Update app.py

app.py

@@ -2,9 +2,9 @@ import os
 import streamlit as st
 from huggingface_hub import InferenceApi
 
-# ── Configuration ──────────────────────────────────────────────────────────
+# ── Config ────────────────────────────────────────────────────────────────
 MODEL_ID       = "sourize/phi2-memory-lora"
-HF_TOKEN       = os.getenv("HF_TOKEN")   # set in your Space secrets
+HF_TOKEN       = os.getenv("HF_TOKEN")
 CONTEXT_TURNS  = 6
 MAX_NEW_TOKENS = 128
 TEMPERATURE    = 0.7
@@ -14,11 +14,11 @@ SYSTEM_PROMPT = (
     "You are a helpful assistant for DeepTalks with base Phi-2\n"
     "fine-tuned by Sourish for domain support.\n"
     "Answer **only** using the conversation context below.\n"
-    "Do NOT output any lines beginning with 'User:' or 'Assistant:'.\n"
+    "Do NOT output lines beginning with 'User:' or 'Assistant:'.\n"
     "If you don't know, say \"I don't know.\"\n"
 )
 
-# ── Initialize the low-level Inference API ─────────────────────────────────
+# ── Inference API ─────────────────────────────────────────────────────────
 inference = InferenceApi(
     repo_id=MODEL_ID,
     token=HF_TOKEN,
@@ -26,22 +26,18 @@ inference = InferenceApi(
 )
 
 def query_hf(prompt: str) -> str:
-    """
-    Call the text-generation endpoint via InferenceApi.
-    """
+    # All generation kwargs go at top level
     outputs = inference(
         inputs=prompt,
-        parameters={
-            "max_new_tokens": MAX_NEW_TOKENS,
-            "temperature": TEMPERATURE,
-            "top_p": TOP_P,
-            "do_sample": True,
-            "return_full_text": False
-        }
+        max_new_tokens=MAX_NEW_TOKENS,
+        do_sample=True,
+        temperature=TEMPERATURE,
+        top_p=TOP_P,
+        return_full_text=False,
     )
-    # outputs is a list of dicts: [{"generated_text": "..."}]
+    # returns a list of dicts: [{"generated_text": "..."}]
     text = outputs[0]["generated_text"].strip()
-    # strip any stray markers
+    # strip stray speaker markers
     for marker in ("User:", "Assistant:"):
         if marker in text:
             text = text.split(marker)[0].strip()
@@ -53,38 +49,33 @@ st.title("🧠 DeepTalks (Inference API)")
 st.subheader("Your personal AI Companion")
 
 if "history" not in st.session_state:
-    st.session_state.history = []  # list of (role, text)
+    st.session_state.history = []
 
-# Render previous messages
+# render past chat
 for role, msg in st.session_state.history:
     st.chat_message("user" if role=="You" else "assistant").write(msg)
 
-# New user input
+# user input
 user_input = st.chat_input("Type your message…")
 if user_input:
-    # Display and record user
     st.chat_message("user").write(user_input)
     st.session_state.history.append(("You", user_input))
 
-    # Build context of last 2*CONTEXT_TURNS entries
+    # build context
     recent = st.session_state.history[-CONTEXT_TURNS*2:]
     context = "\n".join(text for _, text in recent)
-
-    prompt = (
+    prompt  = (
         f"{SYSTEM_PROMPT}\n\n"
         f"Context:\n{context}\n\n"
-        f"User: {user_input}\n"
-        f"Assistant:"
+        f"User: {user_input}\nAssistant:"
     )
 
-    # Call Hugging Face Inference API
     with st.spinner("Thinking…"):
         try:
             reply = query_hf(prompt)
         except Exception as e:
-            st.error(f"Inference API error: {e}")
+            st.error(f"Inference error: {e}")
             reply = "I’m sorry, something went wrong."
 
-    # Display and record assistant
     st.chat_message("assistant").write(reply)
     st.session_state.history.append(("Bot", reply))