Update app.py

app.py

@@ -3,125 +3,132 @@ from supabase import create_client
 import requests, tempfile
 import gradio as gr
 from llama_cpp import Llama
-from typing import List, Dict
 
-# ---------------- Model Setup ----------------
+#MODEL DOWNLOAD & LOAD
 MODEL_URL = "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf"
-
-llm = None
-def get_model():
-    global llm
-    if llm is None:
-        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...")
-        llm = Llama(model_path=model_path, n_ctx=512, n_threads=2)
-    return llm
-
-# ---------------- Supabase Setup ----------------
+#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"
+
+print("Downloading model into temporary memory... This may take a while ⏳")
+response = requests.get(MODEL_URL, stream=True)
+response.raise_for_status()
+
+#Save to a temporary file (deleted when app ends)
+with tempfile.NamedTemporaryFile(delete=False, suffix=".gguf") as tmp_file:
+    for chunk in response.iter_content(chunk_size=8192):
+        tmp_file.write(chunk)
+    temp_model_path = tmp_file.name
+
+#Load model
+llm = Llama(
+    model_path=temp_model_path,
+    n_ctx=512,   # smaller context
+    n_threads=2  # adjust based on your CPU
+)
+
+#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 WITH ROLLING CONTEXT
+MAX_HISTORY = 4  # Keep last 2 user+assistant exchanges for stability
 
-def chat_with_infinite_agent(message: str, lang: str = "English", history: List[Dict[str, str]] = None):
+def chat_with_infinite_agent(message, lang, history):
     history = history or []
+
+    #Keep only the last MAX_HISTORY items
     history = history[-MAX_HISTORY:]
 
-    # Build context
+    #Build context from previous messages (trimmed for stability)
     context_lines = []
     for m, a in zip(history[::2], history[1::2]):
         if m['role']=='user' and a['role']=='assistant':
-            context_lines.append(f"User: {m['content'][:300]}\nAgent: {a['content'][:300]}")
+            #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 = "\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.
-
-{context}
-User: {message}
-Agent:
-"""
+    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.
+    {context}
+    User: {message}
+    Agent:
+    """
+
+    #Generate response safely
     try:
-        llm_model = get_model()
-        response = llm_model(prompt, max_tokens=512)
+        response = llm(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
 
-    # Update history
+    #Append new messages to history
     history.append({"role": "user", "content": message})
     history.append({"role": "assistant", "content": output})
 
-    save_to_supabase("anonymous", message, output, MODEL_URL, tokens_used)
-    return output, history
+    save_to_supabase(
+    user_id="anonymous",       
+    prompt=message,           
+    response=output,                            
+    model=MODEL_URL,           
+    tokens=len(message)
+)
+    return history, history
 
-# ---------------- 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 ----------------
+#GRADIO UI WITH DARK MODE
 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 & header
-    gr.Image("avatar.png", elem_id="agent-avatar")
+    #Avatar image
+    gr.Image("avatar.png", elem_id="agent-avatar")  # No shape styling, just default
+
+    #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 with welcome message
     chatbot = gr.Chatbot(type="messages", value=[
         {"role": "system", "content": "🌙 Welcome to Infinite Agent — your guide to emotional clarity, self-worth, and life direction."}
     ])
+
+    #Input textbox and clear button
     msg = gr.Textbox(placeholder="Ask about your emotions, direction, or purpose...")
     clear = gr.Button("Clear")
 
-    # Actions
+    #Submit & 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")
+    clear.click(lambda: [], None, chatbot, queue=False)  # clears chat
 
-# Launch the app
 demo.launch()