Fixed what I broke I think this time.

core/memory.py

@@ -1,16 +1,17 @@
+# core/memory.py
 def build_prompt_from_history(history, user_text, k=4) -> str:
     """
     history is a list of [user, bot] pairs (Gradio Chatbot format).
-    We embed short storefront reminders to reduce hallucinations.
+    Keep a compact, factual system preface to ground the model.
     """
     lines = [
-        "You are answering questions about a university graduation storefront.",
-        "Products:",
-        "- Cap & Gown Set (CG-SET) — $59.00: Tassel included; ships until 10 days before the event",
-        "- Parking Pass (PK-1) — $10.00: Multiple passes allowed per student",
-        "Venue rules: Formal attire recommended (not required). No muscle shirts. No sagging pants.",
-        "Parking rules: No double parking. Vehicles parked in handicap spaces will be towed.",
-        "Answer concisely using these facts. If unsure, say what’s known from the list above."
+        "System: Answer questions about the university graduation storefront using the facts below.",
+        "System: Be concise. If unsure, say what is known.",
+        "Facts:",
+        "- Cap & Gown Set (CG-SET): $59.00, tassel included; ships until 10 days before the event.",
+        "- Parking Pass (PK-1): $10.00; multiple passes allowed per student.",
+        "- Venue: formal attire recommended; no muscle shirts; no sagging pants.",
+        "- Parking: no double parking; vehicles in handicap spaces will be towed.",
     ]
     for u, b in (history or [])[-k:]:
         if u: lines.append(f"User: {u}")
@@ -18,3 +19,4 @@ def build_prompt_from_history(history, user_text, k=4) -> str:
     lines.append(f"User: {user_text}")
     lines.append("Assistant:")
     return "\n".join(lines)
+

core/model.py

@@ -1,3 +1,4 @@
+# core/model.py
 import os
 from transformers import pipeline
 

core/storefront.py

@@ -97,3 +97,28 @@ def storefront_qna(data, user_text: str):
             return "\n".join(lines)
 
     return None
+
+# app_storefront.py
+
+def clean_generation(text: str) -> str:
+    s = (text or "").strip()
+
+    # If the prompt contained "Assistant:", keep only what comes after the last one
+    last = s.rfind("Assistant:")
+    if last != -1:
+        s = s[last + len("Assistant:"):].strip()
+
+    # If it accidentally continued into a new "User:" or instructions, cut there
+    cut_marks = ["\nUser:", "\nYOU ARE ANSWERING", "\nProducts:", "\nVenue rules:", "\nParking rules:"]
+    cut_positions = [s.find(m) for m in cut_marks if s.find(m) != -1]
+    if cut_positions:
+        s = s[:min(cut_positions)].strip()
+
+    # Collapse repeated lines like "Yes, multiple parking passes..." spam
+    lines, out = s.splitlines(), []
+    seen = set()
+    for ln in lines:
+        # dedupe only exact consecutive repeats; keep normal conversation lines
+        if not out or ln != out[-1]:
+            out.append(ln)
+    return "\n".join(out).strip()

space_app.py

@@ -1,41 +0,0 @@
-import os
-import gradio as gr
-from transformers import pipeline
-
-MODEL_NAME = os.getenv("HF_MODEL_GENERATION", "distilgpt2")
-
-_pipe = None
-def _get_pipe():
-    global _pipe
-    if _pipe is None:
-        _pipe = pipeline("text-generation", model=MODEL_NAME)
-    return _pipe
-
-def chat_fn(message, max_new_tokens=128, temperature=0.8, top_p=0.95):
-    message = (message or "").strip()
-    if not message:
-        return "Please type something!"
-    pipe = _get_pipe()
-    out = pipe(
-        message,
-        max_new_tokens=int(max_new_tokens),
-        do_sample=True,
-        temperature=float(temperature),
-        top_p=float(top_p),
-        pad_token_id=50256
-    )
-    return out[0]["generated_text"]
-
-with gr.Blocks(title="Agentic-Chat-bot") as demo:
-    gr.Markdown("# 🤖 Agentic Chat Bot\nGradio + Transformers demo")
-    prompt = gr.Textbox(label="Prompt", placeholder="Ask me anything…", lines=4)
-    out = gr.Textbox(label="Response", lines=6)
-    max_new = gr.Slider(32, 512, 128, 1, label="Max new tokens")
-    temp = gr.Slider(0.1, 1.5, 0.8, 0.05, label="Temperature")
-    topp = gr.Slider(0.1, 1.0, 0.95, 0.05, label="Top-p")
-    btn = gr.Button("Send")
-    btn.click(chat_fn, [prompt, max_new, temp, topp], out)
-    prompt.submit(chat_fn, [prompt, max_new, temp, topp], out)
-
-if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", "7860")))