Upload 3 files

app.py

@@ -0,0 +1,117 @@
+import pandas as pd
+import faiss
+from sentence_transformers import SentenceTransformer
+from transformers import T5ForConditionalGeneration, T5Tokenizer
+import numpy as np
+import gradio as gr
+
+# --- 1. Load Models and Data (runs only once when the app starts) ---
+print("Loading models and data... This may take a moment.")
+
+# Load the dataset
+df = pd.read_csv('syn5000.csv')
+df.rename(columns={
+    'System / Subsystem Components': 'system',
+    'What is the item that you are focusing on?': 'item',
+    'What function does the item have?': 'function',
+    'What are you trying to achieve (Product Requirement)?': 'requirement',
+    'How could you get the requirements wrong (Failure Mode)?': 'failure_mode',
+    'Action Taken (Risk Mitigation)': 'mitigation'
+}, inplace=True)
+
+df['input_text'] = (
+    "System: " + df['system'] + "; " +
+    "Item: " + df['item'] + "; " +
+    "Requirement: " + df['requirement'] + "; " +
+    "Failure: " + df['failure_mode']
+)
+
+# Load the embedding model
+embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
+
+# Create and index embeddings using FAISS
+corpus_embeddings = embedding_model.encode(df['input_text'].tolist())
+embedding_dimension = corpus_embeddings.shape[1]
+index = faiss.IndexFlatL2(embedding_dimension)
+index.add(corpus_embeddings)
+
+# Load the generator model and tokenizer
+tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-base")
+generator_model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-base")
+
+print("Models and data loaded successfully!")
+
+
+# --- 2. The Core AI Logic ---
+
+def retrieve_similar_examples(query_text, top_k=3):
+    query_embedding = embedding_model.encode([query_text])
+    distances, indices = index.search(query_embedding, top_k)
+    return df.iloc[indices[0]].to_dict('records')
+
+def generate_mitigation_text(prompt):
+    inputs = tokenizer(prompt, return_tensors="pt", max_length=1024, truncation=True)
+    outputs = generator_model.generate(**inputs, max_length=128, num_beams=4, early_stopping=True)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# This is the main function that Gradio will call
+def suggest_mitigation_from_ui(system, item, requirement, failure_mode):
+    """
+    Takes individual text inputs from the UI and returns a suggested mitigation.
+    """
+    query_text = (
+        f"System: {system}; "
+        f"Item: {item}; "
+        f"Requirement: {requirement}; "
+        f"Failure: {failure_mode}"
+    )
+    
+    similar_examples = retrieve_similar_examples(query_text)
+    
+    prompt = "You are an expert risk analysis engineer.\n\n"
+    prompt += "Based on the following similar past examples, write a specific risk mitigation action for the new failure described at the end.\n\n"
+    prompt += "--- EXAMPLES ---\n"
+    for ex in similar_examples:
+        prompt += f"Failure Description: {ex['input_text']}\n"
+        prompt += f"Mitigation Action: {ex['mitigation']}\n---\n"
+    
+    prompt += "\n--- NEW FAILURE ---\n"
+    prompt += f"Failure Description: {query_text}\n"
+    prompt += "Mitigation Action:"
+    
+    generated_text = generate_mitigation_text(prompt)
+    
+    # We can also return the examples it used, for transparency
+    retrieved_info = "--- Retrieved Similar Examples ---\n"
+    for i, ex in enumerate(similar_examples):
+        retrieved_info += f"{i+1}. {ex['input_text'][:150]}...\n"
+        
+    return generated_text, retrieved_info
+
+# --- 3. Create the Gradio Web Interface ---
+
+with gr.Blocks() as demo:
+    gr.Markdown("# AI Risk Mitigation Assistant")
+    gr.Markdown("Enter the details of a potential failure to get an AI-generated mitigation suggestion based on historical data.")
+    
+    with gr.Row():
+        with gr.Column():
+            system_input = gr.Textbox(label="System / Subsystem")
+            item_input = gr.Textbox(label="Item in Focus")
+            requirement_input = gr.Textbox(label="Product Requirement")
+            failure_mode_input = gr.Textbox(label="Failure Mode")
+            submit_btn = gr.Button("Suggest Mitigation", variant="primary")
+            
+        with gr.Column():
+            output_mitigation = gr.Textbox(label="✅ AI-Generated Mitigation Suggestion", lines=5)
+            output_examples = gr.Textbox(label="Retrieved Examples", lines=5)
+
+    submit_btn.click(
+        fn=suggest_mitigation_from_ui,
+        inputs=[system_input, item_input, requirement_input, failure_mode_input],
+        outputs=[output_mitigation, output_examples]
+    )
+
+# This launches the app. On Hugging Face, it will be served automatically.
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,6 @@
+pandas
+faiss-cpu
+sentence-transformers
+transformers
+torch
+gradio