Update app.py

app.py

@@ -16,12 +16,14 @@ from moviepy.video.io.ImageSequenceClip import ImageSequenceClip
 from settings import EVENTS
 from player import Player
 from utils import PROMPT, INPUT_PROMPT, extract_after_final_answer
+from llm import *
 
 # Initialize the HF Inference Client
 # It will automatically pick up the HF_TOKEN environment variable if it's set
 
 # Define the models we want to use from the Hugging Face Hub
 LLM_MODEL = "Qwen/Qwen2.5-7B-Instruct"
+MODEL_CONFIG = HUGGINGFACE_LITE
 
 target_size=(1024, 1024)
 
@@ -113,22 +115,26 @@ def play_turn(human_response, loc, player, cur, custom_token, request: gr.Reques
         return
 
     # 1. Process AI Model Response
+    full_msg = PROMPT + INPUT_PROMPT.format(human_response)
     messages = [
         {
             "role": "user", 
-            "content": PROMPT + INPUT_PROMPT.format(human_response)
+            "content": full_msg
         }
     ]
+    
     try:
-        client = InferenceClient(
-            api_key=chosen_token
-        )
-        response = client.chat_completion(
-            model=LLM_MODEL,
-            messages=messages,
-            max_tokens=500,
-            temperature=0.7
-        )
+        model = setup_model(MODEL_CONFIG)
+        response = model.invoke(full_msg)
+        # client = InferenceClient(
+        #     api_key=chosen_token
+        # )
+        # response = client.chat_completion(
+        #     model=LLM_MODEL,
+        #     messages=messages,
+        #     max_tokens=500,
+        #     temperature=0.7
+        # )
     except Exception as e:
         yield (
             f"Error calling API: {str(e)}", 
@@ -139,7 +145,9 @@ def play_turn(human_response, loc, player, cur, custom_token, request: gr.Reques
         )
         return
     
-    processed_response = extract_after_final_answer(response.choices[0].message.content)
+    # processed_response = extract_after_final_answer(response.choices[0].message.content)
+    processed_response = extract_after_final_answer(response.content)
+    
     print(processed_response)
     # 2. Update Game State / Location
     loc = cur._next(processed_response, player)