app.py

# 1. IMPORTS AND SETUP
import os
import uuid
import gradio as gr
from typing import TypedDict, List, Optional
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.tools import tool
from pydantic import BaseModel, Field # CORRECTED LINE: Import directly from Pydantic
from langgraph.graph import StateGraph, END
import requests
from bs4 import BeautifulSoup
from pypdf import PdfReader # CORRECTED LINE: Import from 'pypdf' instead of 'PyPDF2'
from threading import Thread

print("--- Libraries imported. ---")

# Set API keys from Hugging Face Space Secrets
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
os.environ["LANGCHAIN_API_KEY"] = os.getenv("LANGCHAIN_API_KEY")
os.environ["TAVILY_API_KEY"] = os.getenv("TAVILY_API_KEY")
os.environ["LANGCHAIN_TRACING_V2"] = "true"
os.environ["LANGCHAIN_PROJECT"] = "Deployed Career Navigator"


# 2. LANGGRAPH AGENT BACKEND (The "Brain" of the App)

class SkillAnalysis(BaseModel):
    technical_skills: List[str] = Field(description="List of top 5 technical skills.")
    soft_skills: List[str] = Field(description="List of top 3 soft skills.")

class ResumeFeedback(BaseModel):
    strengths: List[str] = Field(description="Resume strengths.")
    gaps: List[str] = Field(description="Missing skills or experiences.")
    suggestions: List[str] = Field(description="Suggestions for improvement.")

class CareerActionPlan(BaseModel):
    career_overview: str = Field(description="Overview of the chosen career.")
    skill_analysis: SkillAnalysis
    resume_feedback: ResumeFeedback
    learning_roadmap: str = Field(description="Markdown-formatted learning plan.")
    portfolio_plan: str = Field(description="Markdown-formatted portfolio project plan.")

class TeamState(TypedDict):
    student_interests: str
    student_resume: str
    career_options: List[str]
    chosen_career: str
    market_analysis: Optional[SkillAnalysis]
    resume_analysis: Optional[ResumeFeedback]
    final_plan: Optional[CareerActionPlan]

@tool
def scrape_web_content(url: str) -> str:
    """Scrapes text content from a URL."""
    try:
        response = requests.get(url, timeout=10)
        soup = BeautifulSoup(response.content, 'html.parser')
        return soup.get_text(separator=' ', strip=True)[:10000]
    except requests.RequestException:
        return "Error: Could not scrape the URL."

llm = ChatOpenAI(model="gpt-4o", temperature=0)

def job_market_analyst_agent(state: TeamState):
    print("--- 🕵️ Agent: Job Market Analyst ---")
    structured_llm = llm.with_structured_output(SkillAnalysis)
    prompt = ChatPromptTemplate.from_template(
        "You are an expert job market analyst. Based on the career of '{career}', identify the top 5 technical skills and top 3 soft skills required."
    )
    chain = prompt | structured_llm
    analysis = chain.invoke({"career": state['chosen_career']})
    return {"market_analysis": analysis}

def resume_reviewer_agent(state: TeamState):
    print("--- 📝 Agent: Resume Reviewer ---")
    structured_llm = llm.with_structured_output(ResumeFeedback)
    prompt = ChatPromptTemplate.from_messages([
        ("system", "You are an expert career coach. Compare the user's resume with the provided analysis of in-demand skills and provide feedback."),
        ("human", "User's Resume:\n{resume}\n\nRequired Skills Analysis:\n{skill_analysis}")
    ])
    chain = prompt | structured_llm
    feedback = chain.invoke({
        "resume": state["student_resume"],
        "skill_analysis": state["market_analysis"].dict()
    })
    return {"resume_analysis": feedback}

def lead_agent_node(state: TeamState):
    print("--- 👑 Agent: Lead Agent (Synthesizing & Planning) ---")
    structured_llm = llm.with_structured_output(CareerActionPlan)
    prompt = ChatPromptTemplate.from_template(
        "You are the lead career strategist. Synthesize all the provided information into a comprehensive Career Action Plan. "
        "Create a detailed 8-week learning roadmap and suggest 3 portfolio projects.\n\n"
        "Chosen Career: {career}\n"
        "Required Skills: {skills}\n"
        "Resume Feedback: {resume_feedback}"
    )
    chain = prompt | structured_llm
    final_plan = chain.invoke({
        "career": state["chosen_career"],
        "skills": state["market_analysis"].dict(),
        "resume_feedback": state["resume_analysis"].dict()
    })
    return {"final_plan": final_plan}

graph_builder = StateGraph(TeamState)
graph_builder.add_node("analyze_market", job_market_analyst_agent)
graph_builder.add_node("review_resume", resume_reviewer_agent)
graph_builder.add_node("create_final_plan", lead_agent_node)
graph_builder.set_entry_point("analyze_market")
graph_builder.add_edge("analyze_market", "review_resume")
graph_builder.add_edge("review_resume", "create_final_plan")
graph_builder.add_edge("create_final_plan", END)
navigator_agent = graph_builder.compile()

print("--- LangGraph Agent Backend is ready. ---")


# 3. HELPER FUNCTIONS FOR GRADIO
def extract_text_from_pdf(pdf_file_obj):
    if not pdf_file_obj:
        return "", "Please upload a resume to begin."
    try:
        reader = PdfReader(pdf_file_obj.name)
        text = "".join(page.extract_text() or "" for page in reader.pages)
        if not text.strip():
            return "", "Error: Could not extract text from the PDF. Please try a different file."
        return text, ""
    except Exception as e:
        return "", f"An error occurred while reading the PDF: {e}"


def run_agent_and_update_ui(resume_text, chosen_career):
    if not resume_text:
        return {
            output_col: gr.update(visible=True),
            output_overview: gr.update(value="<h3 style='color:red;'>Please upload a resume first.</h3>", visible=True)
        }
        
    yield {
        output_col: gr.update(visible=True),
        input_row: gr.update(visible=False),
        output_overview: gr.update(value="### 🧠 The AI agent team is analyzing your profile...", visible=True)
    }

    initial_state = { "student_resume": resume_text, "chosen_career": chosen_career }
    final_state = navigator_agent.invoke(initial_state)
    plan = final_state['final_plan']

    # Final update
    yield {
        output_plan_state: plan,
        output_overview: gr.update(value=f"## 1. Career Overview: {plan.career_overview}"),
        output_skills: gr.update(
            value=f"## 2. Job Market Skill Analysis\n**Top Technical Skills:** {', '.join(plan.skill_analysis.technical_skills)}\n\n**Top Soft Skills:** {', '.join(plan.skill_analysis.soft_skills)}",
            visible=True
        ),
        output_resume_feedback: gr.update(
            value=f"## 3. Your Resume Feedback\n**Strengths:** {' '.join(plan.resume_feedback.strengths)}\n\n**Gaps to Fill:** {', '.join(plan.resume_feedback.gaps)}\n\n**Suggestions:**\n" + "\n".join([f"- {s}" for s in plan.resume_feedback.suggestions]),
            visible=True
        ),
        output_learning_plan: gr.update(value=f"## 4. Your 8-Week Learning Roadmap\n{plan.learning_roadmap}", visible=True),
        output_portfolio_plan: gr.update(value=f"## 5. Your Portfolio Project Plan\n{plan.portfolio_plan}", visible=True),
        chat_row: gr.update(visible=True)
    }

def chat_with_agent(message, history, plan_state):
    if not plan_state:
        return "An error occurred. Please generate a new plan."
    
    prompt = ChatPromptTemplate.from_messages([
        ("system", "You are a helpful career coach. The user has just received the following career action plan. Answer their follow-up questions based ONLY on this plan.\n\n--- CAREER PLAN ---\n{plan_text}"),
        ("user", "{user_question}")
    ])
    chat_chain = prompt | llm
    
    plan_text = f"Career: {plan_state.career_overview}\nSkills: {plan_state.skill_analysis.dict()}\nResume Feedback: {plan_state.resume_feedback.dict()}\nLearning Plan: {plan_state.learning_roadmap}\nPortfolio Plan: {plan_state.portfolio_plan}"
    
    response = chat_chain.invoke({"plan_text": plan_text, "user_question": message})
    return response.content


# 4. GRADIO UI DEFINITION
with gr.Blocks(theme=gr.themes.Soft(), css="footer {visibility: hidden}") as demo:
    output_plan_state = gr.State()
    resume_text_state = gr.State()
    error_text_state = gr.State()
    
    gr.Markdown("# 🚀 Your AI Career Navigator")
    gr.Markdown("Upload your resume, select a target career, and get a personalized, data-driven action plan from a team of AI agents.")

    with gr.Row(visible=True) as input_row:
        with gr.Column(scale=2):
            input_pdf_resume = gr.File(label="Upload Your Resume (PDF)", file_types=[".pdf"])
            input_career_choice = gr.Dropdown(
                label="Select Your Target Career",
                choices=["Data Analyst", "Software Engineer", "Product Manager", "UX Designer", "AI/ML Engineer"],
                value="Data Analyst"
            )
        with gr.Column(scale=1, min_width=200):
            submit_button = gr.Button("Generate My Action Plan", variant="primary", scale=2)

    with gr.Column(visible=False) as output_col:
        output_overview = gr.Markdown(visible=False)
        output_skills = gr.Markdown(visible=False)
        output_resume_feedback = gr.Markdown(visible=False)
        output_learning_plan = gr.Markdown(visible=False)
        output_portfolio_plan = gr.Markdown(visible=False)

    with gr.Row(visible=False) as chat_row:
         chat_interface = gr.ChatInterface(
            chat_with_agent,
            chatbot=gr.Chatbot(height=500, label="Chat with your Career Coach"),
            additional_inputs=[output_plan_state],
            title="Ask Follow-up Questions",
            description="Ask any questions about your generated plan."
        )

    # Event Handling Logic
    input_pdf_resume.upload(
        fn=extract_text_from_pdf,
        inputs=[input_pdf_resume],
        outputs=[resume_text_state, error_text_state]
    )
    
    submit_button.click(
        fn=run_agent_and_update_ui,
        inputs=[resume_text_state, input_career_choice],
        outputs=[
            output_plan_state,
            output_overview,
            output_skills,
            output_resume_feedback,
            output_learning_plan,
            output_portfolio_plan,
            input_row,
            output_col,
            chat_row
        ]
    )

if __name__ == "__main__":
    demo.launch(debug=True)