Update app.py

app.py

@@ -1,66 +1,77 @@
+import os
 import faiss
 import numpy as np
 import torch
-from sentence_transformers import SentenceTransformer
-from transformers import AutoModelForCausalLM, AutoTokenizer
 import gradio as gr
-import os
+from transformers import AutoModelForCausalLM, GemmaTokenizer
 from huggingface_hub import hf_hub_download
 
 # 🔹 Hugging Face Credentials
-HF_REPO = "Futuresony/future_ai_12_10_2024.gguf"
-HF_TOKEN = os.getenv('HUGGINGFACEHUB_API_TOKEN')
+HF_REPO = "Futuresony/my_model"  # Ensure this is correct
+HF_TOKEN = os.getenv('HUGGINGFACEHUB_API_TOKEN')  # Ensure this is set in your environment
 
-# 🔹 Paths
+# 🔹 FAISS Index Path
 FAISS_PATH = "asa_faiss.index"
-DATASET_PATH = "responses.txt"
+DATASET_PATH = "responses.txt"  # Ensure this file contains indexed responses
 
-# 🔹 Load FAISS Index
-faiss_local_path = hf_hub_download(HF_REPO, FAISS_PATH, token=HF_TOKEN)
-faiss_index = faiss.read_index(faiss_local_path)
-print("✅ FAISS index loaded successfully!")
+# ✅ Load FAISS Index from Hugging Face if not available locally
+if not os.path.exists(FAISS_PATH):
+    print("🔄 Downloading FAISS index...")
+    FAISS_PATH = hf_hub_download(HF_REPO, "asa_faiss.index", token=HF_TOKEN)
 
-# 🔹 Load Dataset Responses
-with open(DATASET_PATH, "r", encoding="utf-8") as f:
-    dataset = f.readlines()
-print("✅ Responses dataset loaded!")
+print(f"📂 Loading FAISS index from {FAISS_PATH}...")
+faiss_index = faiss.read_index(FAISS_PATH)
+print("✅ FAISS index loaded successfully!")
 
-# 🔹 Load Model & Tokenizer Correctly
-tokenizer = AutoTokenizer.from_pretrained(HF_REPO, token=HF_TOKEN)
+# ✅ Load responses dataset
+if os.path.exists(DATASET_PATH):
+    with open(DATASET_PATH, "r", encoding="utf-8") as f:
+        dataset = f.readlines()
+    print("✅ Responses dataset loaded!")
+else:
+    print(f"⚠️ Warning: {DATASET_PATH} not found!")
+    dataset = []
+
+# ✅ Load model & tokenizer
+print("🔄 Loading tokenizer and model...")
+tokenizer = GemmaTokenizer.from_pretrained(HF_REPO, token=HF_TOKEN)
 model = AutoModelForCausalLM.from_pretrained(HF_REPO, token=HF_TOKEN)
+print("✅ Model and tokenizer loaded!")
 
-# 🔹 Load Sentence Transformer (Ensures proper FAISS embedding match)
-embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
-
-# 🔹 Adjust FAISS Threshold
-THRESHOLD = 100  # Adjust based on actual FAISS distances
-
+# 🔹 Set FAISS distance threshold (lower values = more strict matches)
+THRESHOLD = 80  # Adjusted threshold for better accuracy
 
 def embed(text):
-    """Convert text into FAISS-compatible vector using the same method as training."""
-    return embedder.encode([text], convert_to_tensor=True).cpu().numpy()
-
+    """Convert text to FAISS-compatible vector."""
+    tokens = tokenizer.encode(text, add_special_tokens=True)
+    return np.array(tokens, dtype=np.float32).reshape(1, -1)
 
 def chatbot_response(user_query):
-    """Fetch response from FAISS or generate with model if needed."""
-    query_vector = embed(user_query)  # Convert input to vector
-    D, I = faiss_index.search(query_vector, k=1)  # Search FAISS index
+    """Fetches response from FAISS or falls back to the model."""
+    query_vector = embed(user_query)
+    D, I = faiss_index.search(query_vector, k=1)
 
-    print(f"Closest FAISS match index: {I[0][0]}, Distance: {D[0][0]}")  # Debugging info
+    print(f"🔍 Closest FAISS match index: {I[0][0]}, Distance: {D[0][0]}")
 
-    if D[0][0] < THRESHOLD:  # If FAISS match is good
-        response = dataset[I[0][0]].strip()  # Retrieve FAISS response
+    if D[0][0] < THRESHOLD and 0 <= I[0][0] < len(dataset):
+        response = dataset[I[0][0]].strip()
         print("✅ FAISS response used!")
     else:
-        # 🔥 Fallback: Generate response with model
-        print("⚠️ FAISS match too weak, using model instead.")
+        print("⚠️ FAISS match too weak, generating response using model.")
         inputs = tokenizer(user_query, return_tensors="pt")
         outputs = model.generate(**inputs, max_new_tokens=150)
         response = tokenizer.decode(outputs[0], skip_special_tokens=True)
 
     return response
 
-
 # 🔹 Gradio UI
-iface = gr.Interface(fn=chatbot_response, inputs="text", outputs="text", title="ASA Microfinance Chatbot")
-iface.launch()
+iface = gr.Interface(
+    fn=chatbot_response,
+    inputs="text",
+    outputs="text",
+    title="ASA Microfinance Chatbot",
+    description="A chatbot that provides information using FAISS and a language model."
+)
+
+if __name__ == "__main__":
+    iface.launch()