add new

.gitignore

@@ -0,0 +1,2 @@
+__pycache__
+venv

Dockerfile

@@ -0,0 +1,20 @@
+# Use an official Python runtime as a parent image
+FROM python:3.9-slim
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Copy the requirements file and install dependencies first to leverage caching
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy all of your project files from the repository into the container
+COPY . .
+
+# Tell Docker that the container listens on port 7860
+# Hugging Face Spaces expects applications to run on this port
+EXPOSE 7860
+
+# Define the command to run your app
+# This assumes your main file is `main.py` and the FastAPI variable is `app`
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

GEO_CACHE.txt

@@ -0,0 +1 @@
+{"jakarta": [-6.1754049, 106.827168], "jayapura": [-2.5387539, 140.7037389], "gorontalo": [0.7186174, 122.4555927], "pekanbaru": [0.5262455, 101.4515727], "pangkal pinang": [-2.1206733, 106.1134604], "tomohon": [1.3255914, 124.838605], "administrasi jakarta pusat": [-6.18234, 106.8428715], "parepare": [-4.0202504, 119.6611703], "gunungsitoli": [1.2900569, 97.6150768], "dumai": [1.6631967, 101.4470369], "batu": [-7.8711667, 112.5269482], "solok": [-0.93103, 100.804328], "malang": [-7.9771308, 112.6340265], "samarinda": [-0.5017804, 117.1393089], "kendari": [-3.9918068, 122.5180066], "semarang": [-6.9903988, 110.4229104], "bogor": [-6.5962986, 106.7972421], "padangpanjang": [-0.4654636, 100.3932441], "palu": [-0.9051548, 119.8722373], "tebing tinggi": [3.3273686, 99.1623025], "langsa": [4.4730892, 97.9681841], "bitung": [1.44344, 125.1940836], "lubuklinggau": [-3.2919136, 102.8715732], "kupang": [-10.1632209, 123.6017755], "pariaman": [-0.6263889, 100.1177778], "binjai": [3.6063964, 98.4899865], "manado": [1.4900578, 124.8408708], "surabaya": [-7.2462836, 112.7377674], "mataram": [-8.5837726, 116.10685], "bukittinggi": [-0.3051954, 100.3694921], "blitar": [-8.1311983, 112.3115572], "metro": [-5.1078839, 105.3078642], "subulussalam": [2.6403146, 98.0052459], "prabumulih": [-3.4382688, 104.2310181], "bekasi": [-6.2349858, 106.9945444], "tangerang": [-6.1761924, 106.6382161], "kotamobagu": [0.7352231, 124.3154057], "tanjung pinang": [0.9236915, 104.446094], "tanjungbalai": [2.9703419, 99.8020903], "pekalongan": [-6.8905065, 109.6761489], "palopo": [-2.9996306, 120.1920679], "bau-bau": [-3.5671, 120.3863], "medan": [3.5894617, 98.6741623], "serang": [-6.1169662, 106.1518079], "singkawang": [0.9069861, 108.9889657], "sabang": [5.8927453, 95.3225751], "administrasi jakarta timur": [-6.2628908, 106.8822289], "tasikmalaya": [-7.3262484, 108.2201154], "administrasi jakarta barat": [-6.168982, 106.7895028], "yogyakarta": [-7.8012646, 110.3646857], "tangerang selatan": [-6.3227016, 106.7085737], "banjarmasin": [-3.3187496, 114.5925828], "kediri": [-7.8111057, 112.0046051], "bima": [-8.5647631, 118.762474], "madiun": [-7.6290837, 111.5168819], "palangka raya": [-2.2072919, 113.9164372], "jambi": [-1.6394711, 102.9454264], "palembang": [-2.9888243, 104.7568507], "tarakan": [3.3000169, 117.6330159], "padang": [-0.9247587, 100.3632561], "sukabumi": [-6.9199289, 106.9265095], "cirebon": [-6.7137044, 108.5608483], "payakumbuh": [-0.2242687, 100.6319419], "ambon": [-3.6959434, 128.178785], "mojokerto": [-7.5413378, 112.5094975], "sorong": [-0.8634105, 131.2544805], "tual": [-5.6389933, 132.7429249], "makassar": [-5.1342962, 119.4124282], "sawahlunto": [-0.6818141, 100.778552], "batam": [1.1030815, 104.0383696], "lhokseumawe": [5.1789659, 97.1480544], "sungai penuh": [-2.0706799, 101.3961054], "bandar lampung": [-5.4460713, 105.2643742], "denpasar": [-8.6524973, 115.2191175], "bandung": [-6.9218457, 107.6070833], "ternate": [0.7852043, 127.3832342], "bontang": [0.1236548, 117.471708], "administrasi jakarta utara": [-6.136197, 106.9006902], "probolinggo": [-7.7441461, 113.2158401], "banjarbaru": [-3.4430389, 114.8308816], "bengkulu": [-3.5186763, 102.5359834], "magelang": [-7.4770747, 110.2182164], "tegal": [-6.8674488, 109.1378271]}

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Kevin Bramasta Arvyto Wardhana
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

api/index.py

@@ -0,0 +1 @@
+from app.main import app

app/api/routes/recommendations.py

@@ -0,0 +1,15 @@
+from fastapi import APIRouter
+from ...schemas.recommendation import ProfileRequest, RecommendationRequest
+from ...services import ranking
+
+router = APIRouter()
+
+@router.post("/predict-category", tags=["Predictions"])
+def predict_category(request: ProfileRequest):
+    category = ranking.get_category_prediction(request.profile_text)
+    return {"predicted_category": category}
+
+@router.post("/recommend-internships", tags=["Predictions"])
+def recommend_internships(request: RecommendationRequest):
+    ranked_ids = ranking.get_ranked_internships(request)
+    return {"recommendations": ranked_ids}

app/core/models.py

@@ -0,0 +1,85 @@
+import joblib
+from sentence_transformers import SentenceTransformer
+import os
+import requests
+from huggingface_hub import snapshot_download # <-- Add this import
+
+TEMP_DIR = "/tmp"
+
+# This function for your .pkl files is still correct and needed
+def download_and_load_model(model_url, model_filename):
+    # ... (no changes needed in this function)
+    local_path = os.path.join(TEMP_DIR, model_filename)
+    if not os.path.exists(local_path):
+        print(f"Downloading model from {model_url} to {local_path}...")
+        try:
+            response = requests.get(model_url, stream=True)
+            response.raise_for_status()
+            with open(local_path, "wb") as f:
+                for chunk in response.iter_content(chunk_size=8192):
+                    f.write(chunk)
+            print("Download complete.")
+        except requests.exceptions.RequestException as e:
+            print(f"Failed to download model {model_filename}. Error: {e}")
+            return None
+    try:
+        return joblib.load(local_path)
+    except Exception as e:
+        print(f"Failed to load model {local_path}. Error: {e}")
+        return None
+
+# --- NEW FUNCTION FOR THE SENTENCE TRANSFORMER ---
+def get_sentence_transformer(model_name='all-MiniLM-L6-v2'):
+    """
+    Downloads the SentenceTransformer model to /tmp if it doesn't exist,
+    then loads it from there.
+    """
+    local_model_path = os.path.join(TEMP_DIR, model_name)
+
+    if not os.path.exists(local_model_path):
+        print(f"Downloading SentenceTransformer model '{model_name}' to {local_model_path}...")
+        # Use snapshot_download to get all files for the model from Hugging Face
+        try:
+            snapshot_download(repo_id=f"sentence-transformers/{model_name}",
+                              local_dir=local_model_path,
+                              local_dir_use_symlinks=False) # This is important for Vercel
+            print("Download complete.")
+        except Exception as e:
+            print(f"Failed to download SentenceTransformer model. Error: {e}")
+            return None
+
+    # Load the model from the local path in /tmp
+    try:
+        print(f"Loading SentenceTransformer model from {local_model_path}...")
+        return SentenceTransformer(local_model_path)
+    except Exception as e:
+        print(f"Failed to load SentenceTransformer model from {local_model_path}. Error: {e}")
+        return None
+
+# --- Main Model Loading Logic ---
+try:
+    # Your .pkl model loading remains the same
+    TFIDF_URL = "https://pub-4a389a9b2dc842a2a55678d2db0ec0c6.r2.dev/tfidf_vectorizer.pkl"
+    LE_URL = "https://pub-4a389a9b2dc842a2a55678d2db0ec0c6.r2.dev/label_encoder.pkl"
+    RF_MODEL_URL = "https://pub-4a389a9b2dc842a2a55678d2db0ec0c6.r2.dev/random_forest_model.pkl"
+
+    tfidf_vectorizer = download_and_load_model(TFIDF_URL, "tfidf_vectorizer.pkl")
+    le = download_and_load_model(LE_URL, "label_encoder.pkl")
+    rf_model = download_and_load_model(RF_MODEL_URL, "random_forest_model.pkl")
+
+    if all([tfidf_vectorizer, le, rf_model]):
+        print("Classification models loaded successfully.")
+    else:
+        print("One or more classification models failed to load.")
+
+    # --- THIS IS THE LINE TO CHANGE ---
+    # Old line: sentence_model = SentenceTransformer('all-MiniLM-L6-v2')
+    # New line:
+    sentence_model = get_sentence_transformer()
+    if sentence_model:
+        print("SentenceTransformer model loaded successfully.")
+    else:
+        print("SentenceTransformer model failed to load.")
+
+except Exception as e:
+    print(f"An unexpected error occurred during model loading: {e}")

app/main.py

@@ -0,0 +1,14 @@
+from fastapi import FastAPI
+from .api.routes import recommendations
+
+app = FastAPI(
+    title="Student Recommendation API",
+    description="An API that uses machine learning to predict job categories and recommend internships.",
+    version="1.0.0"
+)
+
+app.include_router(recommendations.router, prefix="/api/v1")
+
+@app.get("/", tags=["Root"])
+def read_root():
+    return {"message": "Welcome to the Student Recommendation API"}

app/schemas/recommendation.py

@@ -0,0 +1,16 @@
+from pydantic import BaseModel
+from typing import List, Optional
+
+class ProfileRequest(BaseModel):
+    profile_text: str
+
+class InternshipItem(BaseModel):
+    id: int
+    internship_text: str
+    location: str
+
+class RecommendationRequest(BaseModel):
+    profile_text: str
+    predicted_category: Optional[str] = None
+    preferred_location: str
+    internships: List[InternshipItem]

app/services/ranking.py

@@ -0,0 +1,121 @@
+from geopy.distance import geodesic
+import json
+import os
+from geopy.geocoders import Nominatim
+
+from ..core.models import tfidf_vectorizer, le, rf_model, sentence_model
+from ..schemas.recommendation import RecommendationRequest
+from ..utils.text import clean_resume
+
+geolocator = Nominatim(user_agent="student_recommendation_api_v1")
+
+GEO_CACHE_FILE = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "GEO_CACHE.txt")
+GEO_CACHE = {}
+
+def load_geo_cache():
+    global GEO_CACHE
+    if os.path.exists(GEO_CACHE_FILE):
+        with open(GEO_CACHE_FILE, "r") as f:
+            try:
+                GEO_CACHE = json.load(f)
+            except json.JSONDecodeError:
+                GEO_CACHE = {}
+
+def save_geo_cache():
+    with open(GEO_CACHE_FILE, "w") as f:
+        json.dump(GEO_CACHE, f)
+
+load_geo_cache()
+
+def geo_coords(city_name: str) -> tuple | None:
+    """
+    Geocodes a city name to (latitude, longitude).
+    Uses an in-memory cache to avoid repeated API calls.
+    """
+    city_name = city_name.lower().strip()
+    if city_name in GEO_CACHE:
+        return GEO_CACHE[city_name]
+    try:
+        print(f"--- Geocoding and caching new city: {city_name} ---")
+        location = geolocator.geocode(f"{city_name}, Indonesia")
+
+        if location:
+            coords = (location.latitude, location.longitude)
+            GEO_CACHE[city_name] = coords
+            save_geo_cache()
+            return coords
+        else:
+            print(f"Location not found for {city_name}")
+            GEO_CACHE[city_name] = None
+            save_geo_cache()
+            return None
+    except Exception as e:
+        print(f"Error geocoding {city_name}: {e}")
+        return None
+
+def get_category_prediction(profile_text: str) -> str:
+    """Processes text and predicts the job category."""
+    cleaned_text = profile_text.lower()
+    vectorized_text = tfidf_vectorizer.transform([cleaned_text])
+    prediction_encoded = rf_model.predict(vectorized_text)[0]
+    category = le.inverse_transform([prediction_encoded])[0]
+    return category
+
+def  get_ranked_internships(request: RecommendationRequest) -> list[int]:
+    """Performs two-stage ranking with dynamic geocoding."""
+
+    profile_text_to_encode = request.profile_text
+
+    if request.predicted_category:
+        profile_text_to_encode = f"The user's predicted job category is {request.predicted_category}. Based on that, consider their profile: {request.profile_text}"
+
+    profile_embedding = sentence_model.encode(profile_text_to_encode)
+    internship_texts = [internship.internship_text for internship in request.internships]
+
+    if not internship_texts:
+        return []
+
+    internship_embeddings = sentence_model.encode(internship_texts)
+    cosine_score = sentence_model.similarity(profile_embedding, internship_embeddings)[0].tolist()
+
+    print("--- FastAPI Debugging ---")
+    print(f"Received {len(internship_texts)} internships to rank.")
+    print(f"Calculated Cosine Scores: {cosine_score}")
+    print("--------------------------")
+
+    ranked_by_similarity = []
+    for i, internship in enumerate(request.internships):
+        ranked_by_similarity.append({
+            "id": internship.id,
+            "similarity_score": cosine_score[i],
+            "location": internship.location,
+        })
+
+    final_ranked_list = []
+    user_coords = geo_coords(request.preferred_location)
+
+    print(user_coords, request.preferred_location)
+
+
+    for internship in ranked_by_similarity:
+        final_score = internship['similarity_score']
+
+        if user_coords:
+            internship_coords = geo_coords(internship['location'])
+            if internship_coords:
+                distance_km = geodesic(user_coords, internship_coords).kilometers
+                if distance_km < 1:
+                    final_score += 2.0
+                elif distance_km < 150:
+                    final_score += 0.75
+
+        internship['final_score'] = final_score
+        final_ranked_list.append(internship)
+
+    final_ranked_list.sort(key=lambda x: x['final_score'], reverse=True)
+
+    final_ids = [item['id'] for item in final_ranked_list]
+
+    print(final_ranked_list)
+
+    return final_ids

app/utils/text.py

@@ -0,0 +1,17 @@
+import re
+
+def clean_resume(text: str) -> str:
+    """
+    Cleans the input resume text by removing URLs, special characters,
+    and extra whitespace, and converting to lowercase.
+    """
+    # Remove URLs
+    text = re.sub(r'http\S+|www\S+', '', text)
+    # Remove non-alphanumeric characters (keeps only letters and spaces)
+    text = re.sub(r'[^A-Za-z\s]', '', text)
+    # Convert to lowercase
+    text = text.lower()
+    # Remove extra whitespace
+    text = re.sub(r'\s+', ' ', text).strip()
+
+    return text

requirements.txt

@@ -0,0 +1,17 @@
+# Core FastAPI Framework
+fastapi
+uvicorn
+requests
+
+# Machine Learning & Data
+scikit-learn
+joblib
+sentence-transformers
+torch
+geopy
+huggingface-hub
+# Pydantic is a dependency of FastAPI, but we list it for clarity
+pydantic
+
+# Good practice for managing environment variables
+python-dotenv

vercel.json

@@ -0,0 +1,14 @@
+{
+  "functions": {
+    "api/index.py": {
+      "maxDuration": 60,
+      "memory": 3008
+    }
+  },
+  "routes": [
+    {
+      "src": "/(.*)",
+      "dest": "api/index.py"
+    }
+  ]
+}