Update app.py

app.py

@@ -24,6 +24,7 @@ except Exception:
 from transformers import AutoTokenizer, AutoModelForCausalLM
 from huggingface_hub import login, HfApi
 
+
 # -------------------
 # Configuration
 # -------------------
@@ -37,8 +38,9 @@ HF_TOKEN = (
 COHERE_API_KEY = os.getenv("COHERE_API_KEY")
 USE_HOSTED_COHERE = bool(COHERE_API_KEY and _HAS_COHERE)
 
+
 # -------------------
-# Helpers (for page header / connection card only)
+# Helpers (used for the connection card only)
 # -------------------
 def local_now_str(user_tz: str | None) -> tuple[str, str]:
     """Returns (label, formatted_time). Falls back to UTC if tz missing/invalid."""
@@ -54,7 +56,9 @@ def local_now_str(user_tz: str | None) -> tuple[str, str]:
             label = "UTC"
     return label, dt.strftime("%Y-%m-%d %H:%M:%S")
 
+
 def header(processing_time=None, user_tz: str | None = None):
+    """Only used in the connection status panel (not in chat replies)."""
     tz_label, now_str = local_now_str(user_tz)
     s = (
         f"Current Date and Time ({tz_label} - YYYY-MM-DD HH:MM:SS formatted): {now_str}\n"
@@ -64,6 +68,7 @@ def header(processing_time=None, user_tz: str | None = None):
         s += f"Processing Time: {processing_time:.2f} seconds\n"
     return s
 
+
 def pick_dtype_and_map():
     if torch.cuda.is_available():
         return torch.float16, "auto"
@@ -71,6 +76,7 @@ def pick_dtype_and_map():
         return torch.float16, {"": "mps"}
     return torch.float32, "cpu"  # CPU path (likely too big for R7B)
 
+
 def is_identity_query(message: str, history) -> bool:
     """Detects identity questions in current message or most recent user turn."""
     patterns = [
@@ -85,17 +91,23 @@ def is_identity_query(message: str, history) -> bool:
         r"\byour\s+name\b",
         r"\bwho\s+am\s+i\s+chatting\s+with\b",
     ]
+
     def hit(text: str | None) -> bool:
         t = (text or "").strip().lower()
         return any(re.search(p, t) for p in patterns)
+
     if hit(message):
         return True
+
     if history:
+        # Gradio history: List[Tuple[user, assistant]]
         last_user = history[-1][0] if isinstance(history[-1], (list, tuple)) and history[-1] else None
         if hit(last_user):
             return True
+
     return False
 
+
 # -------------------
 # Cohere Hosted Path
 # -------------------
@@ -103,6 +115,7 @@ _co_client = None
 if USE_HOSTED_COHERE:
     _co_client = cohere.Client(api_key=COHERE_API_KEY)
 
+
 def _cohere_parse(resp):
     # v5+ responses.create
     if hasattr(resp, "output_text") and resp.output_text:
@@ -116,6 +129,7 @@ def _cohere_parse(resp):
         return resp.text.strip()
     return "Sorry, I couldn't parse the response from Cohere."
 
+
 def cohere_chat(message, history):
     try:
         # Prefer modern API
@@ -143,6 +157,7 @@ def cohere_chat(message, history):
     except Exception as e:
         return f"Error calling Cohere API: {e}"
 
+
 # -------------------
 # Local HF Path
 # -------------------
@@ -153,20 +168,31 @@ def load_local_model():
             "HUGGINGFACE_HUB_TOKEN (or HF_TOKEN) is not set. "
             "Either set it, or provide COHERE_API_KEY to use Cohere's hosted API."
         )
+
     login(token=HF_TOKEN, add_to_git_credential=False)
+
     dtype, device_map = pick_dtype_and_map()
     tok = AutoTokenizer.from_pretrained(
-        MODEL_ID, token=HF_TOKEN, use_fast=True, model_max_length=4096,
-        padding_side="left", trust_remote_code=True,
+        MODEL_ID,
+        token=HF_TOKEN,
+        use_fast=True,
+        model_max_length=4096,
+        padding_side="left",
+        trust_remote_code=True,
     )
     mdl = AutoModelForCausalLM.from_pretrained(
-        MODEL_ID, token=HF_TOKEN, device_map=device_map, low_cpu_mem_usage=True,
-        torch_dtype=dtype, trust_remote_code=True,
+        MODEL_ID,
+        token=HF_TOKEN,
+        device_map=device_map,
+        low_cpu_mem_usage=True,
+        torch_dtype=dtype,
+        trust_remote_code=True,
     )
     if mdl.config.eos_token_id is None and tok.eos_token_id is not None:
         mdl.config.eos_token_id = tok.eos_token_id
     return mdl, tok
 
+
 def build_inputs(tokenizer, message, history):
     msgs = []
     for u, a in (history or []):
@@ -177,6 +203,7 @@ def build_inputs(tokenizer, message, history):
         msgs, tokenize=True, add_generation_prompt=True, return_tensors="pt"
     )
 
+
 def local_generate(model, tokenizer, input_ids, max_new_tokens=350):
     input_ids = input_ids.to(model.device)
     with torch.no_grad():
@@ -194,6 +221,7 @@ def local_generate(model, tokenizer, input_ids, max_new_tokens=350):
     text = tokenizer.decode(gen_only, skip_special_tokens=True)
     return text.strip()
 
+
 # -------------------
 # Chat callback (no header/meta in chat replies)
 # -------------------
@@ -218,6 +246,7 @@ def chat_fn(message, history, user_tz):
     except Exception as e:
         return f"Error during chat: {e}"
 
+
 # -------------------
 # Connection check (keeps header/meta)
 # -------------------
@@ -243,22 +272,23 @@ def check_connection(user_tz=None):
     except Exception as e:
         return f"{header(user_tz=user_tz)}Connection Status: ❌ Error\nDetails: {e}"
 
+
 # -------------------
 # UI
 # -------------------
 with gr.Blocks(theme=gr.themes.Default()) as demo:
-    # Capture browser timezone via JS and store in state
+    # Hold browser timezone (e.g., "America/Toronto")
     user_tz_state = gr.State("")
-    # On load, capture browser timezone via JS and store in user_tz_state
-demo.load(
-    fn=lambda tz: tz,                    # echo the JS value back to Gradio
-    inputs=None,
-    outputs=[user_tz_state],             # outputs must be a LIST
-    js="() => Intl.DateTimeFormat().resolvedOptions().timeZone"
-)
 
+    # On load, capture browser timezone via JS and store in user_tz_state
+    demo.load(
+        fn=lambda tz: tz,                     # echo the JS value
+        inputs=None,
+        outputs=[user_tz_state],              # outputs must be a LIST
+        js="() => Intl.DateTimeFormat().resolvedOptions().timeZone"
+    )
 
-    gr.Markdown(f"# Medical Decision Support AI\n{header(user_tz=None)}")
+    gr.Markdown("# Medical Decision Support AI")
 
     with gr.Row():
         btn = gr.Button("Check Connection Status")
@@ -273,7 +303,7 @@ demo.load(
     chat = gr.ChatInterface(
         fn=chat_fn,
         type="messages",
-        additional_inputs=[user_tz_state],  # pass timezone into chat_fn
+        additional_inputs=[user_tz_state],  # pass timezone into chat_fn (for future use)
         description="A medical decision support system that provides healthcare-related information and decision making support.",
         examples=[
             ["What are the symptoms of hypertension?", ""],
@@ -283,12 +313,10 @@ demo.load(
         cache_examples=False,
     )
 
+    # Wire timezone into the connection check as well
     btn.click(fn=check_connection, inputs=user_tz_state, outputs=status)
 
 if __name__ == "__main__":
     demo.launch()
 
 
-
-
-