graph.py

from langgraph.graph import StateGraph
from langchain_groq import ChatGroq
from langchain.prompts import PromptTemplate
from typing import TypedDict
import os

# Load API key from Hugging Face Secret later
import os
llm = ChatGroq(temperature=0, model="llama3-70b-8192")

class ResumeState(TypedDict):
    resume_text: str
    jd_text: str
    summary: str
    result: str
    reasoning: str

def summarize_node(state: ResumeState) -> ResumeState:
    prompt = PromptTemplate.from_template("Summarize this resume:\n{resume}")
    summary = llm.invoke(prompt.format(resume=state["resume_text"])).content
    state["summary"] = summary
    return state

def jd_match_node(state: ResumeState) -> ResumeState:
    prompt = PromptTemplate.from_template(
        "Given the resume and job description, classify the candidate as Suitable or Not Suitable. Explain your reasoning.\n\nResume:\n{resume}\n\nJob Description:\n{jd}"
    )
    response = llm.invoke(prompt.format(resume=state["resume_text"], jd=state["jd_text"])).content.strip()

    if "not suitable" in response.lower():
        result = "Not Suitable"
    elif "suitable" in response.lower():
        result = "Suitable"
    else:
        result = "Unclear"

    state["result"] = result
    state["reasoning"] = response
    return state

def build_graph():
    builder = StateGraph(ResumeState)
    builder.add_node("summarize", summarize_node)
    builder.add_node("jd_match", jd_match_node)

    builder.set_entry_point("summarize")
    builder.add_edge("summarize", "jd_match")
    builder.set_finish_point("jd_match")

    return builder.compile()