Update agents/reasoner.py

agents/reasoner.py

@@ -1,107 +1,58 @@
-import matplotlib.pyplot as plt
-import io
-from matplotlib.figure import Figure
+import pandas as pd
+import numpy as np
 
-# -------------------------------------------------------------
-# UTILITY: Convert Matplotlib Figure → PNG bytes (Gradio compatible)
-# -------------------------------------------------------------
-def fig_to_png_bytes(fig: Figure):
-    """Convert a matplotlib figure to PNG bytes for Gradio Image()."""
-    buf = io.BytesIO()
-    fig.savefig(buf, format="png", dpi=120, bbox_inches="tight")
-    buf.seek(0)
-    return buf.getvalue()
 
-
-# -------------------------------------------------------------
-# SLOTING ANALYSIS
-# -------------------------------------------------------------
 def run_slotting_analysis(message, slotting_df):
-    reasoning = []
-    reasoning.append("Interpreting slotting optimization request.")
-
-    # Example strategy:
-    strategy = "Move slow movers outward to free prime space"
-    reasoning.append(f"Selected strategy: {strategy}")
-
-    # Generate a mock optimized slotting table:
-    optimized = slotting_df.copy()
-    optimized["Aisle"] = [1, 11, 20]   # example optimized aisles
-    optimized["Rack"] = [18, 17, 7]    # example optimized racks
-
-    reasoning.append("Generated optimized slotting plan.")
-
-    explanation = " ".join(reasoning)
-
-    return explanation, optimized
-
-
-
-# -------------------------------------------------------------
-# PICKING ROUTE OPTIMIZATION (WITH PNG OUTPUT)
-# -------------------------------------------------------------
-def run_picking_optimization(message, picking_df):
-    reasoning = []
-    reasoning.append("Analyzing picking request…")
-
-    # Convert to integer pairs
-    coords = list(zip(
-        picking_df["Aisle"].astype(int).tolist(),
-        picking_df["Rack"].astype(int).tolist()
-    ))
-
-    # --- Nearest neighbour route selection ---
-    ordered = [coords.pop(0)]
-    while coords:
-        last = ordered[-1]
-        next_point = min(coords, key=lambda c: abs(c[0] - last[0]) + abs(c[1] - last[1]))
-        ordered.append(next_point)
-        coords.remove(next_point)
-
-    # --- Build matplotlib figure ---
-    fig, ax = plt.subplots(figsize=(5, 5))
-    xs = [x for x, _ in ordered]
-    ys = [y for _, y in ordered]
-    ax.plot(xs, ys, marker="o")
-    ax.set_title("Optimized Picking Route")
-    ax.set_xlabel("Aisle")
-    ax.set_ylabel("Rack")
-
-    reasoning.append("Generated optimized picking route.")
-
-    explanation = " ".join(reasoning)
-
-    # Convert FIG → PNG bytes
-    png_bytes = fig_to_png_bytes(fig)
-
-    # return explanation + PNG bytes (NO slotting table)
-    return explanation, png_bytes
-
-
-
-# -------------------------------------------------------------
-# FULL WAREHOUSE REPORT (WITH GRAPH + TABLE)
-# -------------------------------------------------------------
-def run_full_report(message, slotting_df, picking_df):
-    reasoning = []
-    reasoning.append("Building unified warehouse operations report…")
-
-    # Summary counts
-    fast = (slotting_df["Velocity"] == "Fast").sum()
-    med = (slotting_df["Velocity"] == "Medium").sum()
-    slow = (slotting_df["Velocity"] == "Slow").sum()
-
-    summary = f"""
-    Warehouse Summary:
-    - Fast movers: {fast}
-    - Medium movers: {med}
-    - Slow movers: {slow}
-    """
-
-    explanation = " ".join(reasoning) + "\n" + summary
-
-    # Include picking graph in the full report
-    _, fig_bytes = run_picking_optimization(message, picking_df)
-
-    # Return: explanation, PNG route graph, AND slotting table
-    return explanation, fig_bytes, slotting_df.copy()
+    reasoning_steps = []
+
+    df = slotting_df.copy()
+
+    # ---------------------------------------------------------
+    # 1️⃣ Convert velocity to numerical weight
+    # ---------------------------------------------------------
+    velocity_map = {
+        "fast": 3,
+        "medium": 2,
+        "slow": 1
+    }
+
+    df["VelocityNorm"] = df["Velocity"].str.lower().map(velocity_map)
+    reasoning_steps.append("Mapped velocity categories to numerical weights.")
+
+    # ---------------------------------------------------------
+    # 2️⃣ Normalize frequency (usage)
+    # ---------------------------------------------------------
+    df["FreqNorm"] = (df["Frequency"] - df["Frequency"].min()) / (
+        df["Frequency"].max() - df["Frequency"].min() + 1e-8
+    )
+    reasoning_steps.append("Normalized frequency to 0–1 scale.")
+
+    # ---------------------------------------------------------
+    # 3️⃣ Compute Slotting Score
+    # ---------------------------------------------------------
+    # Weighted: 60% velocity + 40% frequency
+    df["Score"] = (0.6 * df["VelocityNorm"]) + (0.4 * df["FreqNorm"])
+    reasoning_steps.append("Computed weighted slotting score (60% velocity, 40% frequency).")
+
+    # ---------------------------------------------------------
+    # 4️⃣ Assign Aisles and Racks based on score
+    # ---------------------------------------------------------
+    df = df.sort_values("Score", ascending=False).reset_index(drop=True)
+
+    # Prime aisles 1–5 → best locations
+    aisles = np.arange(1, len(df) + 1)
+    racks = np.linspace(1, 20, len(df)).astype(int)
+
+    df["Aisle"] = aisles
+    df["Rack"] = racks
+    reasoning_steps.append("Assigned optimal aisle & rack positions based on ranking.")
+
+    explanation = (
+        "### 📦 Slotting Optimization\n"
+        "The SKUs were evaluated using a weighted model combining velocity and picking frequency.\n"
+        "High-velocity & high-frequency items are placed closer to prime aisles.\n\n"
+        "#### 🔍 Key Reasoning Steps:\n"
+        + "\n".join([f"- {r}" for r in reasoning_steps])
+    )
+
+    return explanation, df