Implement streaming responses for LLM API

app.py

@@ -48,7 +48,7 @@ llm = HuggingFacePipeline(pipeline=pipeline(
     tokenizer=tokenizer,
     max_new_tokens=512,  # Adjust as needed for desired response length
     return_full_text=False, # Crucial for getting only the AI's response, esp when ans is small
-    temperature=0.5,      # Controls randomness (0.0 for deterministic, 1.0 for very creative)
+    temperature=0.3,      # Controls randomness (0.0 for deterministic, 1.0 for very creative)
     do_sample=True        # Enable sampling for more varied outputs
     ))