Update app.py

app.py

@@ -1,6 +1,12 @@
+# app.py (Gradio) — Delight AI Agent Prototype with RFQ Simulation
+# ✅ Stable on Hugging Face Gradio Spaces
+# ✅ Dummy supplier DB (~50) stored in /tmp
+# ✅ Inquiry → extract → shortlist → RFQ simulation → pick best → quote draft → download
+
 import os
 import re
 import random
+import uuid
 from datetime import datetime, timedelta
 
 import pandas as pd
@@ -85,7 +91,7 @@ def regenerate_db():
 
 
 # ----------------------------
-# Lightweight inquiry parsing (no LLM)
+# Parsing
 # ----------------------------
 def normalize_text(t: str) -> str:
     return re.sub(r"\s+", " ", (t or "").strip().lower())
@@ -182,6 +188,9 @@ def pick_margin(pricing_mode: str, base_margin: float):
     return base_margin
 
 
+# ----------------------------
+# Offers + RFQ Simulation
+# ----------------------------
 def compute_offers(req: dict, suppliers: pd.DataFrame, margin_pct: float):
     category = req.get("category")
     brand = req.get("brand")
@@ -207,14 +216,14 @@ def compute_offers(req: dict, suppliers: pd.DataFrame, margin_pct: float):
     rows = []
     for _, s in candidates.iterrows():
         factor = float(s["price_competitiveness_factor"])
-        supplier_cost = market_mid * factor * random.uniform(0.92, 1.06)
+        supplier_cost_est = market_mid * factor * random.uniform(0.92, 1.06)
 
         region = str(s["region"])
         if region in ["Johor", "KL", "Batam"]:
-            supplier_cost *= 1.05
+            supplier_cost_est *= 1.05
 
-        supplier_cost = round(supplier_cost, 2)
-        sell_price = round(supplier_cost / (1 - margin_pct / 100.0), 2)
+        supplier_cost_est = round(supplier_cost_est, 2)
+        sell_price = round(supplier_cost_est / (1 - margin_pct / 100.0), 2)
 
         reliability = float(s["reliability_score"])
         lead = int(s["lead_time_days"])
@@ -227,7 +236,7 @@ def compute_offers(req: dict, suppliers: pd.DataFrame, margin_pct: float):
             "reliability_score": reliability,
             "lead_time_days": lead,
             "moq": int(s["moq"]),
-            "est_supplier_cost_sgd": supplier_cost,
+            "est_supplier_cost_sgd": supplier_cost_est,
             "recommended_sell_price_sgd": sell_price,
             "score": round(score, 4),
             "contact_email": s["contact_email"],
@@ -237,27 +246,72 @@ def compute_offers(req: dict, suppliers: pd.DataFrame, margin_pct: float):
     return offers, (market_lo, market_hi)
 
 
-def build_quote_text(req, offers_df, pricing_mode, margin_used, company_name, customer_name, valid_days):
+def simulate_rfq(offers_df: pd.DataFrame, qty: int, n: int = 5) -> pd.DataFrame:
+    """
+    Simulate supplier quote replies for top N suppliers.
+    - Price = est cost with small variance
+    - Lead time = existing lead +/- small variance
+    - Response time depends on reliability
+    """
     if offers_df.empty:
-        return (
-            "No matching suppliers found.\n\n"
-            "New Product Mode:\n"
-            "1) Search market range online\n"
-            "2) Identify supplier categories\n"
-            "3) Send RFQs to shortlisted suppliers\n"
-            "4) Update internal catalog once confirmed\n"
-        )
+        return pd.DataFrame()
 
+    top = offers_df.head(n).copy()
+
+    rfq_rows = []
+    for _, row in top.iterrows():
+        reliability = float(row["reliability_score"])
+        base_cost = float(row["est_supplier_cost_sgd"])
+        base_lead = int(row["lead_time_days"])
+
+        # More reliable suppliers give tighter pricing variance and faster response
+        price_variance = (1.5 - reliability) * 0.08  # 0.04–0.07ish
+        quoted_unit = base_cost * random.uniform(1 - price_variance, 1 + price_variance)
+        quoted_unit = round(quoted_unit, 2)
+
+        # Volume discount for higher qty (simple demo)
+        if qty >= 100:
+            quoted_unit = round(quoted_unit * 0.97, 2)
+        elif qty >= 50:
+            quoted_unit = round(quoted_unit * 0.985, 2)
+
+        lead = base_lead + random.choice([-2, -1, 0, 1, 2])
+        lead = max(2, lead)
+
+        response_hours = int(max(1, (1.2 - reliability) * random.uniform(3, 12)))
+
+        rfq_rows.append({
+            "rfq_id": f"RFQ-{uuid.uuid4().hex[:6].upper()}",
+            "supplier_name": row["supplier_name"],
+            "supplier_id": row["supplier_id"],
+            "region": row["region"],
+            "reliability_score": reliability,
+            "quoted_unit_price_sgd": quoted_unit,
+            "quoted_total_sgd": round(quoted_unit * qty, 2),
+            "quoted_lead_days": lead,
+            "response_time_hours": response_hours,
+        })
+
+    rfq_df = pd.DataFrame(rfq_rows)
+
+    # Best = lowest unit price, tie-breaker: lead time, then reliability
+    rfq_df = rfq_df.sort_values(
+        ["quoted_unit_price_sgd", "quoted_lead_days", "reliability_score"],
+        ascending=[True, True, False]
+    ).reset_index(drop=True)
+
+    rfq_df.insert(0, "rank", range(1, len(rfq_df) + 1))
+    return rfq_df
+
+
+def build_quote_text(req, unit_price_sgd, lead_days, pricing_mode, margin_used, company_name, customer_name, valid_days):
     qty = int(req.get("quantity") or 10)
     category = req.get("category") or "Lighting Product"
     brand = req.get("brand") or "Brand-agnostic"
     wattage = f"{req.get('wattage')}W" if req.get("wattage") else ""
-    unit_price = float(offers_df.iloc[0]["recommended_sell_price_sgd"])
-    total = round(unit_price * qty, 2)
+    total = round(unit_price_sgd * qty, 2)
     valid_until = (datetime.today() + timedelta(days=int(valid_days))).strftime("%Y-%m-%d")
 
-    best = offers_df.iloc[0]
-
     return f"""Subject: Quotation - {brand} {wattage} {category} (Qty: {qty})
 
 Hi {customer_name},
@@ -266,10 +320,10 @@ Thanks for your inquiry. Please find our quotation below:
 
 Item: {brand} {wattage} {category}
 Quantity: {qty}
-Unit Price: SGD {unit_price:.2f}
+Unit Price: SGD {unit_price_sgd:.2f}
 Total: SGD {total:.2f}
 
-Estimated Lead Time: {int(best["lead_time_days"])} days
+Estimated Lead Time: {int(lead_days)} days
 Validity: Until {valid_until}
 Pricing Mode: {pricing_mode} (Margin applied: {margin_used:.0f}%)
 
@@ -279,41 +333,111 @@ Sales Team
 """
 
 
+# ----------------------------
+# Agent runners
+# ----------------------------
 def run_agent(inquiry_text, base_margin, pricing_mode, top_n, company_name, customer_name, valid_days):
     req = parse_inquiry(inquiry_text)
     margin_used = pick_margin(pricing_mode, float(base_margin))
 
     offers_df, market_rng = compute_offers(req, SUPPLIERS_DF, margin_used)
 
-    market_text = "Estimated market range: "
     if market_rng:
-        market_text += f"SGD {market_rng[0]:.2f} – {market_rng[1]:.2f} per unit"
+        market_text = f"Estimated market range: SGD {market_rng[0]:.2f} – {market_rng[1]:.2f} per unit"
     else:
         lo, hi = estimate_market_range(req.get("category"), req.get("wattage"))
-        market_text += f"SGD {lo:.2f} – {hi:.2f} per unit"
+        market_text = f"Estimated market range: SGD {lo:.2f} – {hi:.2f} per unit"
 
     steps = []
     steps.append(f"Step 1 — Extracted requirement: {req}")
     steps.append(f"Step 2 — Market intelligence: {market_text}")
     steps.append(f"Step 3 — Pricing mode: {pricing_mode} | Margin applied: {margin_used:.0f}%")
+
     if offers_df.empty:
-        steps.append("Step 4 — No internal matches found → New Product Mode triggered.")
-    else:
-        steps.append(f"Step 4 — Shortlisted {min(len(offers_df), int(top_n))} suppliers; top recommendation: {offers_df.iloc[0]['supplier_name']}")
+        steps.append("Step 4 — No internal matches found → New Product Mode.")
+        quote_text = (
+            "No matching suppliers found.\n\n"
+            "New Product Mode:\n"
+            "1) Research market range online\n"
+            "2) Identify supplier categories\n"
+            "3) Send RFQs to shortlisted suppliers\n"
+            "4) Update internal catalog once confirmed\n"
+        )
 
-    offers_view = offers_df.head(int(top_n)) if not offers_df.empty else pd.DataFrame(
-        columns=["supplier_id","supplier_name","region","reliability_score","lead_time_days","moq","est_supplier_cost_sgd","recommended_sell_price_sgd","score","contact_email"]
+        empty_offers = pd.DataFrame(columns=[
+            "supplier_id","supplier_name","region","reliability_score","lead_time_days","moq",
+            "est_supplier_cost_sgd","recommended_sell_price_sgd","score","contact_email"
+        ])
+        empty_rfq = pd.DataFrame(columns=[
+            "rank","rfq_id","supplier_name","supplier_id","region","reliability_score",
+            "quoted_unit_price_sgd","quoted_total_sgd","quoted_lead_days","response_time_hours"
+        ])
+
+        quote_path = "/tmp/quote_draft.txt"
+        with open(quote_path, "w", encoding="utf-8") as f:
+            f.write(quote_text)
+
+        return req, "\n".join(steps), market_text, empty_offers, empty_rfq, "", quote_text, quote_path, DATA_PATH
+
+    steps.append(f"Step 4 — Shortlisted top {int(top_n)} suppliers from internal DB.")
+
+    offers_view = offers_df.head(int(top_n)).copy()
+
+    # Default quote based on internal estimate from best supplier
+    best = offers_view.iloc[0]
+    est_cost = float(best["est_supplier_cost_sgd"])
+    unit_price = round(est_cost / (1 - margin_used / 100.0), 2)
+    lead_days = int(best["lead_time_days"])
+
+    quote_text = build_quote_text(req, unit_price, lead_days, pricing_mode, margin_used, company_name, customer_name, valid_days)
+
+    quote_path = "/tmp/quote_draft.txt"
+    with open(quote_path, "w", encoding="utf-8") as f:
+        f.write(quote_text)
+
+    # RFQ area empty until user clicks simulate
+    empty_rfq = pd.DataFrame(columns=[
+        "rank","rfq_id","supplier_name","supplier_id","region","reliability_score",
+        "quoted_unit_price_sgd","quoted_total_sgd","quoted_lead_days","response_time_hours"
+    ])
+
+    summary = (
+        f"Current recommendation (internal estimate): {best['supplier_name']} | "
+        f"Est Cost SGD {est_cost:.2f}/unit → Sell SGD {unit_price:.2f}/unit | Lead {lead_days} days"
+    )
+
+    return req, "\n".join(steps), market_text, offers_view, empty_rfq, summary, quote_text, quote_path, DATA_PATH
+
+
+def run_rfq_simulation(req, offers_df, base_margin, pricing_mode, company_name, customer_name, valid_days):
+    if offers_df is None or len(offers_df) == 0:
+        return pd.DataFrame(), "No offers available for RFQ simulation.", "", "", None
+
+    qty = int(req.get("quantity") or 10)
+    margin_used = pick_margin(pricing_mode, float(base_margin))
+
+    rfq_df = simulate_rfq(offers_df, qty=qty, n=min(5, len(offers_df)))
+
+    best = rfq_df.iloc[0]
+    rfq_best_unit = float(best["quoted_unit_price_sgd"])
+    rfq_best_lead = int(best["quoted_lead_days"])
+
+    # Update selling price based on RFQ best unit cost
+    sell_unit = round(rfq_best_unit / (1 - margin_used / 100.0), 2)
+
+    summary = (
+        f"RFQ best supplier: {best['supplier_name']} | "
+        f"Quoted Cost SGD {rfq_best_unit:.2f}/unit → Sell SGD {sell_unit:.2f}/unit | "
+        f"Lead {rfq_best_lead} days | Response ~{int(best['response_time_hours'])}h"
     )
 
-    quote_text = build_quote_text(req, offers_df, pricing_mode, margin_used, company_name, customer_name, valid_days)
+    quote_text = build_quote_text(req, sell_unit, rfq_best_lead, pricing_mode, margin_used, company_name, customer_name, valid_days)
 
-    # Write quote text to a file for download
     quote_path = "/tmp/quote_draft.txt"
     with open(quote_path, "w", encoding="utf-8") as f:
         f.write(quote_text)
 
-    # Return: parsed req (json-like), agent steps, market text, offers table, quote text, downloadable file, downloadable csv
-    return req, "\n".join(steps), market_text, offers_view, quote_text, quote_path, DATA_PATH
+    return rfq_df, "✅ RFQ simulation completed. Best quote selected.", summary, quote_text, quote_path
 
 
 # ----------------------------
@@ -323,16 +447,16 @@ with gr.Blocks(title="Delight AI Agent (Prototype)") as demo:
     gr.Markdown(
         """
 # 💡 Delight AI Agent (Prototype)
-Paste a customer inquiry → agent extracts requirement → ranks suppliers → recommends pricing → generates quotation draft.
+Paste a customer inquiry → agent extracts requirement → ranks suppliers → (simulated RFQ) → recommends pricing → generates quotation draft.
 """
     )
 
-    with gr.Row():
-        inquiry = gr.Textbox(
-            label="Customer Inquiry",
-            lines=6,
-            value="Hi, please quote best price for 50 pcs Philips 18W LED panel light. Delivery to Singapore in 2 weeks."
-        )
+    # Inputs
+    inquiry = gr.Textbox(
+        label="Customer Inquiry",
+        lines=6,
+        value="Hi, please quote best price for 50 pcs Philips 18W LED panel light. Delivery to Singapore in 2 weeks."
+    )
 
     with gr.Row():
         base_margin = gr.Slider(5, 40, value=20, step=1, label="Base Margin (%)")
@@ -344,8 +468,10 @@ Paste a customer inquiry → agent extracts requirement → ranks suppliers →
         customer_name = gr.Textbox(label="Customer name", value="Customer")
         valid_days = gr.Slider(1, 30, value=7, step=1, label="Quote validity (days)")
 
+    # Buttons
     run_btn = gr.Button("🚀 Run Agent")
 
+    # Outputs
     with gr.Row():
         parsed_req = gr.JSON(label="Extracted Requirement")
         agent_steps = gr.Textbox(label="Agent Steps", lines=10)
@@ -353,11 +479,19 @@ Paste a customer inquiry → agent extracts requirement → ranks suppliers →
     market_info = gr.Textbox(label="Market Intelligence", lines=2)
 
     offers_table = gr.Dataframe(
-        label="Recommended Supplier Options (Top N)",
+        label="Supplier Shortlist (Top N)",
         interactive=False,
         wrap=True
     )
 
+    gr.Markdown("## 📡 RFQ Simulation (Agentic Step)")
+    rfq_btn = gr.Button("📨 Simulate RFQ to Top Suppliers")
+
+    rfq_status = gr.Textbox(label="RFQ Status", lines=2)
+    rfq_table = gr.Dataframe(label="RFQ Responses (Simulated)", interactive=False, wrap=True)
+
+    recommendation = gr.Textbox(label="Recommendation Summary", lines=2)
+
     quote_text = gr.Textbox(label="Generated Quote Draft", lines=14)
 
     with gr.Row():
@@ -369,10 +503,25 @@ Paste a customer inquiry → agent extracts requirement → ranks suppliers →
         db_preview = gr.Dataframe(label="DB Preview (Top 10)", interactive=False)
         db_file = gr.File(label="Download Fresh DB (.csv)")
 
+    # State
+    state_req = gr.State({})
+    state_offers = gr.State(pd.DataFrame())
+
+    # Wire actions
     run_btn.click(
         fn=run_agent,
         inputs=[inquiry, base_margin, pricing_mode, top_n, company_name, customer_name, valid_days],
-        outputs=[parsed_req, agent_steps, market_info, offers_table, quote_text, quote_file, supplier_csv],
+        outputs=[parsed_req, agent_steps, market_info, offers_table, rfq_table, recommendation, quote_text, quote_file, supplier_csv],
+    ).then(
+        fn=lambda req, offers: (req, offers),
+        inputs=[parsed_req, offers_table],
+        outputs=[state_req, state_offers],
+    )
+
+    rfq_btn.click(
+        fn=run_rfq_simulation,
+        inputs=[state_req, state_offers, base_margin, pricing_mode, company_name, customer_name, valid_days],
+        outputs=[rfq_table, rfq_status, recommendation, quote_text, quote_file],
     )
 
     regen_btn.click(
@@ -382,5 +531,4 @@ Paste a customer inquiry → agent extracts requirement → ranks suppliers →
     )
 
 if __name__ == "__main__":
-    # Hugging Face uses 7860 by default; this makes it explicit and stable.
     demo.launch(server_name="0.0.0.0", server_port=7860)