app.py

import pandas as pd
from fuzzywuzzy import process
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import gradio as gr

# Load dataset from GitHub
url = "https://raw.githubusercontent.com/devbernie/Medimatch/refs/heads/main/dataset.csv"
data = pd.read_csv(url)

# Preprocess data
columns_to_process = ['Composition', 'Uses', 'Medicine Name']
for column in columns_to_process:
    data[column] = data[column].fillna('').str.lower()

# Fuzzy search function
def fuzzy_search(query, data_column, threshold=80):
    matches = process.extract(query.lower(), data_column, limit=5)
    results = [match[0] for match in matches if match[1] >= threshold]
    return results

# Vectorize data
vectorizer_composition = TfidfVectorizer()
composition_matrix = vectorizer_composition.fit_transform(data['Composition'])

vectorizer_uses = TfidfVectorizer()
uses_matrix = vectorizer_uses.fit_transform(data['Uses'])

# Calculate similarities
composition_similarity = cosine_similarity(composition_matrix)
uses_similarity = cosine_similarity(uses_matrix)
overall_similarity = (composition_similarity + uses_similarity) / 2

# Recommendation function
def recommend_medicine(input_name, top_n=5):
    matches = fuzzy_search(input_name, data['Medicine Name'])
    if not matches:
        return f"Không tìm thấy thuốc gần đúng với '{input_name}'. Vui lòng thử lại."
    
    selected_medicine = matches[0]
    idx = data[data['Medicine Name'] == selected_medicine].index[0]
    sim_scores = list(enumerate(overall_similarity[idx]))
    sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True)
    
    recommendations = []
    for i in sim_scores[1:top_n+1]:
        med_info = data.iloc[i[0]]
        recommendations.append({
            "Medicine Name": med_info['Medicine Name'],
            "Composition": med_info['Composition'],
            "Uses": med_info['Uses'],
            "Side Effects": med_info['Side_effects'],
            "Manufacturer": med_info['Manufacturer'],
            "Image URL": med_info['Image URL']
        })
    return recommendations

# Gradio interface function
def gradio_recommend(input_name):
    results = recommend_medicine(input_name)
    if isinstance(results, str):  # Error message
        return results
    
    output = ""
    for med in results:
        output += f"### Tên thuốc: {med['Medicine Name'].capitalize()}\n"
        output += f"**Thành phần:** {med['Composition']}\n"
        output += f"**Công dụng:** {med['Uses']}\n"
        output += f"**Tác dụng phụ:** {med['Side Effects']}\n"
        output += f"**Nhà sản xuất:** {med['Manufacturer']}\n"
        if med['Image URL']:
            output += f"![Hình ảnh thuốc]({med['Image URL']})\n"
        output += "\n"
    return output

# Gradio app
interface = gr.Interface(
    fn=gradio_recommend,
    inputs="text",
    outputs="markdown",
    title="Gợi ý Thuốc Tương Tự",
    description="Nhập tên thuốc để nhận danh sách các thuốc tương tự dựa trên thành phần và công dụng."
)

if __name__ == "__main__":
    interface.launch()