feat: cleanup

README.md

@@ -1,6 +1,6 @@
 # CaféCast ☕📊
 
-**CaféCast** is an AI-powered sales forecasting and product analysis application. This project focuses on testing and enhancing my knowledge of machine learning models, hyperparameter fine-tuning, and model interpretability. It serves as a learning platform to explore advanced techniques, refine my skills, and deepen my understanding of how to apply machine learning to complex real-world problems. The main goal is to practice and improve while exploring the capabilities of various machine learning models and methodologies.
+**CaféCast** is a data-focused platform designed to help users explore and analyze datasets using AI-driven insights. This project currently focuses on manual data cleaning, shaping, and model training workflows, with plans to evolve into a fully interactive and user-friendly platform.
 
 ---
 
@@ -10,17 +10,15 @@
   - Supports **TensorFlow (CPU)** for macOS users to ensure compatibility and efficient local execution.
   - Utilizes **PyTorch (CUDA)** for Windows users with GPU support for accelerated training and predictions.
   - Automatic detection of the operating system to ensure the correct model implementation is selected.
-- **Broad Forecasting Capabilities:**
-  - Handles **multiple concurrent predictions**, enabling comprehensive sales and demand insights.
-  - Adaptable to a wide range of café datasets and forecasting requirements.
-- **Sales Forecasting Models:**
+- **Data Cleaning and Shaping:**
+  - Manual workflows for preparing datasets for analysis.
+- **Model Training:**
   - Advanced LSTM-based models with Bayesian optimization and iterative tuning for precise forecasts.
   - Time Series Transformers for capturing complex temporal dependencies.
   - Classical ARIMA modeling for interpretable, short-term linear trend forecasts.
 - **Enhanced Model Flexibility:**
   - Configurable hyperparameters to suit the unique characteristics of each dataset.
   - Dynamic support for both platform-optimized and manually fine-tuned workflows.
-- **Explainable AI:** SHAP values and attention mechanisms provide transparency into model decisions.
 - **Data Visualization:** Visual insights into sales trends, seasonality, and demand patterns.
 
 ---
@@ -83,6 +81,40 @@ This project is a testament to continuous learning and experimentation in the fi
 
 ---
 
+## Future Expansions 🚀
+
+CafeCast is set to evolve into a more interactive and user-friendly platform with the following key features:
+
+### 1. **User-Focused Frontend Deployment**
+
+- A dedicated web-based interface for seamless interaction.
+- Users can upload datasets directly through the frontend, removing the need for manual backend processes.
+- A clean and intuitive design that prioritizes simplicity and accessibility.
+- Technologies considered include **React**, **Next.js**, or other modern web frameworks for dynamic and responsive user experiences.
+
+### 2. **AI-Driven Q&A Functionality**
+
+- Users will be able to ask questions about their uploaded data through the frontend.
+- The platform will leverage pre-trained models saved on the backend to provide accurate and meaningful answers.
+- Natural language processing (NLP) capabilities will interpret user queries and deliver real-time responses.
+- Support for follow-up queries to explore datasets more deeply.
+
+### 3. **Automated Data Cleaning and Preprocessing**
+
+- Automate the current manual data cleaning workflow.
+- The platform will:
+  - Validate datasets upon upload.
+  - Detect and resolve missing values, outliers, and inconsistencies automatically.
+  - Present a preview of cleaned data for user confirmation.
+
+### 4. **Model Training and Customization**
+
+- Allow users to train or fine-tune models directly through the platform using their uploaded datasets.
+- Include visualization tools for monitoring training progress, accuracy, and performance metrics.
+- Ensure that users can easily customize their models to suit specific use cases.
+
+---
+
 ## Tech Stack 🛠️
 
 - **Programming Language:** Python

api.py

@@ -0,0 +1,98 @@
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+import numpy as np
+import joblib
+import tensorflow as tf
+import torch
+
+app = FastAPI()
+
+# Paths for models and scaler
+SCALER_PATH = "models/scaler.pkl"
+LSTM_MODEL_PATH = "models/scaler_lstm.pkl"
+LSTM_BAYESIAN_MODEL_PATH = "models/scaler_lstm_bayesian.pkl"
+TRANSFORMER_MODEL_PATH = "models/scaler_ts_transformer.pkl"
+
+# Globals for models and scaler
+scaler = None
+lstm_model = None
+lstm_bayesian_model = None
+transformer_model = None
+
+@app.on_event("startup")
+def load_models_and_scaler():
+    global scaler, lstm_model, lstm_bayesian_model, transformer_model
+
+    # Load scaler
+    try:
+        with open(SCALER_PATH, "rb") as f:
+            scaler = joblib.load(f)
+        print("Scaler loaded successfully.")
+    except FileNotFoundError:
+        raise HTTPException(status_code=500, detail="Scaler file not found.")
+
+    # Load LSTM model
+    try:
+        with open(LSTM_MODEL_PATH, "rb") as f:
+            lstm_model = joblib.load(f)
+        print("LSTM model loaded successfully.")
+    except Exception as e:
+        print(f"Error loading LSTM model: {e}")
+
+    # Load Bayesian LSTM model
+    try:
+        with open(LSTM_BAYESIAN_MODEL_PATH, "rb") as f:
+            lstm_bayesian_model = joblib.load(f)
+        print("Bayesian LSTM model loaded successfully.")
+    except Exception as e:
+        print(f"Error loading Bayesian LSTM model: {e}")
+
+    # Load Transformer model
+    try:
+        with open(TRANSFORMER_MODEL_PATH, "rb") as f:
+            transformer_model = torch.load(f, map_location="cpu")
+            print(transformer_model.state_dict().keys())
+            transformer_model.eval()
+        print("Transformer model loaded successfully.")
+    except Exception as e:
+        print(f"Error loading Transformer model: {e}")
+
+# Request model
+class PredictionRequest(BaseModel):
+    model: str  # 'lstm', 'lstm_bayesian', 'transformer'
+    input_values: list  # Raw input values
+
+@app.post("/predict")
+def predict(data: PredictionRequest):
+    if scaler is None:
+        raise HTTPException(status_code=500, detail="Scaler not loaded.")
+
+    # Preprocess input
+    try:
+        input_array = np.array(data.input_values).reshape(1, -1)
+        scaled_input = scaler.transform(input_array)
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"Invalid input: {e}")
+
+    # Predict based on model
+    if data.model == "lstm":
+        if lstm_model:
+            prediction = lstm_model.predict(scaled_input)
+        else:
+            raise HTTPException(status_code=500, detail="LSTM model not loaded.")
+    elif data.model == "lstm_bayesian":
+        if lstm_bayesian_model:
+            prediction = lstm_bayesian_model.predict(scaled_input)
+        else:
+            raise HTTPException(status_code=500, detail="Bayesian LSTM model not loaded.")
+    elif data.model == "transformer":
+        if transformer_model:
+            with torch.no_grad():
+                tensor_input = torch.tensor(scaled_input, dtype=torch.float32)
+                prediction = transformer_model(tensor_input).numpy()
+        else:
+            raise HTTPException(status_code=500, detail="Transformer model not loaded.")
+    else:
+        raise HTTPException(status_code=400, detail="Invalid model type.")
+
+    return {"model": data.model, "prediction": prediction.tolist()}