Update app.py

app.py

@@ -3,133 +3,125 @@ from supabase import create_client
 import requests, tempfile
 import gradio as gr
 from llama_cpp import Llama
+from typing import List, Dict
 
-#Model Download & Load
+# ---------------- Model Setup ----------------
 MODEL_URL = "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf"
-#MODEL_URL = "https://huggingface.co/TheBloke/bling-stable-lm-3b-4e1t-v0-GGUF/resolve/main/bling-stable-lm-3b-4e1t-v0.Q4_K_M.gguf"
 
 llm = None
 def get_model():
     global llm
     if llm is None:
-        print("Downloading model")
+        print("Downloading model...")
         response = requests.get(MODEL_URL, stream=True, timeout=600)
         response.raise_for_status()
-
         with tempfile.NamedTemporaryFile(delete=False, suffix=".gguf") as tmp_file:
             for chunk in response.iter_content(chunk_size=8192):
                 tmp_file.write(chunk)
             model_path = tmp_file.name
-
-        print("Loading model into memory")
+        print("Loading model into memory...")
         llm = Llama(model_path=model_path, n_ctx=512, n_threads=2)
     return llm
 
-#Supabase setup to store data in db
+# ---------------- Supabase Setup ----------------
 SUPABASE_URL = os.getenv("SUPABASE_URL")
 SUPABASE_KEY = os.getenv("SUPABASE_KEY")
-
 supabase = create_client(SUPABASE_URL, SUPABASE_KEY)
 
 def save_to_supabase(user_id, prompt, response, model, tokens=None):
     try:
         supabase.table("ai_logs").insert({
-            "user_id": user_id,      
-            "prompt": prompt,       
-            "response": response,           
-            "model": model,         
-            "tokens": tokens          
+            "user_id": user_id,
+            "prompt": prompt,
+            "response": response,
+            "model": model,
+            "tokens": tokens
         }).execute()
     except Exception as e:
         print("Supabase insert failed:", e)
 
-#Chat Function
-MAX_HISTORY = 4  
+# ---------------- Chat Function ----------------
+MAX_HISTORY = 4
 
-def chat_with_infinite_agent(message, lang, history):
+def chat_with_infinite_agent(message: str, lang: str = "English", history: List[Dict[str, str]] = None):
     history = history or []
-
-    #Keep only the last MAX_HISTORY items
     history = history[-MAX_HISTORY:]
 
-    #Build context from previous messages (trimmed for stability)
+    # Build context
     context_lines = []
     for m, a in zip(history[::2], history[1::2]):
         if m['role']=='user' and a['role']=='assistant':
-            #Trim very long messages to reduce memory
-            user_msg = m['content'][:300]
-            assistant_msg = a['content'][:300]
-            context_lines.append(f"User: {user_msg}\nAgent: {assistant_msg}")
+            context_lines.append(f"User: {m['content'][:300]}\nAgent: {a['content'][:300]}")
     context = "\n".join(context_lines)
 
-    #Language adjustment
+    # Language adjustment
     if lang == "Svenska":
         message = f"Svara på svenska: {message}"
     elif lang == "Türkçe":
         message = f"Türkçe cevapla: {message}"
 
-    #Full prompt with your Human Design & Gene Keys
     prompt = f"""
-    You are Infinite Agent — the AI embodiment of Tugce Ozdeger,
-    a 5/1 Emotional Manifestor whose wisdom flows through Gene Keys:
-    12, 22, 11, 37, 21, 61, 31, 39, 46, and 25.
-
-    Guide others toward emotional clarity, self-worth, and life direction
-    with empathy, depth, and grace.
+You are Infinite Agent — the AI embodiment of Tugce Ozdeger,
+a 5/1 Emotional Manifestor whose wisdom flows through Gene Keys:
+12, 22, 11, 37, 21, 61, 31, 39, 46, and 25.
 
-    {context}
-    User: {message}
-    Agent:
-    """
+Guide others toward emotional clarity, self-worth, and life direction
+with empathy, depth, and grace.
 
-    #Generate response safely
+{context}
+User: {message}
+Agent:
+"""
     try:
-        llm = get_model()
-        response = llm(prompt, max_tokens=512)
+        llm_model = get_model()
+        response = llm_model(prompt, max_tokens=512)
         output = response["choices"][0]["text"].strip()
+        tokens_used = response.get("usage", {}).get("total_tokens", None)
     except Exception as e:
         print("LLM error:", e)
         output = "⚠️ Sorry, the model failed to respond. Please try again."
+        tokens_used = None
 
-    #Append new messages to history
+    # Update history
     history.append({"role": "user", "content": message})
     history.append({"role": "assistant", "content": output})
 
-    save_to_supabase(
-    user_id="anonymous",       
-    prompt=message,           
-    response=output,                            
-    model=MODEL_URL,           
-    tokens=len(message)
-)
-    return history, history
+    save_to_supabase("anonymous", message, output, MODEL_URL, tokens_used)
+    return output, history
 
-#Gradio UI
+# ---------------- API Function ----------------
+def api_predict(message: str, lang: str = "English", history: List[Dict[str, str]] = None) -> dict:
+    output, updated_history = chat_with_infinite_agent(message, lang, history or [])
+    return {"response": output, "history": updated_history}
+
+# ---------------- Gradio UI ----------------
 css_path = os.path.join(os.path.dirname(__file__), "style.css")
 
 with gr.Blocks(title="Infinite Agent", css=open(css_path).read()) as demo:
-    #Avatar image
+    # Avatar & header
     gr.Image("avatar.png", elem_id="agent-avatar")
-
-    #Header
     gr.Markdown(
         "### 🌙 Infinite Agent — Emotional Clarity & Life Direction\n"
         "_Guided by Tugce Ozdeger’s Human Design & Gene Keys_",
         elem_classes="header-text"
     )
 
-    #Language selector
+    # Language selector
     language = gr.Dropdown(["English", "Svenska", "Türkçe"], label="Choose language", value="English")
 
-    #Chatbot
+    # Chatbot
     chatbot = gr.Chatbot(type="messages", value=[
         {"role": "system", "content": "🌙 Welcome to Infinite Agent — your guide to emotional clarity, self-worth, and life direction."}
     ])
-
     msg = gr.Textbox(placeholder="Ask about your emotions, direction, or purpose...")
     clear = gr.Button("Clear")
 
+    # Actions
     msg.submit(chat_with_infinite_agent, [msg, language, chatbot], [chatbot, chatbot])
     clear.click(lambda: [], None, chatbot, queue=False)
 
+    # Expose API endpoint
+    gr.api(api_predict, api_name="/api/predict")
+
+# Launch the app
 demo.launch()