Switch to OpenAI-compatible chat API with Qwen2.5-7B

Old HF Inference API format (inputs/parameters) is dead. New router
uses OpenAI-compatible /v1/chat/completions with messages array.
Mistral-7B no longer available on free tier; switch to Qwen2.5-7B
which is recommended and warm. v0.9.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app.py

@@ -10,7 +10,7 @@ app = Flask(__name__, template_folder=os.path.join(BASE_DIR, 'templates'))
 
 # --- CONFIGURATION ---
 HF_TOKEN = os.environ.get("HF_TOKEN", "")
-LLM_MODEL = "mistralai/Mistral-7B-Instruct-v0.2"
+LLM_MODEL = "Qwen/Qwen2.5-7B-Instruct-1M"
 
 # British female voice — sharp RP diction
 TTS_VOICE = "en-GB-SoniaNeural"
@@ -53,27 +53,26 @@ def chat():
     if not user_msg:
         return jsonify({"response": "Darling, silence is only for those who have nothing to say."})
 
-    prompt = f"""<s>[INST] {SYSTEM_PROMPT}
-
- User: {user_msg} [/INST]"""
-
     try:
-        headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+        headers = {
+            "Authorization": f"Bearer {HF_TOKEN}",
+            "Content-Type": "application/json"
+        }
         payload = {
-            "inputs": prompt,
-            "parameters": {
-                "max_new_tokens": 100,
-                "temperature": 0.9,
-                "top_p": 0.95,
-                "return_full_text": False
-            }
+            "model": LLM_MODEL,
+            "messages": [
+                {"role": "system", "content": SYSTEM_PROMPT},
+                {"role": "user", "content": user_msg}
+            ],
+            "max_tokens": 150,
+            "temperature": 0.9,
+            "top_p": 0.95
         }
 
-        response = requests.post(f"https://router.huggingface.co/hf-inference/models/{LLM_MODEL}", headers=headers, json=payload)
+        response = requests.post("https://router.huggingface.co/v1/chat/completions", headers=headers, json=payload)
 
         if response.status_code == 200:
-            ai_msg = response.json()[0]['generated_text'].strip()
-            ai_msg = ai_msg.replace("Assistant:", "").strip()
+            ai_msg = response.json()["choices"][0]["message"]["content"].strip()
             return jsonify({"response": ai_msg})
         else:
             err = response.text[:200]

templates/index.html

@@ -97,7 +97,7 @@
                 <input type="text" id="userInput" placeholder="Type here..." autocomplete="off">
                 <button id="sendBtn">SEND</button>
             </div>
-            <div id="status" class="status">Ready. <span style="float:right;color:#444;">v0.7</span></div>
+            <div id="status" class="status">Ready. <span style="float:right;color:#444;">v0.9</span></div>
             <audio id="player"></audio>
         </div>
     </div>