Update app.py

app.py

@@ -4,18 +4,18 @@ import os
 from huggingface_hub import InferenceClient
 
 app = Flask(__name__)
-CORS(app)  # Keeps your Acode connection unlocked permanently!
+CORS(app)  # Keeps your Acode terminal connected safely!
 
-# Pulls your hidden secret token safely from Space Settings
+# Pulls your hidden WRITE token from Space secrets
 HF_TOKEN = os.getenv("HF_TOKEN")
 MODEL_REPO = "Kelvin000010191/Krypton-1"
 
-# Initializes the client with the authentication token
+# Initializes the official native client wrapper
 client = InferenceClient(token=HF_TOKEN)
 
 @app.route("/", methods=["GET"])
 def home():
-    return "Krypton-1 Core Node Operational."
+    return "Krypton-1 Core Chat Node Operational."
 
 @app.route("/api/chat", methods=["POST"])
 def chat():
@@ -23,29 +23,30 @@ def chat():
         data = request.json
         user_prompt = data.get("prompt", "")
         
-        # Explicitly targets the model repo and handles text generation
-        response = client.text_generation(
-            prompt=user_prompt,
+        # Uses standard Chat Completion matrix format instead of raw text blocks
+        response = client.chat_completion(
             model=MODEL_REPO,
-            max_new_tokens=250,
+            messages=[{"role": "user", "content": user_prompt}],
+            max_tokens=250,
             temperature=0.7
         )
         
-        return jsonify({"reply": response})
+        # Pulls out the clean text string answer generated by the model
+        reply_text = response.choices[0].message.content
+        return jsonify({"reply": reply_text})
         
     except Exception as e:
         raw_error = str(e)
         
-        # Flags if the newly generated token isn't saved right
+        # Catch if the token setup isn't fully saved yet
         if "401" in raw_error or "unauthorized" in raw_error.lower():
-            return jsonify({"reply": "SYS_AUTH_ERR: Token unauthorized. Make sure your fresh WRITE token is saved as HF_TOKEN in Space Secrets."})
+            return jsonify({"reply": "SYS_AUTH_ERR: Token unauthorized. Double check your HF_TOKEN inside Space Secrets."})
             
-        # Flags if the massive 15GB clusters are waking up
+        # Catch if your 15GB model layer is waking up its hardware node
         if "loading" in raw_error.lower() or "503" in raw_error:
-            return jsonify({"reply": "SYS_BOOT: Krypton-1 engine is loading weights into active memory. Give the cloud 60 seconds, then try again!"})
+            return jsonify({"reply": "SYS_BOOT: Krypton-1 engine is loading weights into active memory clusters. Give the cloud 60 seconds, then try again!"})
             
-        # Prints any other raw issue directly to your phone screen
-        return jsonify({"reply": f"SYS_ALERT: Connected to server, but model errored. Details: {raw_error}"})
+        return jsonify({"reply": f"SYS_ALERT: Handshake complete, but model rejected structure. Error Details: {raw_error}"})
 
 if __name__ == "__main__":
     app.run(host="0.0.0.0", port=7860)