Upload 2 files

app_jerry_logistic.py

@@ -0,0 +1,336 @@
+
+import gradio as gr
+import json, re, math
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+# --------------------
+# State/HUD structure
+# --------------------
+DEFAULTS = {
+    "model": "logistic",
+    # Logistic yield: Y(X) = L / (1 + exp(-k*(X - x0)))
+    "L": None,     # max yield
+    "k": None,     # slope
+    "x0": None,    # inflection X
+    # Prices & costs
+    "Px": None,    # input price ($/lb of fert)  -> MIC
+    "Py": None,    # output price ($/unit yield)
+    "Other": 0.0,  # other cost per acre (fixed wrt X)
+    # Range
+    "x_min": 0.0,
+    "x_max": 150.0,
+    "x_step": 5.0,
+    # Computed last table/plot
+    "table_ready": False,
+    "want_plot": False,
+    "last_table_cols": [],
+    "last_focus_x": None  # for showing sample calc on a specific X
+}
+
+HELP = '''Say things like:
+- "use logistic: L=1200, k=0.06, x0=60"
+- "set Py=0.5 and Px=0.25 and other cost=300"
+- "range 0 to 150 by 5"
+- "make table"
+- "graph yield" (or "plot mpp" / "plot profit")
+- "compute APP" / "compute MPP" / "compute MVP" / "compute MIC"
+- "show production stages"
+- "calculate profit"
+- "find optimal" (uses rule MVP = MIC in Stage II)
+- "change Py to 0.6" or "change Px 0.3"
+'''
+
+# --------------------
+# Core economics
+# --------------------
+def Y_logistic(X, L, k, x0):
+    return L / (1.0 + np.exp(-k * (X - x0)))
+
+def MPP_logistic(X, L, k, x0):
+    # derivative of logistic: dY/dX = L*k*exp(-k*(X - x0)) / (1 + exp(-k*(X - x0)))^2
+    e = np.exp(-k * (X - x0))
+    return L * k * e / (1.0 + e)**2
+
+def table_from_hud(hud):
+    L, k, x0 = hud["L"], hud["k"], hud["x0"]
+    x_min, x_max, x_step = hud["x_min"], hud["x_max"], hud["x_step"]
+    if None in (L, k, x0):
+        raise ValueError("Set logistic params first (L, k, x0).")
+    X = np.arange(float(x_min), float(x_max) + 1e-9, float(x_step))
+    Y = Y_logistic(X, L, k, x0)
+    APP = np.divide(Y, X, out=np.zeros_like(Y), where=X>0)
+    MPP = MPP_logistic(X, L, k, x0)
+    data = {
+        "X": X,
+        "Y": Y,
+        "APP": APP,
+        "MPP": MPP
+    }
+    # Add MVP, MIC, Profit if prices exist
+    if hud.get("Py") is not None:
+        data["MVP"] = MPP * float(hud["Py"])
+    if hud.get("Px") is not None:
+        data["MIC"] = np.full_like(X, float(hud["Px"]), dtype=float)
+    if hud.get("Py") is not None and hud.get("Px") is not None:
+        Other = float(hud.get("Other", 0.0))
+        data["Profit"] = Y * float(hud["Py"]) - X * float(hud["Px"]) - Other
+    df = pd.DataFrame(data)
+    # Stage classification
+    stage = []
+    for i, row in df.iterrows():
+        if row["MPP"] > 0:
+            if row["APP"] >= row["MPP"]:
+                s = "I"
+            else:
+                s = "II"
+        else:
+            s = "III"
+        stage.append(s)
+    df["Stage"] = stage
+    return df
+
+def find_optimal(hud):
+    # Find X that maximizes profit by MVP = MIC (Stage II)
+    df = table_from_hud(hud)
+    if "MVP" not in df.columns or "MIC" not in df.columns:
+        raise ValueError("Need Py (for MVP) and Px (for MIC).")
+    # Find index minimizing |MVP - MIC| with MPP>0 (Stage II)
+    mask = (df["Stage"] == "II")
+    if not mask.any():
+        # fallback: allow all
+        mask = df["MPP"] > 0
+    sub = df[mask].copy()
+    sub["gap"] = (sub["MVP"] - sub["MIC"]).abs()
+    j = sub["gap"].idxmin()
+    row = df.loc[j]
+    return row  # contains X, Y, Profit, etc.
+
+# --------------------
+# NLP parser
+# --------------------
+def parse(user, hud):
+    updated = []
+
+    txt = user.strip()
+
+    # logistic params
+    # "L=1200, k=0.06, x0=60" (order flexible)
+    mL = re.search(r'\bL\s*=\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    mk = re.search(r'\bk\s*=\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    mx0 = re.search(r'\bx0\s*=\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    if mL:
+        hud["L"] = float(mL.group(1)); updated.append("L")
+    if mk:
+        hud["k"] = float(mk.group(1)); updated.append("k")
+    if mx0:
+        hud["x0"] = float(mx0.group(1)); updated.append("x0")
+
+    if re.search(r'logistic', txt, flags=re.I):
+        hud["model"] = "logistic"; updated.append("model")
+
+    # prices
+    mPx = re.search(r'(?:Px|MIC|input price)\s*=?\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    mPy = re.search(r'(?:Py|output price)\s*=?\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    mOther = re.search(r'(?:Other|other cost)\s*=?\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    if mPx:
+        hud["Px"] = float(mPx.group(1)); updated.append("Px")
+    if mPy:
+        hud["Py"] = float(mPy.group(1)); updated.append("Py")
+    if mOther:
+        hud["Other"] = float(mOther.group(1)); updated.append("Other")
+
+    # quick "change Py to 0.6" etc.
+    mchg = re.search(r'change\s+(px|py|other)\s*(?:to|=)\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    if mchg:
+        key = mchg.group(1).title()
+        hud[key] = float(mchg.group(2)); updated.append(key)
+
+    # range
+    mrange = re.search(r'(?:range|fertilize|apply|from)\s*([0-9]*\.?[0-9]+)\s*(?:to|-)\s*([0-9]*\.?[0-9]+)\s*(?:by|in|step)\s*([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    if mrange:
+        hud["x_min"] = float(mrange.group(1)); updated.append("x_min")
+        hud["x_max"] = float(mrange.group(2)); updated.append("x_max")
+        hud["x_step"] = float(mrange.group(3)); updated.append("x_step")
+
+    # focus X for blackboard example: "at X=80" or "at 80 lbs"
+    mfocus = re.search(r'at\s*(?:x\s*=?\s*)?([0-9]*\.?[0-9]+)', txt, flags=re.I)
+    if mfocus:
+        hud["last_focus_x"] = float(mfocus.group(1)); updated.append("last_focus_x")
+
+    # intents
+    intent = None
+    if re.search(r'\bmake table|\btable', txt, flags=re.I):
+        intent = "table"
+    if re.search(r'graph|plot', txt, flags=re.I):
+        # accept "plot yield", "plot mpp", "plot profit", default "yield"
+        intent = "plot"
+    if re.search(r'compute\s+app|\bAPP\b', txt, flags=re.I):
+        intent = "app"
+    if re.search(r'compute\s+mpp|\bMPP\b', txt, flags=re.I):
+        intent = "mpp"
+    if re.search(r'compute\s+mvp|\bMVP\b', txt, flags=re.I):
+        intent = "mvp"
+    if re.search(r'compute\s+mic|\bMIC\b', txt, flags=re.I):
+        intent = "mic"
+    if re.search(r'profit', txt, flags=re.I):
+        intent = "profit"
+    if re.search(r'optimal|find optimal|mvp\s*=\s*mic', txt, flags=re.I):
+        intent = "optimal"
+    if re.search(r'stage', txt, flags=re.I):
+        intent = "stage"
+    if re.search(r'help', txt, flags=re.I):
+        intent = "help"
+
+    return hud, updated, intent, txt
+
+# --------------------
+# Blackboard examples
+# --------------------
+def sample_work(hud, df):
+    # Return a short textual sample calculation at a representative X.
+    if hud.get("last_focus_x") is not None:
+        x0 = hud["last_focus_x"]
+        j = int(round((x0 - hud["x_min"]) / hud["x_step"]))
+        j = max(0, min(j, len(df)-1))
+    else:
+        j = len(df)//2
+    row = df.iloc[j]
+    lines = []
+    lines.append(f"At X = {row['X']:.3g}:")
+    lines.append(f"  Y = L/(1+exp(-k*(X-x0))) with L={hud['L']}, k={hud['k']}, x0={hud['x0']}")
+    lines.append(f"  => Y ≈ {row['Y']:.4g}")
+    lines.append(f"  APP = Y/X ≈ {row['APP']:.4g}")
+    lines.append(f"  MPP = dY/dX ≈ {row['MPP']:.4g}")
+    if 'MVP' in df.columns:
+        lines.append(f"  MVP = MPP·Py ≈ {row['MVP']:.4g}")
+    if 'MIC' in df.columns:
+        lines.append(f"  MIC = Px ≈ {row['MIC']:.4g}")
+    if 'Profit' in df.columns:
+        lines.append(f"  Profit = Py·Y − Px·X − Other ≈ {row['Profit']:.4g}")
+    lines.append(f"  Stage = {row['Stage']}")
+    return "\n".join(lines)
+
+# --------------------
+# Controller
+# --------------------
+def controller(user_text, hud_json):
+    # load/ensure hud
+    try:
+        hud = json.loads(hud_json) if hud_json else {}
+    except Exception:
+        hud = {}
+    for k, v in DEFAULTS.items():
+        hud.setdefault(k, v)
+
+    hud, updated, intent, raw = parse(user_text, hud)
+
+    # Decide action
+    reply = ""
+    table = pd.DataFrame()
+    plot = None
+    blackboard = ""
+
+    # Some actions require table
+    need_table = intent in ("table","plot","app","mpp","mvp","mic","profit","stage","optimal")
+    if need_table:
+        try:
+            df = table_from_hud(hud)
+        except Exception as e:
+            reply = f"Jerry → {e} \n\n{HELP}"
+            return reply, json.dumps(hud, indent=2), ", ".join(updated) or "(none)", blackboard, table, plot
+
+    if intent == "help" or intent is None:
+        reply = "Jerry → How can I help?\n\n" + HELP
+
+    elif intent == "table":
+        hud["table_ready"] = True
+        hud["last_table_cols"] = list(df.columns)
+        reply = f"Built table for X from {hud['x_min']} to {hud['x_max']} by {hud['x_step']}."
+        table = df
+        blackboard = sample_work(hud, df)
+
+    elif intent == "plot":
+        # choose what to plot
+        metric = "Y"
+        if re.search(r'mpp', raw, flags=re.I): metric = "MPP"
+        elif re.search(r'app', raw, flags=re.I): metric = "APP"
+        elif re.search(r'mvp', raw, flags=re.I) and "MVP" in df.columns: metric = "MVP"
+        elif re.search(r'profit', raw, flags=re.I) and "Profit" in df.columns: metric = "Profit"
+        fig = plt.figure()
+        ax = fig.add_subplot(111)
+        ax.plot(df["X"], df[metric])
+        ax.set_xlabel("X (fertilizer)")
+        ax.set_ylabel(metric)
+        ax.set_title(f"{metric} vs X")
+        plot = fig
+        table = df
+        reply = f"Plotted {metric} vs X."
+        blackboard = sample_work(hud, df)
+
+    elif intent in ("app","mpp","mvp","mic","profit","stage"):
+        table = df
+        parts = []
+        if intent in ("app","stage"): parts.append("APP")
+        if intent in ("mpp","stage"): parts.append("MPP")
+        if intent in ("mvp","stage") and "MVP" in df.columns: parts.append("MVP")
+        if intent in ("mic","stage") and "MIC" in df.columns: parts.append("MIC")
+        if intent in ("profit","stage") and "Profit" in df.columns: parts.append("Profit")
+        if intent == "stage": parts.append("Stage")
+        show = [c for c in parts if c in df.columns] + ["X"]
+        show = list(dict.fromkeys(["X"] + show))  # ensure X first, unique
+        reply = "Computed metrics. Showing relevant columns."
+        table = df[show]
+        blackboard = sample_work(hud, df)
+        if intent == "mvp" and "MVP" not in df.columns:
+            reply += " (Set Py first)"
+        if intent == "mic" and "MIC" not in df.columns:
+            reply += " (Set Px first)"
+        if intent == "profit" and "Profit" not in df.columns:
+            reply += " (Set Px and Py first)"
+
+    elif intent == "optimal":
+        try:
+            row = find_optimal(hud)
+            reply = (f"Optimal (MVP ≈ MIC in Stage II):\n"
+                     f"X* ≈ {row['X']:.4g}, Y* ≈ {row['Y']:.4g}, "
+                     f"MVP ≈ {row.get('MVP', float('nan')):.4g}, MIC ≈ {row.get('MIC', float('nan')):.4g}")
+            if 'Profit' in row:
+                reply += f", Profit* ≈ {row['Profit']:.4g}"
+            df = table_from_hud(hud)
+            cols = ['X','Y','MPP']
+            if 'MVP' in df.columns: cols.append('MVP')
+            if 'MIC' in df.columns: cols.append('MIC')
+            if 'Profit' in df.columns: cols.append('Profit')
+            table = df[cols]
+            blackboard = sample_work(hud, df) + "\n\nRule: choose X where MVP = MIC (in Stage II)."
+        except Exception as e:
+            reply = f"Jerry → {e} \n(Set Px and Py, and logistic params)."
+
+    else:
+        reply = "Okay — updated HUD. Ask me to 'make table', 'plot yield', 'compute MPP', 'profit', 'find optimal', etc."
+
+    return reply, json.dumps(hud, indent=2), ", ".join(updated) or "(none)", blackboard, table, plot
+
+with gr.Blocks() as demo:
+    gr.Markdown("### Jerry — MIC/Profit Coach (Logistic Yield, NLP HUD)")
+    with gr.Row():
+        with gr.Column(scale=1):
+            user = gr.Textbox(label="Talk to Jerry", value="use logistic L=1200, k=0.06, x0=60; set Py=0.5, Px=0.25, other cost=300; range 0 to 150 by 5; make table", lines=4)
+            btn = gr.Button("Ask Jerry", variant="primary")
+        with gr.Column(scale=1):
+            out = gr.Textbox(label="Jerry says", lines=10)
+    with gr.Row():
+        with gr.Column(scale=1):
+            hud = gr.Textbox(label="Student HUD — what Jerry sees (persists)", value=json.dumps(DEFAULTS, indent=2), lines=18)
+        with gr.Column(scale=1):
+            changed = gr.Textbox(label="Fields updated by last prompt", value="(none)")
+    black = gr.Textbox(label="Blackboard — show the work", value="(none)", lines=12)
+    table = gr.Dataframe(label="Main Output Table", wrap=True)
+    plot = gr.Plot(label="Figure (on demand)")
+
+    btn.click(controller, inputs=[user, hud], outputs=[out, hud, changed, black, table, plot])
+
+if __name__ == "__main__":
+    demo.launch()

requirements_logistic.txt

@@ -0,0 +1,4 @@
+gradio>=4.44.0
+pandas
+numpy
+matplotlib