resume_parser / app.py
akshansh36's picture
removed printing of llm sherpa json
2069f24 verified
import streamlit as st
from pymongo import MongoClient
import fitz # PyMuPDF
import ast
import re
from groq import Groq
import concurrent.futures
import pandas as pd
import io
import json
import requests
DB_NAME = 'akshansh_db'
try:
client = MongoClient('mongodb+srv://akshansh:HzLqyintpUfmcC4D@dev001.4fkwn.mongodb.net/')
db = client[DB_NAME]
collection = db['parsed_resume_streamlit']
print("MongoDB connection established.")
except Exception as e:
print(f"Error connecting to MongoDB: {e}")
groq_api = "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"
llmsherpa_api_url = " http://65.2.175.211:5010/api/parseDocument?renderFormat=all&applyOcr=yes"
def sanitize_text(text):
# Escape single quotes and other potentially problematic characters
return text.replace("'", "\\'")
def process_using_llm(text):
try:
sanitized_text = sanitize_text(text)
client = Groq(api_key=groq_api)
prompt=f"""
1. Given is the text content of a resume, please extract information from it and output the result in a dictionary format which is defined below along with the expected data structure, strictly adhere to the dictionary format given below, if any field is not present leave it empty.
Note: 1. Do not skip any information and do not add any information which is not present in the input content.
2. In case of github urls, linkedin urls, email id, add only if the url is present else leave it empty.
3. For the work experience only the latest work experience is required that is the one which is presntly being done or done at the last.
4. In the format of extracted_content, do not give any other things, like comments or anything
Input: {sanitized_text}
Expected output format: "extracted_content: {{
'name': 'String',
'email': 'String',
'phone': 'String',
'location': 'String',
'linkedin': 'String',
'github':'String',
'inter_personal_skills': [
'String'
],
'technical_skills': [
'String'
],
'soft_skills':[
'String'
],
'programming_languages':[
'String'
],
'linguistic_languages':[
'String'
],
'latest_work_experience':{{
'company': 'String',
'role': 'String',
'duration': 'String',
'work_location': 'String',
}},
'graduation_details':{{
'course':'String',
'institution':'String',
'course_type':'String',
'year_of_graduation':'String',
'percentage_or_cgpa':'String'
}},
'higher_secondary_education':{{
'institution':'String',
'education_board_type':'String',
'year_of_completion':'String',
'percentage_or_cgpa':'String'
}},
'secondary_education':{{
'institution':'String',
'education_board_type':'String',
'year_of_completion':'String',
'percentage_or_cgpa':'String'
}}
}}"
"""
chat_completion = client.chat.completions.create(
messages=[
{
"role": "user",
"content": prompt
}
],
model="llama3-70b-8192"
)
return chat_completion.choices[0].message.content
except Exception as e:
print(f"An error occurred in LLM part: {e}")
return None
def extract(output):
match = re.search(r'extracted_content:\s*(\{.*\})', output, re.DOTALL)
if match:
extracted_content = match.group(1)
return ast.literal_eval(extracted_content)
else:
print("No extracted content found in parsing llm's output")
return {}
def process_resume(pdf_content):
response = requests.post(llmsherpa_api_url, files={'file': ('resume.pdf', pdf_content, 'application/pdf')})
# Check if the response is valid JSON
try:
response_json = response.json()
except json.JSONDecodeError:
print("Failed to decode JSON response")
return None
if 'return_dict' in response_json and 'result' in response_json['return_dict']:
blocks = response_json['return_dict']['result']['blocks']
content=""
for block in blocks:
tag=block.get('tag',None)
if tag=="table":
table_rows=block['table_rows']
for row in table_rows:
cells=row.get('cells',None)
if cells:
cells=row['cells']
for cell in cells:
value=cell['cell_value']
if isinstance(value,dict):
sentences=value.get('sentences',None)
for sentence in sentences:
content+=sentence+'\n'
elif value !='':
content+=value+'\n'
else:
value=row.get('cell_value',None)
if value:
content+=value+'\n'
else:
sentences=block.get('sentences')
for s in sentences:
content+=s+'\n'
if content:
result = {}
processed_text = process_using_llm(content)
if processed_text:
extracted_output = extract(processed_text)
result=extracted_output
return result
def json_to_excel(data): # data is a list of JSON
try:
# Define the specific order of columns
column_order = [
'Name', 'Phone', 'Location', 'Email', 'Linkedin', 'Github',
'Graduation Details', 'Graduation Institution', 'Graduation Course Type',
'Year of Graduation', 'Aggregate Percentage in Graduation',
'Higher Secondary Institute Name', 'Higher Secondary Education Board Type',
'Year of Completion of Higher Secondary Education',
'Aggregate Percentage in Higher Secondary Education',
'Secondary Education Institute Name', 'Secondary Education Board Type',
'Year of Completion of Secondary Education', 'Aggregate Percentage in Secondary Education',
'Current Working Organization', 'Current Designation', 'Current Work Duration',
'Current Work Location', 'Inter Personal Skills', 'Technical Skills',
'Soft Skills', 'Programming Languages', 'Languages'
]
flat_data = []
for item in data:
flat_item = {}
if "name" in item:
name = item.get("name", None)
if name:
flat_item['Name'] = name
if "phone" in item:
phone = item.get('phone', None)
if phone:
flat_item['Phone'] = phone
if "location" in item:
location = item.get("location", None)
if location:
flat_item['Location'] = location
if "email" in item:
email = item.get("email", None)
if email:
flat_item['Email'] = email
if "linkedin" in item:
linkedin = item.get('linkedin', None)
if linkedin:
flat_item['Linkedin'] = linkedin
if 'github' in item:
github = item.get('github', None)
if github:
flat_item['Github'] = github
if "graduation_details" in item:
ed = item["graduation_details"]
course = ed.get('course', None)
if course:
flat_item['Graduation Details'] = course
institution = ed.get('institution', None)
if institution:
flat_item['Graduation Institution'] = institution
course_type = ed.get('course_type', None)
if course_type:
flat_item['Graduation Course Type'] = course_type
year = ed.get('year_of_graduation', None)
if year:
flat_item['Year of Graduation'] = year
marks = ed.get('percentage_or_cgpa', None)
if marks:
flat_item['Aggregate Percentage in Graduation'] = marks
if "higher_secondary_education" in item:
ed = item.get('higher_secondary_education')
institution = ed.get('institution', None)
if institution:
flat_item['Higher Secondary Institute Name'] = institution
board = ed.get('education_board_type', None)
if board:
flat_item['Higher Secondary Education Board Type'] = board
year = ed.get('year_of_completion', None)
if year:
flat_item['Year of Completion of Higher Secondary Education'] = year
marks = ed.get('percentage_or_cgpa', None)
if marks:
flat_item['Aggregate Percentage in Higher Secondary Education'] = marks
if "secondary_education" in item:
ed = item.get('secondary_education')
institution = ed.get('institution', None)
if institution:
flat_item['Secondary Education Institute Name'] = institution
board = ed.get('education_board_type', None)
if board:
flat_item['Secondary Education Board Type'] = board
year = ed.get('year_of_completion', None)
if year:
flat_item['Year of Completion of Secondary Education'] = year
marks = ed.get('percentage_or_cgpa', None)
if marks:
flat_item['Aggregate Percentage in Secondary Education'] = marks
if 'latest_work_experience' in item:
current_work = item.get('latest_work_experience', None)
if current_work:
company = current_work.get('company', None)
if company:
flat_item['Current Working Organization'] = company
role = current_work.get('role', None)
if role:
flat_item['Current Designation'] = role
duration = current_work.get('duration', None)
if duration:
flat_item['Current Work Duration'] = duration
location = current_work.get('work_location', None)
if location:
flat_item['Current Work Location'] = location
if "inter_personal_skills" in item:
flat_item["Inter Personal Skills"] = ", ".join(item["inter_personal_skills"])
if "technical_skills" in item:
flat_item["Technical Skills"] = ", ".join(item["technical_skills"])
if "soft_skills" in item:
flat_item["Soft Skills"] = ", ".join(item["soft_skills"])
if "programming_languages" in item:
flat_item["Programming Languages"] = ", ".join(item["programming_languages"])
if "linguistic_languages" in item:
flat_item["Languages"] = ", ".join(item["linguistic_languages"])
flat_data.append(flat_item)
# Create DataFrame
df = pd.DataFrame(flat_data)
# Reorder columns according to the specified order
df = df[[col for col in column_order if col in df.columns]]
return df
except Exception as e:
print(f"Error occurred in converting JSON to Excel: {e}")
return None
def main():
st.title('Resume Parser')
# Allow the user to specify the maximum number of resumes to upload
max_resumes = st.number_input("Maximum number of resumes to upload, limit: 5", min_value=1, max_value=5, value=1, step=1)
# Allow the user to upload the resumes
uploaded_files = st.file_uploader("Upload your resumes", type=["pdf"], accept_multiple_files=True)
if uploaded_files:
if len(uploaded_files) != max_resumes:
st.warning(f"Please upload exactly {max_resumes} resumes.")
else:
submit_button = st.button("Process Resumes")
if submit_button:
try:
with st.spinner("Your resumes are being processed..."):
with concurrent.futures.ThreadPoolExecutor() as executor:
# Reading the PDF content for each uploaded file
pdf_contents = [file.read() for file in uploaded_files[:max_resumes]]
# Process each PDF content using the process_resume function
results = list(executor.map(process_resume, pdf_contents))
successful_resumes = []
failed_resumes_count = 0
for result in results:
if result:
successful_resumes.append(result)
collection.insert_one(result)
else:
failed_resumes_count += 1
if successful_resumes:
if failed_resumes_count > 0:
st.warning(f"{failed_resumes_count} resumes could not be processed. Do you still want to download the successfully processed resumes?")
user_response = st.radio("Please select:", ("Yes", "No"))
if user_response == "Yes":
# Convert the processed resume data to a pandas DataFrame
df = json_to_excel(successful_resumes)
if df is not None:
# Create an Excel file in memory
excel_file = io.BytesIO()
with pd.ExcelWriter(excel_file, engine='xlsxwriter') as writer:
df.to_excel(writer, index=False, sheet_name='Resumes')
st.download_button(
label="Download XLSX file",
data=excel_file.getvalue(),
file_name="resume_data.xlsx",
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
)
else:
st.error("Aw! Snap, could not process any of the resumes. Please try again later.")
elif user_response == "No":
st.info("Then try again after some time.")
else:
# Convert the processed resume data to a pandas DataFrame
df = json_to_excel(successful_resumes)
if df is not None:
# Create an Excel file in memory
excel_file = io.BytesIO()
with pd.ExcelWriter(excel_file, engine='xlsxwriter') as writer:
df.to_excel(writer, index=False, sheet_name='Resumes')
st.download_button(
label="Download XLSX file",
data=excel_file.getvalue(),
file_name="resume_data.xlsx",
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
)
st.success(f"Resumes processed successfully! {len(successful_resumes)} out of {max_resumes} resumes processed.")
else:
st.error("Aw! Snap, could not process any of the resumes. Please try again later.")
else:
st.error("Aw! Snap, could not process any of the resumes. Please try again later.")
except Exception as e:
st.error("Aw! Snap, could not process your resumes. Please try again later.")
print(f"Error processing resumes: {e}")
if __name__ == "__main__":
main()