Update app.py

app.py

@@ -3,25 +3,25 @@ import streamlit as st
 from huggingface_hub import InferenceClient
 
 # ── Configuration ──────────────────────────────────────────────────────────
-HF_TOKEN        = os.getenv("HF_TOKEN")  # store your token in Space Secrets
-MODEL_ID        = "sourize/phi2-memory-lora"
-CONTEXT_TURNS   = 7
-MAX_NEW_TOKENS  = 128
-SYSTEM_PROMPT   = (
-    "You are a helpful assistant for DeepTalks with base Phi-2 "
-    "fine-tuned by Sourish.\n"
+HF_TOKEN       = os.getenv("HF_TOKEN")           
+MODEL_ID       = "sourize/phi2-memory-lora"      
+CONTEXT_TURNS  = 6
+MAX_NEW_TOKENS = 128
+
+SYSTEM_PROMPT = (
+    "You are a helpful assistant for DeepTalks, based on Phi-2 fine-tuned by Sourish.\n"
     "Answer **only** using the conversation context below.\n"
     "Do NOT output any lines beginning with 'User:' or 'Assistant:'.\n"
     "If you don't know, say \"I don't know.\"\n"
 )
 
-# ── HF Inference client ─────────────────────────────────────────────────────
+# ── Initialize the HF Inference client ─────────────────────────────────────
 client = InferenceClient(token=HF_TOKEN)
 
 def query_hf(prompt: str) -> str:
-    out = client.text_generation(
+    resp = client.text_generation(
         model=MODEL_ID,
-        inputs=prompt,
+        prompt=prompt,
         parameters={
             "max_new_tokens": MAX_NEW_TOKENS,
             "do_sample": True,
@@ -30,47 +30,49 @@ def query_hf(prompt: str) -> str:
             "return_full_text": False
         },
     )
-    text = out.generated_text.strip()
-    # strip any stray markers
-    for marker in ["User:", "Assistant:"]:
+    text = resp.generated_text.strip()
+    # strip any accidental echoes
+    for marker in ("User:", "Assistant:"):
         if marker in text:
             text = text.split(marker)[0].strip()
     return text or "I don't know."
 
 # ── Streamlit UI ──────────────────────────────────────────────────────────
 st.set_page_config(layout="centered")
-st.title("🧠 DeepTalks (Inference API)")
-st.subheader("Your personal AI Companion")
+st.title("🧠 DeepTalks")
+st.subheader("Your personal AI Companion", divider='grey')
 
 if "history" not in st.session_state:
-    st.session_state.history = []  # tuples of (role, text)
+    st.session_state.history = []  # list of (role, text)
 
-# render history
-for role, text in st.session_state.history:
-    st.chat_message("user" if role=="You" else "assistant").write(text)
+# render past messages
+for role, txt in st.session_state.history:
+    st.chat_message("user" if role == "You" else "assistant").write(txt)
 
-# new input
+# new user input
 user_input = st.chat_input("Type your message…")
 if user_input:
+    # display & store user
     st.chat_message("user").write(user_input)
     st.session_state.history.append(("You", user_input))
 
     # build context
     recent = st.session_state.history[-CONTEXT_TURNS*2:]
-    ctx = "\n".join(text for _, text in recent)
-    prompt = (
+    context = "\n".join(msg for _, msg in recent)
+    prompt  = (
         f"{SYSTEM_PROMPT}\n\n"
-        f"Context:\n{ctx}\n\n"
+        f"Context:\n{context}\n\n"
         f"User: {user_input}\nAssistant:"
     )
 
-    # call HF Inference API
+    # call the HF Inference API
     with st.spinner("Thinking…"):
         try:
-            reply = query_hf(prompt)
+            answer = query_hf(prompt)
         except Exception as e:
-            reply = "Error generating response."
-            st.error(e)
+            st.error(f"Inference API error: {e}")
+            answer = "I’m sorry, something went wrong."
 
-    st.chat_message("assistant").write(reply)
-    st.session_state.history.append(("Bot", reply))
+    # display & store assistant
+    st.chat_message("assistant").write(answer)
+    st.session_state.history.append(("Bot", answer))