MCP Resume Scorer with Leaderboard

from fastapi import FastAPI, UploadFile, Form, File, Query, HTTPException from fastapi.middleware.cors import CORSMiddleware from sqlalchemy import create_engine, Column, String, Float, Table, MetaData import google.generativeai as genai import io from pdfminer.high_level import extract_text import re import traceback import logging # --- Logging Setup --- logging.basicConfig(level=logging.INFO) # --- FastAPI Setup --- app = FastAPI() app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"], ) # --- Gemini API Setup --- genai.configure(api_key="AIzaSyArFO5d86n2kyCvdwZ9WdaBBlCywiaAbQA") # 🔑 Replace with your actual key # --- SQLite Setup --- engine = create_engine("sqlite:///leaderboard.db") metadata = MetaData() leaderboard = Table( "leaderboard", metadata, Column("name", String, primary_key=True), Column("score", Float), Column("feedback", String), ) metadata.create_all(engine) # --- Resume Scoring Function --- def score_resume_with_gemini(resume_text: str, target_role: str) -> tuple[float, str]: prompt = f""" You are a resume evaluator. Score the following resume for the role '{target_role}'. Give a score out of 100 and a brief feedback. Resume: {resume_text} """ model = genai.GenerativeModel(model_name="gemini-1.5-flash") response = model.generate_content(prompt) content = response.text logging.info("Gemini response: %s", content) match = re.search(r"(\d{1,3})", content) score = float(match.group(1)) if match else 0.0 return score, content # --- Upload and Verify Resume --- @app.post("/tools/verify_resume") async def verify_resume( name: str = Form(...), target_role: str = Form(...), file: UploadFile = File(...), debug: bool = Query(False) ): try: resume_bytes = await file.read() resume_text = extract_text(io.BytesIO(resume_bytes)) if not resume_text.strip(): raise ValueError("Resume text could not be extracted. Please upload a valid PDF.") score, feedback = score_resume_with_gemini(resume_text, target_role) with engine.connect() as conn: conn.execute(leaderboard.insert().values(name=name, score=score, feedback=feedback)) return {"name": name, "score": score, "feedback": feedback} except Exception as e: logging.error("Error during resume verification", exc_info=True) if debug: tb = traceback.format_exc() return {"error": str(e), "traceback": tb} raise HTTPException(status_code=500, detail=str(e)) # --- Leaderboard Endpoint --- @app.get("/tools/leaderboard") async def get_leaderboard(): with engine.connect() as conn: result = conn.execute(leaderboard.select()).fetchall() sorted_entries = sorted(result, key=lambda x: x.score, reverse=True) return [{"name": r.name, "score": r.score, "feedback": r.feedback} for r in sorted_entries[:10]]