Update app.py

app.py

@@ -48,7 +48,7 @@ logger = logging.getLogger("fireai")
 logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("httpcore").setLevel(logging.WARNING)
 
-# Space owner token strictly used for high-speed container asset downloads
+# Enforce Space environment token & hf_transfer for accelerated model downloads
 HF_TOKEN = os.environ.get("HF_TOKEN") or os.environ.get("HUGGING_FACE_HUB_TOKEN")
 os.environ.setdefault("HF_HUB_ENABLE_HF_TRANSFER", "1")
 
@@ -87,6 +87,7 @@ _plan_lock = threading.Lock()
 
 
 def hub_kwargs() -> dict[str, Any]:
+    # Always passes HF_TOKEN to authenticate and enable multi-threaded high-speed transfers
     return {"token": HF_TOKEN} if HF_TOKEN else {}
 
 
@@ -181,12 +182,12 @@ def load_model(mode: str) -> dict[str, Any]:
             clear_loaded_model()
 
         repo = MODELS[mode]
-        logger.info("Loading tokenizer: %s", repo)
+        logger.info("Loading tokenizer with HF_TOKEN: %s", repo)
         tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True, **hub_kwargs())
         if tokenizer.pad_token is None:
             tokenizer.pad_token = tokenizer.eos_token
 
-        logger.info("Loading %s as NF4 on large slice: %s", mode, repo)
+        logger.info("Loading %s as NF4 on large slice with HF_TOKEN: %s", mode, repo)
         started = time.monotonic()
 
         model = AutoModelForCausalLM.from_pretrained(
@@ -233,7 +234,7 @@ def history_to_prompt(history: list, mode: str, tokenizer) -> str:
     for item in history or []:
         if not isinstance(item, dict):
             continue
-        if (item.get("metadata") or {}).get("title") == "Raw thinking":
+        if (item.get("metadata") or {}).get("title") in ("Thinking", "Raw thinking"):
             continue
         role = item.get("role")
         content = item.get("content")
@@ -323,7 +324,6 @@ def generate_response(
     enabled = gr.update(interactive=True)
     info = get_account(profile, token)
 
-    # Strictly block unauthenticated calls from using ZeroGPU
     if not info["logged_in"]:
         history.append(
             {
@@ -347,7 +347,7 @@ def generate_response(
         history.append({
             "role": "assistant",
             "content": "Thinking...",
-            "metadata": {"title": "Raw thinking"},
+            "metadata": {"title": "Thinking"},
         })
         thought_index = len(history) - 1
         history.append({