Update app.py

app.py

@@ -52,6 +52,14 @@ def render_profile(profile: Optional[gr.OAuthProfile]) -> str:
     display_name = getattr(profile, "name", None) or getattr(profile, "username", "Hugging Face user")
     return f"Signed in as **{display_name}**."
 
+def refresh_auth_info(
+    profile: Optional[gr.OAuthProfile],
+    oauth_token: Optional[gr.OAuthToken],
+):
+    """Capture OAuth credentials for downstream callbacks."""
+    return oauth_token, render_profile(profile)
+
+
 def respond(
     message,
     image_files,  # Changed parameter name and structure
@@ -75,18 +83,22 @@ def respond(
     print(f"Frequency Penalty: {frequency_penalty}, Seed: {seed}")
     print(f"Selected model (custom_model): {custom_model}")  
     print(f"Model search term: {model_search_term}")
-    print(f"Selected model from radio: {selected_model}")
 
     api_token: Optional[str] = None
+    token_source = "anonymous"
 
     if oauth_token is not None and getattr(oauth_token, "token", None):
         api_token = oauth_token.token
+        token_source = "user"
         print("Using OAuth token from signed-in user for inference.")
     elif HF_READ_TOKEN:
         api_token = HF_READ_TOKEN
+        token_source = "space"
         print("Using server-configured Hugging Face token for inference.")
+    else:
+        print("No default token configured; relying on user sign-in or anonymous access.")
 
-    # Initialize the Inference Client with default HF inference
+    # Initialize the Inference Client with appropriate credentials
     client_kwargs = {}
     if api_token is not None:
         client_kwargs["token"] = api_token
@@ -224,6 +236,24 @@ def respond(
     except HfHubHTTPError as e:
         status = getattr(e.response, "status_code", None)
         if status in (401, 403):
+            if token_source == "user":
+                raise gr.Error(
+                    (
+                        "Failed to generate response: {}\n\n"
+                        "Your Hugging Face session must grant the **inference-api** (Make calls to Inference Providers) "
+                        "permission. Sign out, then sign back in and approve the requested scopes, or update the Space "
+                        "metadata to request that scope."
+                    ).format(e)
+                ) from e
+            if token_source == "space":
+                raise gr.Error(
+                    (
+                        "Failed to generate response: {}\n\n"
+                        "The Space-level token lacks Inference Provider access. Update the `HF_READ_TOKEN` secret with "
+                        "a token that has the `Make calls to Inference Providers` permission, or have users sign in with "
+                        "their own accounts."
+                    ).format(e)
+                ) from e
             raise gr.Error(
                 "Failed to generate response: {}. Sign in with your Hugging Face account and retry."
                 .format(e)
@@ -372,6 +402,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
 
     # Chat history state
     chat_history = gr.State([])
+    oauth_token_state = gr.State(None)
     
     # Function to filter models
     def filter_models(search_term):
@@ -441,7 +472,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
         custom_model,
         search_term,
         selected_model,
-        oauth_token: Optional[gr.OAuthToken] = None,
+        oauth_token_obj: Optional[gr.OAuthToken] = None,
     ):
         # Check if history is valid
         if not history or len(history) == 0:
@@ -492,7 +523,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
                 custom_model,
                 search_term,
                 selected_model,
-                oauth_token
+                oauth_token=oauth_token_obj,
             ):
                 history[-1][1] = response
                 yield history
@@ -511,7 +542,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
                 custom_model,
                 search_term,
                 selected_model,
-                oauth_token
+                oauth_token=oauth_token_obj,
             ):
                 history[-1][1] = response
                 yield history
@@ -526,7 +557,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
         bot,
         [chatbot, system_message_box, max_tokens_slider, temperature_slider, top_p_slider, 
          frequency_penalty_slider, seed_slider, custom_model_box, 
-         model_search_box, featured_model_radio],
+         model_search_box, featured_model_radio, oauth_token_state],
         [chatbot]
     ).then(
         lambda: {"text": "", "files": []},  # Clear inputs after submission
@@ -550,7 +581,13 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
     )
     print("Featured model radio button change event linked.")
 
-    demo.load(render_profile, inputs=None, outputs=auth_status)
+    login_button.click(
+        refresh_auth_info,
+        inputs=None,
+        outputs=[oauth_token_state, auth_status],
+    )
+
+    demo.load(refresh_auth_info, inputs=None, outputs=[oauth_token_state, auth_status])
 
 print("Gradio interface initialized.")