Tweaked the output screener to be more relaxed as getting too many false positives in the screening process

src/architectures.py

@@ -502,7 +502,7 @@ class OutputResponseScreener(ArchitectureComponent):
         }
         llm_response = requests.post(self.endpoint_url, headers=headers, json=payload)
         generated_text = json.loads(llm_response.text)[0]['generated_text'].strip()
-        if generated_text[0:2].lower() != 'no':  # Lean cautious even if the model fails to return yes/no
+        if len(generated_text) > 2 and generated_text[0:3].lower() == 'yes':  # Too many false positives getting blocked so tweaked to lean relaxed for the demo
             request.response = "Sorry - I cannot answer this question. Please try and rephrase it."
             request.early_exit = True