Update src/streamlit_app.py

adding criterions analysis and weights for screening process

src/streamlit_app.py

@@ -28,17 +28,177 @@ client = AzureOpenAI(
 )
 
 # Function to extract information from CV using Azure OpenAI GPT-4 model
-def extract_info_from_azure(text):
+def extract_PD_info_from_azure(text):
     prompt = f"""
-    Extract the following information from this CV text:
-    1. Job title
-    2. Location
-    3. Skills
-    4. Years of experience
-    5. Education level
-
-    CV Text:
+    Act as an HR recruiter. Carefully evaluate the provided job description (PD) document and extract the following information into the specified structured format. Ensure that all fields are populated. If a field is not explicitly mentioned in the text, use ""N/A"" as the default value.
+
+               - Requirements Categories to Extract:
+
+               - Location
+            
+               - Country
+
+               - City
+
+               - Department
+
+               - Required Years of Experience
+
+               - Responsibilities
+
+               - Skills
+
+               - Experience
+
+               - Education
+
+               - Other Factors
+
+               - Languages
+
+
+    PD Text:
+    {text}
+    """
+    response = client.chat.completions.create(
+    messages=[
+            {
+                "role": "system",
+                "content": "You are a helpful assistant.",
+            },
+            {
+                "role": "user",
+                "content": prompt,
+            }
+        ],
+        max_tokens=4096,
+        temperature=1.0,
+        top_p=1.0,
+        model=DEPLOYMENT_NAME
+    )
+
+    try:
+
+        result = response.choices[0].message.content
+        return result
+    except Exception as e:
+        st.error(f"Error occurred while contacting Azure OpenAI: {e}")
+        return None
+
+# Function to extract information from CV using Azure OpenAI GPT-4 model
+def extract_criterions_from_azure(text):
+    prompt = f"""
+    Act as an HR recruiter. Based on the main criteria categories defined below, extract bullet items from the provided PD text. For each bullet item, create a Criterion Object with the following fields:
+    Instructions:
+        - Criterions should represent each item listed under each PD section (Skills,Education, Experience, Other Factors):
+                1- For each item under PD sections mentioned above, ensure to have related criteria item created to measure it. only skip those items may not relevant or negate a state such as ""There are no extra requirements specified for this position"" .
+        - PD sections with NA value, dont create Crtierion for it at all.
+        - Ensure to add Criterions for Languages Proficiency based on its identified level
+        - Ensure to add Criterions that has clue within the PD to measure it, if no clue in PD, do not consider it.
+    
+    **Important:**  
+        - Output all Criterion items together under each other as follows, make sure category part formated in bold:
+            1- Item 1  (Category: Skills)
+            2- Item 2 (Category: Experience)  
+        
+       
+    
+     
+    PD Text:
+    {text}
+    """
+    response = client.chat.completions.create(
+    messages=[
+            {
+                "role": "system",
+                "content": "You are a helpful assistant.",
+            },
+            {
+                "role": "user",
+                "content": prompt,
+            }
+        ],
+        max_tokens=4096,
+        temperature=1.0,
+        top_p=1.0,
+        model=DEPLOYMENT_NAME
+    )
+
+    try:
+
+        result = response.choices[0].message.content
+        return result
+    except Exception as e:
+        st.error(f"Error occurred while contacting Azure OpenAI: {e}")
+        return None
+    
+def extract_weights_from_azure(text):
+    prompt = f"""
+    Act as an HR recruiter. Based on the extracted criteria records from the previous step, evaluate the importance of each requirement category type based on provided criteria and assign a weight to it. The weight should reflect how critical the category is for the job, with a score from 1 to 100, ensuring that the sum of all weights does not exceed 100.
+
+    Requirements Categories to Evaluate and Assign Weights:
+    Location (keyword: Location)
+
+    Required Years of Experience (keyword: Required_Years_of_Experience)
+
+    Responsibilities (keyword: Responsibilities)
+
+    Skills (keyword: Skills)
+
+    Experience (keyword: Experience)
+
+    Education (keyword: Education)
+
+    Other Factors (keyword: Other_Factors)
+
+    Languages (keyword: Language_Proficiency)
+
+
+    Instructions:
+    
+        Dynamic Handling of Categories:
+            **Include categories that have explicit criteria mentioned in the Criterions JSON Content.
+            **Exclude categories that are not present in the content.
+            **Adjust the weights dynamically so the sum of all assigned weights does not exceed 100.
+
+
+        Weight Assignment:
+            **Assign a weight (1–100) to each category based on its importance to the job.
+            **Distribute weights logically, ensuring critical categories like Skills, Experience, Education, and Language Proficiency are well represented.
+            **Ensure the sum of all weights, including Extra_Requirement_Three if present, totals exactly 100. 
+            **If the total exceeds or is less than 100, proportionally adjust the weights of the other categories.
+
+        Category Justification:
+            **For each category, provide a detailed justification of its importance, tied to the specific requirements in the job description.
+            **Explain how the requirements support the assigned weight and their direct relevance to job performance or success.
+
+        Final Adjustment Rule:
+                After assigning initial weights:
+                ** Check if the total weight equals 100.
+                ** If the total is less than 100, proportionally increase the weights of all present categories.
+                ** If the total is more than 100, proportionally decrease the weights of all present categories.
+                ** Preserve explicit weight floors (e.g., Experience greater than or equal to 15%) when adjusting.
+                ** The final output must sum to exactly 100.                    
+        
+        Output Format:
+            **The response must strictly follow below, including the Category_Description for each category.
+            **Ensure that all categories in the PD criteria are represented in the weights, and no category is missing.
+            **If Language_Proficiency is found in the criteria, it should have a weight assigned and be included in the output.
+            **If Extra_Requirement_One is found in the criteria, it should have a weight assigned and be included in the output.
+            **If Extra_Requirement_Two is found in the criteria, it should have a weight assigned and be included in the output.
+            **If Extra_Requirement_Three is found in the criteria, it should have a weight assigned and be included in the output.
+            
+            
+            Example Output:
+            1- Skills (Weight: 35%): <short description about required skills for this job>
+            2- Experience (Weight: 25%): <short description about required experiences for this job>
+            3- Education (Weight: 25%): <short description about required education for this job>
+            4- Language Proficincy (Weight: 15%): <short description about required languages for this job>
+
+    Criterions list: 
     {text}
+    
+    
     """
     response = client.chat.completions.create(
     messages=[
@@ -74,11 +234,11 @@ cv_text = st.text_area("Enter CV Text", height=300)
 
 if cv_text:
     # Display the entered CV text
-    st.subheader("Entered Text from CV")
-    st.text_area("CV Text", cv_text, height=300)
+    # st.subheader("Entered Text from CV")
+    # st.text_area("CV Text", cv_text, height=300)
 
     # Extract relevant information using Azure OpenAI GPT-4
-    extracted_info = extract_info_from_azure(cv_text)
+    extracted_info = extract_PD_info_from_azure(cv_text)
 
     # Display the extracted information
     if extracted_info:
@@ -86,3 +246,24 @@ if cv_text:
         st.write(extracted_info)
     else:
         st.write("Could not extract information. Please try again.")
+
+
+     # Extract Criterions
+    extracted_criterions = extract_criterions_from_azure(extracted_info)
+    
+    # Display the extracted information
+    if extracted_criterions:
+        st.subheader("Extracted Criterion List")
+        st.write(extracted_criterions)
+    else:
+        st.write("Could not extract criterion information. Please try again.")
+        
+     # Extract Criterions
+    extracted_weights = extract_weights_from_azure(extracted_criterions)
+    
+    # Display the extracted information
+    if extracted_weights:
+        st.subheader("Extracted Weights List")
+        st.write(extracted_weights)
+    else:
+        st.write("Could not extract weights information. Please try again.")