automatically get the correct encoding for the given model
+    encoding = tiktoken.encoding_for_model(model_name)
+    # Encode the text and count the tokens
+    return len(encoding.encode(text))
+
+
+def Get_nearest_cases_Json(case,K):
+    vquery=np.array(generate_embeddings(case))
+    vquery=vquery.reshape(1,-1)
+    
+    D, I = index_law.search(vquery, K)  # search
+    cxt_cases=""
+    cxt_list=[]
+    Locs=I[0]
+    for L in Locs:
+        cxt_cases+=str(json_cases[L])
+        
+        
+    return cxt_cases
+    
+
+
+def count_tokens(text):
+    return len(enc.encode(text))
+
+vec_embs=np.load("data/All_cases_embedded.npy")
+
+index_law=faiss.IndexFlatIP(vec_embs.shape[1])
+index_law.add(vec_embs)
+
+#case=Emb_text_list[10]
+
+
+
+# File path
+output_file = 'data/KSA_Legal_cases.json'
+
+# Read the JSON file
+with open(output_file, 'r', encoding='utf-8') as file:
+    json_cases = json.load(file)
+
+
+
+def GPT_AI_Judge_Json(case , cxt):
+    SYS_PROMPT="""
+    
+
+**System Role**:  
+You are an **Arabic Legal Judge Assistant**, specialized in analyzing legal cases and extracting insights from related legal precedences.
+
+
+### **Input Details**:  
+You will be given:
+1. A **legal case** (primary input).  
+2. A **context** containing multiple legal precedences in json format.
+
+
+
+### **Your Tasks**:  
+1. **Analyze the Input Case**:  
+   - Focus on the **description of the case** (توصيف القضية) and its key aspects.  
+
+2. **Identify Relevant legal Precedences**:  
+   - Search the provided context for precedences only **closely related** to the input case.  
+
+3. **Create a Comparative Analysis**:  
+   - Present a **contrastive table** comparing the relevant precedences with columns containing metadata in context
+4. **Discussion of Key Points**:  
+   - Highlight **commonalities and differences** between the input case and the relevant precedences.  
+
+5. **Suggest a Ruling Decision**:  
+   - Provide a **recommendation** for the Judge, based on the rulings of the similar precedences.  
+
+---
+
+
+
+
+### **If No Relevant Precedences**:  
+- Clearly state that no related precedences were identified from the context.  
+- Apologize and note that a ruling recommendation cannot be provided.  
+
+---
+
+### **Response Format**:  
+1. **Comparative Table**:  
+   - compare  relevant precedences in a table.  
+
+2. **RTL Formatting**:  
+   - Use  **right-to-left (RTL)** direction and **right alignment**.  
+   - Ensure all headers, lists, and paragraphs include `dir="rtl"` and `text-align: right`.  
+
+3. **Clear Structure**:  
+   - Provide a well-organized response for proper Arabic rendering.
+
+---
+
+"""
+
+
+    
+    User_Prompt=f"Input Legal Case {case} \n Legal precedences context : {cxt}"
+    message_text=[
+      {
+        "role": "system",
+        "content": SYS_PROMPT
+      },
+      {
+        "role": "user",
+        "content": User_Prompt
+      },
+
+    ]
+    completion = client.chat.completions.create(
+    #model="gpt-35-turbo-16k",  
+    model="gpt-4o",   
+    messages = message_text,
+    temperature=0.0,
+    max_tokens=3500,
+    top_p=0.95,
+    frequency_penalty=0,
+    presence_penalty=0,
+    stop=None
+    )
+    return completion.choices[0].message.content
+
+
+import gradio as gr
+
+# Define the processing function
+def gpt_judge(case,history):
+    cxt=Get_nearest_cases_Json(case,5)
+    print("Tokens ==>",count_tokens(cxt))
+    # Example: Generate a markdown response
+    response = GPT_AI_Judge_Json(case,cxt)
+    # Save the response to a Markdown file
+    # Convert Markdown to HTML
+    return response
+welcome_message="اذكر احداث ووقائع وملابسات القضية وسأقوم بتحليلها و اقتراح الحكم بناءا عى سوابق قضائية مشابهة "
+
+chatbot=gr.Chatbot(value=[(None,welcome_message)],height=800,rtl=True)
+
+tit_html='\n<div style="text-align: right;">\n<p>اذكر الوقائع الخاصة بالقضية وتوصيفها للحصول على الاستشارة القانونية المناسبة.</p>\n</div>\n'
+tit_img = """
+<div style="text-align: right;">
+<img src="https://i.postimg.cc/rytvLcdm/ksa-leg.png" alt="Stars Logo" width="200" height="200">
+</div>
+"""
+with gr.Blocks() as demo:
+    
+    
+    gr.ChatInterface(
+        gpt_judge,
+        chatbot=chatbot,
+        title=tit_img,
+        description=tit_html,
+        theme="soft",
+        retry_btn=None,
+        undo_btn="Delete Previous",
+        clear_btn="Clear"
+    )
+    #btn = gr.Button("توعية قانونية")
+    #btn.click(fn=greet)
+    
+
+demo.launch()