Create app.py

app.py

@@ -0,0 +1,38 @@
+import gradio as gr
+import chromadb
+import numpy as np
+from sentence_transformers import SentenceTransformer
+from transformers import pipeline
+import pickle
+
+# Load pre-trained model and embeddings
+model = SentenceTransformer("all-MiniLM-L6-v2")  # You can upload this model from HF Hub if available
+generator = pipeline("text-generation", model="gpt2")
+
+# Initialize ChromaDB client (using the Chroma database uploaded as a file)
+client = chromadb.Client()
+collection = client.create_collection("documents")
+
+# Manually load your embeddings and document data from the HF Space files
+with open("embeddings.pkl", "rb") as f:
+    embeddings = pickle.load(f)
+
+# Example of adding embeddings to FAISS (if using FAISS as the indexer)
+faiss_index = faiss.IndexFlatL2(512)  # Adjust dimension if needed
+faiss_index.add(np.array(embeddings))
+
+# Example documents loaded manually or fetched via API
+documents = ["What is RAG?", "How does FAISS work?", "Introduction to Chroma."]
+
+def generate_answer(query):
+    query_embedding = model.encode([query])
+    D, I = faiss_index.search(np.array(query_embedding), k=1)  # Retrieve the closest document
+    retrieved_doc = documents[I[0][0]]
+
+    prompt = f"Context: {retrieved_doc}\nQuestion: {query}\nAnswer:"
+    response = generator(prompt, max_length=50)
+    return response[0]['generated_text']
+
+# Gradio interface for manual file uploads and query input
+iface = gr.Interface(fn=generate_answer, inputs="text", outputs="text")
+iface.launch()