faster

app.py

@@ -11,7 +11,7 @@ from langchain_classic.chains import RetrievalQA
 
 # 1. SETUP: DEFINING THE MODELS
 # We use a smaller/faster multilingual embedding model for retrieval
-embedding_model_name = "intfloat/multilingual-e5-small"
+embedding_model_name = "intfloat/multilingual-e5-large"
 
 # Detect device - use CUDA if available, otherwise CPU
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -44,7 +44,7 @@ def process_and_chat(file, query):
             return "**Error:** COHERE_API_KEY environment variable is not set."
         
         # Initialize LLM (Cohere Command R - faster model)
-        llm = ChatCohere(model="command-r7b-12-2024", temperature=0.3, cohere_api_key=api_key)
+        llm = ChatCohere(model="command-a-03-2025", temperature=0.3, cohere_api_key=api_key)
         
         # 2. LOAD & PROCESS DOCUMENT
         # Gradio 6.x returns file path as string directly