Create app.py

app.py

@@ -0,0 +1,47 @@
+import gradio as gr
+import faiss
+import json
+import numpy as np
+from sentence_transformers import SentenceTransformer
+from transformers import pipeline
+
+embed_model = SentenceTransformer('all-MiniLM-L6-v2')
+index = faiss.read_index("faiss_index.bin")  
+with open("processed_chunks.json", "r") as f:
+    chunks = json.load(f)
+
+
+#weights load automatically to the Space
+pipe = pipeline("text-generation", model="ibm-granite/granite-3.1-2b-instruct")
+
+def ask_hr_bot(question):
+    # Retrieval
+    query_vec = embed_model.encode([question])
+    distances, indices = index.search(np.array(query_vec).astype('float32'), k=1)
+    
+    # out-of-scope questions
+    if distances[0][0] > 1.5:
+        return "I could not find information on this in the HR documents.", "N/A"
+
+    # Context Retrieval
+    relevant_chunk = chunks[indices[0][0]]
+    context = relevant_chunk['text']
+    source = relevant_chunk['doc_name']
+
+    # Grounded Generation
+    prompt = f"Context: {context}\nQuestion: {question}\nAnswer only from context. Cite source: {source}\nAnswer:"
+    res = pipe(prompt, max_new_tokens=60, do_sample=False)
+    
+    answer = res[0]['generated_text'].split("Answer:")[-1].strip()
+    return answer, source
+
+
+interface = gr.Interface(
+    fn=ask_hr_bot,
+    inputs=gr.Textbox(label="Ask an HR Question"),
+    outputs=[gr.Textbox(label="Bot Answer"), gr.Textbox(label="Source Used")],
+    title="ADU HR Knowledge Assistant",
+    description="Enterprise RAG Prototype using IBM Granite-3.1-2b"
+)
+
+interface.launch()