# # # from fastapi import FastAPI # # # from pydantic import BaseModel # # # import pickle # # # import pandas as pd # # # from preprocessing import preprocess_input # # # app = FastAPI() # # # # Load model # # # with open("app/models/model.pkl", "rb") as f: # # # model = pickle.load(f) # # # # Load encoders + scaler # # # with open("app/models/encoders.pkl", "rb") as f: # # # encoders, scaler, mean_opv_doses, final_feature_order, df_temp_prep = pickle.load(f) # # # # Mapping # # # class_map = {0: "Positive", 1: "Negative", 2: "NPENT"} # # # # Input schema # # # class PredictRequest(BaseModel): # # # Province: str # # # District: str # # # Case_Contact: str # # # OPV_Doses: str # # # Specimen_Number: str # # # Specimen_condition: str # # # Body_Temp: str # # # Sore_Throat: str # # # Fatigue: str # # # Limb_Discomfort: str # # # @app.post("/predict") # # # def predict(req: PredictRequest): # # # raw = { # # # "Province": req.Province, # # # "District": req.District, # # # "Case/Contact": req.Case_Contact, # # # "OPV Doses": req.OPV_Doses, # # # "Specimen Number": req.Specimen_Number, # # # "Specimen condition": req.Specimen_condition, # # # "Body Temp": req.Body_Temp, # # # "Sore Throat": req.Sore_Throat, # # # "Fatigue": req.Fatigue, # # # "Limb Discomfort": req.Limb_Discomfort, # # # } # # # X = preprocess_input(raw, encoders, scaler, mean_opv_doses, final_feature_order, df_temp_prep) # # # pred_enc = model.predict(X) # # # pred_class = class_map[int(pred_enc.item())] # # # return {"prediction": pred_class} # # # app/main.py # # from fastapi import FastAPI, HTTPException # # from fastapi.middleware.cors import CORSMiddleware # # from pydantic import BaseModel # # import pickle # # import os # # from app.preprocessing import preprocess_input # # # , encoder_province, encoder_district # # APP_ROOT = os.path.dirname(os.path.abspath(__file__)) # # MODELS_DIR = os.path.join(APP_ROOT, "models") # # # Paths (adjust filenames if you used different names) # # LABEL_ENCODERS_PATH = os.path.join(MODELS_DIR, "label_encoders.pkl") # # SCALER_PATH = os.path.join(MODELS_DIR, "scaler.pkl") # # MODEL_PATH = os.path.join(MODELS_DIR, "catboost_model.pkl") # # # Load artifacts # # with open(LABEL_ENCODERS_PATH, "rb") as f: # # label_encoders = pickle.load(f) # # with open(SCALER_PATH, "rb") as f: # # scaler = pickle.load(f) # # with open(MODEL_PATH, "rb") as f: # # model = pickle.load(f) # # # Derive readable class names from your target encoder # # le_target = label_encoders.get("Final_cell_culture_result") # # class_map = {i: name for i, name in enumerate(le_target.classes_)} # # # Choose a sensible mean OPV doses (if you have real mean, replace here) # # mean_opv_doses = 2.0 # # # final_feature_order must match the training feature order: # # # Replace or adjust if you used different order during training # # final_feature_order = ['Province', 'District', 'Case/Contact', 'OPV Doses', # # 'Specimen Number', 'Specimen condition', 'Body Temp', # # 'Sore Throat', 'Fatigue', 'Limb Discomfort'] # # app = FastAPI(title="CatBoost Prediction API") # # # Allow requests from localhost and your Flutter dev server (adjust origins as needed) # # app.add_middleware( # # CORSMiddleware, # # allow_origins=["*"], # in production lock this down to your domains # # allow_credentials=True, # # allow_methods=["*"], # # allow_headers=["*"], # # ) # # # Pydantic request model (snake/camel mapping kept simple) # # class PredictRequest(BaseModel): # # age_in_months: float # # province: str # # district: str # # case_or_contact: str # # specimen_cond: str # # date_of_onset: str # # date_of_collection: str # # date_received: str # # opv_doses_received: float # # sore_throat: str # # fatigue: str # # limb_discomfort: str # # @app.post("/predict") # # def predict(req: PredictRequest): # # # Build raw_input keys matching preprocess_input expected keys # # raw = { # # "Province": req.Province, # # "District": req.District, # # "Case/Contact": req.Case_Contact or "Case", # # "OPV Doses": req.OPV_Doses, # # "Specimen Number": req.Specimen_Number, # # "Specimen condition": req.Specimen_condition or "Good", # # "Body Temp": req.Body_Temp, # # "Sore Throat": req.Sore_Throat, # # "Fatigue": req.Fatigue, # # "Limb Discomfort": req.Limb_Discomfort, # # } # # try: # # X = preprocess_input(raw, label_encoders, scaler, mean_opv_doses, final_feature_order) # # except Exception as e: # # raise HTTPException(status_code=400, detail=f"Preprocessing failed: {e}") # # try: # # # model.predict should accept a DataFrame; if not adjust to numpy # # pred_enc = model.predict(X) # # pred_index = int(pred_enc.item()) if hasattr(pred_enc, "item") else int(pred_enc) # # pred_label = class_map.get(pred_index, str(pred_index)) # # except Exception as e: # # raise HTTPException(status_code=500, detail=f"Prediction failed: {e}") # # return {"prediction": pred_label} # # main.py # from fastapi import FastAPI # from pydantic import BaseModel # import pickle # from app.preprocessing import preprocess_input # app = FastAPI(title="Polio Prediction API") # # Load trained ML model # with open("app/models/model.pkl", "rb") as f: # model = pickle.load(f) # # ---------- CLEAN API INPUT MODEL ---------- # class PredictionInput(BaseModel): # province: str # district: str # case_contact: str # opv_doses: str # specimen_number: str # specimen_condition: str # body_temp: str # sore_throat: str # fatigue: str # limb_discomfort: str # # ---------- API ENDPOINT ---------- # @app.post("/predict") # def predict(data: PredictionInput): # # Convert API field names → original model field names # raw_input = { # "Province": data.province, # "District": data.district, # "Case/Contact": data.case_contact, # "OPV Doses": data.opv_doses, # "Specimen Number": data.specimen_number, # "Specimen condition": data.specimen_condition, # "Body Temp": data.body_temp, # "Sore Throat": data.sore_throat, # "Fatigue": data.fatigue, # "Limb Discomfort": data.limb_discomfort # } # # Preprocess # processed_df = preprocess_input(raw_input) # # Predict # prediction = model.predict(processed_df)[0] # probability = model.predict_proba(processed_df)[0].max() # return { # "prediction": int(prediction), # "probability": float(probability) # } # @app.get("/") # def home(): # return {"message": "Polio Prediction API running."} from fastapi import FastAPI from fastapi.middleware.cors import CORSMiddleware from app.schemas import SampleInput from app.load_artifacts import ( loaded_catboost_model, loaded_label_encoders, loaded_scaler, class_names_readable ) from app.preprocessing import preprocess_single_input import pandas as pd app = FastAPI( title="Poliovirus Detection API", description="Serve CatBoost model predictions via API", version="1.0.0", ) # ✅ Add CORS middleware here app.add_middleware( CORSMiddleware, allow_origins=["*"], # Allow all origins (Flutter mobile + web) allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) @app.post("/predict") def predict(input_data: SampleInput): if loaded_catboost_model is None or loaded_label_encoders is None or loaded_scaler is None: return {"error": "Model or preprocessing components not loaded"} raw_input_dict = input_data.dict() processed_df = preprocess_single_input( raw_input_dict=raw_input_dict ) prediction_encoded = loaded_catboost_model.predict(processed_df) prediction_class = class_names_readable[int(prediction_encoded.item())] return { "prediction": prediction_class, "processed_input": processed_df.to_dict(orient="records")[0] }