works with gemini

README.md

@@ -7,8 +7,5 @@ sdk: gradio
 sdk_version: 5.42.0
 app_file: app.py
 pinned: false
-hf_oauth: true
-hf_oauth_scopes:
-- inference-api
 ---
-An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).
+AskNews-powered news assistant built with [Gradio](https://gradio.app), [`google-genai`](https://pypi.org/project/google-genai/), and the [AskNews SDK](https://pypi.org/project/asknews/).

app.py

@@ -9,10 +9,11 @@ from dotenv import load_dotenv
 load_dotenv()
 
 import gradio as gr
-from huggingface_hub import InferenceClient, InferenceTimeoutError
+from google import genai
+from google.genai import types
 from asknews_sdk import AskNewsSDK
 
-DEFAULT_MODEL = "openai/gpt-oss-20b"
+DEFAULT_MODEL = "gemini-2.5-pro"
 
 
 LOG_LEVEL = os.getenv("ASKNEWS_LOG_LEVEL", "INFO").upper()
@@ -35,7 +36,7 @@ def get_asknews_sdk() -> Optional[AskNewsSDK]:
             client_id=client_id,
             client_secret=client_secret,
             scopes=["news"]
-          )
+        )
         logger.info("AskNews SDK initialised successfully.")
         return sdk
     except Exception as exc:
@@ -102,7 +103,6 @@ def respond(
     max_tokens: int,
     temperature: float,
     top_p: float,
-    hf_token: gr.OAuthToken,
     model_name: str = DEFAULT_MODEL,
     use_asknews: bool = True,
     asknews_hours_back: int = 24*30,
@@ -110,13 +110,26 @@ def respond(
     asknews_domains_csv: str = "afp.com",
 ):
     """
-    Stream chat responses from HF, enriching with AskNews context when enabled.
+    Stream chat responses from Google Gemini, enriching with AskNews context when enabled.
     """
-    # Validate OAuth token for HF
-    if hf_token is None or hf_token.token is None or hf_token.token.strip() == "":
-        yield "Veuillez vous connecter à Hugging Face via le bouton Login dans la barre latérale."
+    api_key = os.getenv("GOOGLE_API_KEY", "").strip()
+    if not api_key:
+        logger.warning("Missing Google API key.")
+        yield (
+            "Définissez GOOGLE_API_KEY dans votre environnement ou saisissez la clé API Google Gemini dans le champ dédié."
+        )
+        return
+
+    try:
+        genai_client = genai.Client(api_key=api_key)
+    except Exception as exc:
+        logger.exception("Failed to initialise Google GenAI client: %s", exc)
+        yield f"Échec d'initialisation du client Google GenAI: {exc}"
         return
 
+    user_message_raw = "" if message is None else str(message)
+    user_message = user_message_raw.strip()
+
     # Prepare AskNews SDK if requested
     sdk = get_asknews_sdk() if use_asknews else None
     asknews_context = ""
@@ -130,7 +143,7 @@ def respond(
         )
         asknews_context = fetch_asknews_context(
             sdk=sdk,
-            query=message,
+            query=user_message,
             hours_back=asknews_hours_back,
             n_articles=asknews_n_articles,
             domains=domains,
@@ -140,40 +153,11 @@ def respond(
         else:
             logger.warning("AskNews context is empty after fetch.")
 
-    # Build messages
-    messages: List[Dict[str, str]] = []
     base_system = system_message.strip() if system_message else "You are a helpful assistant."
-    messages.append({"role": "system", "content": base_system})
 
-    # If we have AskNews context, inject it as an additional system guidance
-    if asknews_context:
-        messages.append({
-            "role": "system",
-            "content": (
-                "Use the following news context when answering. If the user's query is unrelated, ignore it.\n\n"
-                f"{asknews_context}"
-            ),
-        })
-
-    # Include history (roles should be valid)
-    for m in history or []:
-        role = m.get("role")
-        content = m.get("content", "")
-        if role in ("user", "assistant", "system") and content is not None:
-            messages.append({"role": role, "content": str(content)})
-
-    # Current user message
-    if message is None or str(message).strip() == "":
+    if not user_message:
         yield "Veuillez saisir un message."
         return
-    messages.append({"role": "user", "content": str(message).strip()})
-
-    # Initialize HF client
-    try:
-        client = InferenceClient(token=hf_token.token, model=model_name)
-    except Exception as e:
-        yield f"Échec d'initialisation du client d'inférence HF: {e}"
-        return
 
     response_accum = ""
     # Optional prefix informing about context usage (not counted by model, only displayed)
@@ -191,33 +175,69 @@ def respond(
     #     response_accum += "[Contexte AskNews]\n" + context_display + "\n\n"
     #     yield response_accum
 
+    system_instruction = base_system.strip()
+    if asknews_context:
+        system_instruction += (
+            "\n\nUtilise le contexte AskNews suivant pour ta réponse. Si la question est sans rapport, ignore ce contexte.\n"
+            f"{asknews_context}"
+        )
+
+    conversation: List[types.Content] = []
+    for msg in history or []:
+        role = msg.get("role")
+        content = str(msg.get("content", "")).strip()
+        if not content or role not in ("user", "assistant"):
+            continue
+        if role == "user":
+            conversation.append(
+                types.Content(role="user", parts=[types.Part.from_text(text=content)])
+            )
+        else:
+            conversation.append(
+                types.Content(role="model", parts=[types.Part.from_text(text=content)])
+            )
+
+    conversation.append(
+        types.Content(role="user", parts=[types.Part.from_text(text=user_message)])
+    )
+
+    generation_config = types.GenerateContentConfig(
+        systemInstruction=system_instruction,
+        temperature=float(temperature),
+        topP=float(top_p),
+        maxOutputTokens=int(max_tokens),
+    )
+
     try:
-        for chunk in client.chat_completion(
-            messages=messages,
-            max_tokens=max_tokens,
-            stream=True,
-            temperature=temperature,
-            top_p=top_p,
-        ):
+        stream = genai_client.models.generate_content_stream(
+            model=model_name,
+            contents=conversation,
+            config=generation_config,
+        )
+        for chunk in stream:
             try:
-                choices = getattr(chunk, "choices", [])
-                token = ""
-                if choices and getattr(choices[0], "delta", None) is not None:
-                    token_piece = getattr(choices[0].delta, "content", None)
-                    if token_piece:
-                        token = token_piece
+                token = getattr(chunk, "text", None)
+                if not token and getattr(chunk, "candidates", None):
+                    parts: List[str] = []
+                    for candidate in chunk.candidates:
+                        content = getattr(candidate, "content", None)
+                        if content and getattr(content, "parts", None):
+                            for part in content.parts:
+                                piece = getattr(part, "text", None)
+                                if piece:
+                                    parts.append(piece)
+                    token = "".join(parts)
                 if token:
                     response_accum += token
                     yield response_accum
             except Exception:
                 continue
-    except InferenceTimeoutError:
-        yield response_accum + "\n\n[Temps dépassé. Réessayez ou réduisez 'Max new tokens'.]"
     except Exception as e:
+        logger.exception("Google GenAI generation failed: %s", e)
         if response_accum:
             yield response_accum + f"\n\n[Erreur: {e}]"
         else:
-            yield f"Erreur de l'API d'inférence: {e}"
+            yield f"Erreur de génération Gemini: {e}"
 
 
 # ---- Gradio UI ----
@@ -263,14 +283,13 @@ Style :
 )
 
 with gr.Blocks() as demo:
-    gr.Markdown("# Chatbot HF avec contexte AskNews")
+    gr.Markdown("# Chatbot Gemini avec contexte AskNews")
     with gr.Sidebar():
-        gr.LoginButton()
         gr.Markdown(
-            "Connectez-vous avec votre compte Hugging Face.\n\n"
-            "Pour activer AskNews, définissez les variables d'environnement:\n"
+            "Définissez dans votre environnement les variables d'environnement suivantes :\n"
             "- ASKNEWS_CLIENT_ID\n"
             "- ASKNEWS_CLIENT_SECRET\n\n"
+            "Configurer la clé Google Gemini via GOOGLE_API_KEY.\n\n"
             "Ajustez les paramètres pour contrôler le contexte (heures, domaines, nombre d'articles)."
         )
     chatbot.render()