Spaces:
Runtime error
Runtime error
| import ollama | |
| import json | |
| import time | |
| from parser import parse_resume | |
| # --- Step 1: Parse Resume --- | |
| resume_file = r"C:\Users\prana\Downloads\ABDM\Documents\PranavKerkar_resume.pdf" | |
| parsed_data = parse_resume(resume_file) | |
| # --- Step 2: Context for AI --- | |
| context = f""" | |
| You are an AI Interviewer. | |
| Candidate's resume: | |
| {json.dumps(parsed_data, indent=2)} | |
| Rules: | |
| - Start with: "Am I audible?" | |
| - If 'no', retry politely. | |
| - Ask questions based on resume (skills, education, projects). | |
| - Wait 6–7 seconds before repeating if no response. | |
| - If unanswered after 2 tries, mark 'Unanswered' and move on. | |
| - End with: "Thank you for your time, results will be shared via email. Do you have any questions for me?" | |
| """ | |
| # --- Step 3: Helper Function for Rule-based Logic --- | |
| def rule_based_response(user_input, parsed_data): | |
| """Simple rule-based interview flow""" | |
| # Greeting / Audible check | |
| if user_input in ["no", "not really", "can't hear"]: | |
| return "AI Interviewer: Let me try again, can you hear me now?" | |
| if user_input in ["yes", "yeah", "yep"]: | |
| return f"AI Interviewer: Great, how are you {parsed_data.get('name','there')}?" | |
| # Resume skills | |
| skills = parsed_data.get("skills", []) | |
| for skill in skills: | |
| if skill.lower() in user_input: | |
| return f"AI Interviewer: Since you mentioned {skill}, can you rate yourself 1–10 and explain why?" | |
| # Education based | |
| if "b.tech" in user_input or "bachelor" in user_input: | |
| return "AI Interviewer: Can you share a project you did during your Bachelor's?" | |
| if "m.tech" in user_input or "master" in user_input: | |
| return "AI Interviewer: What was your Master's thesis about?" | |
| # Projects fallback | |
| if "project" in user_input: | |
| return "AI Interviewer: Can you walk me through your most challenging project?" | |
| return None # no rule → fallback to AI | |
| # --- Step 4: Interview Loop --- | |
| print("AI Interviewer: Am I audible?") | |
| unanswered_count = 0 | |
| last_question = None | |
| while True: | |
| candidate_input = input("Candidate: ").strip().lower() | |
| if candidate_input in ["exit", "quit", "bye"]: | |
| print("AI Interviewer: Thank you, goodbye.") | |
| break | |
| if candidate_input == "": | |
| if last_question and unanswered_count < 2: | |
| unanswered_count += 1 | |
| print("AI Interviewer: (waiting 6 seconds...)") | |
| time.sleep(6) | |
| print("AI Interviewer: Let me repeat —", last_question) | |
| continue | |
| elif last_question and unanswered_count >= 2: | |
| print("AI Interviewer: Marking this question as 'Unanswered'. Moving on.") | |
| unanswered_count = 0 | |
| last_question = None | |
| continue | |
| # 1. Rule-based first | |
| reply = rule_based_response(candidate_input, parsed_data) | |
| if reply: | |
| print(reply) | |
| last_question = reply.replace("AI Interviewer: ", "") | |
| unanswered_count = 0 | |
| continue | |
| # 2. Fallback to Ollama AI | |
| response = ollama.chat( | |
| model="llama2", | |
| messages=[ | |
| {"role": "system", "content": context}, | |
| {"role": "user", "content": candidate_input} | |
| ] | |
| ) | |
| interviewer_reply = response["message"]["content"] | |
| print("AI Interviewer:", interviewer_reply) | |
| last_question = interviewer_reply | |
| unanswered_count = 0 | |