Revert to old model, login with secret token

app.py

@@ -6,6 +6,15 @@ import torch
 import numpy as np
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 
+import os
+
+# Login automatically if HF_TOKEN is present
+hf_token = os.getenv("HF_TOKEN")
+if hf_token:
+    from huggingface_hub import login
+    login(token=hf_token)
+
+
 # Load corpus
 print("Loading dataset...")
 dataset = load_dataset("rag-datasets/rag-mini-wikipedia", "text-corpus")
@@ -25,8 +34,8 @@ index.add(corpus_embeddings_np)
 reranker = CrossEncoder("cross-encoder/ms-marco-MiniLM-L-6-v2")
 
 # Generator (choose one: local HF model or OpenAI)
-tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-4")
-model = AutoModelForCausalLM.from_pretrained("microsoft/phi-4", device_map="auto", torch_dtype=torch.float16)
+tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
+model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3", device_map="auto", torch_dtype=torch.float16)
 generator = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=150)
 
 def rag_pipeline(query):