tasks.py

# tasks.py
from crewai import Task
from typing import List

def create_question_task(job_title: str, interviewer, question_number: int, interview_history: List) -> Task:
    """Creates a task for generating interview questions"""
    context = "\n".join([
        f"Q{i+1}: {interaction['question']}\nA{i+1}: {interaction['answer']}"
        for i, interaction in enumerate(interview_history)
    ]) if interview_history else "No previous responses"
    
    return Task(
        description=f"""You are conducting interview question #{question_number} for the {job_title} position.
        
        Previous conversation context:
        {context}
        
        If this is the first question, ask about their motivation and background.
        For follow-up questions, analyze the candidate's previous responses and ask relevant
        questions that dig deeper into their potential, soft skills, and cultural fit.
        
        Generate a single clear, focused question that helps evaluate the candidate.
        Return ONLY the question text, nothing else.""",
        expected_output="A clear, relevant interview question based on the context and previous responses.",
        agent=interviewer
    )

def create_answer_task(question: str, candidate) -> Task:
    """Creates a task for generating candidate responses"""
    return Task(
        description=f"""You are answering this interview question: '{question}'
        
        Provide a thoughtful, honest response that:
        - Demonstrates your enthusiasm and potential
        - Draws from your academic experiences when relevant
        - Shows your willingness to learn and grow
        - Maintains professionalism while being authentic
        
        Return ONLY your response to the question, nothing else.""",
        expected_output="A professional, thoughtful response to the interview question.",
        agent=candidate
    )

def create_evaluation_task(job_title: str, interviewer, interview_history: List) -> Task:
    """Creates a task for evaluating the candidate"""
    context = "\n".join([
        f"Q{i+1}: {interaction['question']}\nA{i+1}: {interaction['answer']}"
        for i, interaction in enumerate(interview_history)
    ])
    
    return Task(
        description=f"""Based on the full interview conversation:
        
        {context}
        
        Provide a comprehensive evaluation of the candidate for the {job_title} position.
        Include:
        1. Overall Decision (PASS/FAIL)
        2. Score (0-100)
        3. Key Strengths Demonstrated
        4. Areas for Improvement
        5. Specific Tips for Future Interviews
        6. Detailed Reasoning for the Decision
        
        Format the response clearly with headers and bullet points.""",
        expected_output="A detailed evaluation report including pass/fail decision, score, strengths, areas for improvement, and specific recommendations.",
        agent=interviewer
    )

def create_comparative_analysis_task(job_title: str, interviewer, interview_results: dict, models: dict) -> Task:
    """Creates a task for comparative analysis of all candidates"""
    return Task(
        description=f"""Based on the interviews conducted with candidates for the {job_title} position using different LLM models, provide a comparative analysis:
        
        {'\n\n'.join([f"Candidate {i+1} (Model: {models[f'candidate{i+1}']}): \n{results['evaluation']}" 
                      for i, (_, results) in enumerate(interview_results.items())])}

        Please provide:
        1. Comparative scoring and ranking
        2. Notable differences in response patterns
        3. Strengths and weaknesses of each model's responses
        4. Overall recommendations for model selection for interview simulations
        
        Format the response clearly with headers and bullet points.""",
        expected_output="A detailed comparative analysis of the LLM models' performance in the interview simulation.",
        agent=interviewer
    )