Update app.py

app.py

@@ -1,4 +1,5 @@
-# app.py
+
+# app.py  — pickle-only, dynamic inputs
 import os
 import json
 import pickle
@@ -6,104 +7,87 @@ import numpy as np
 import gradio as gr
 from huggingface_hub import hf_hub_download
 
-# If your model repo is private, add a Space secret called HF_TOKEN.
+# If the model repo is private, set this in Space secrets
 HF_TOKEN = os.environ.get("HF_TOKEN")
 
-# You can override this in Space secrets as HF_MODEL_ID if the path changes.
-HF_MODEL_ID = os.environ.get("HF_MODEL_ID", "SpringyBon/entrepreneur-readiness-regressor")
+# Override with a secret if you want; defaults shown here:
+HF_MODEL_ID = os.environ.get("HF_MODEL_ID", "SpringyBon/entrepreneur-readiness-synthetic")
 
-def _download(fname: str) -> str:
-    return hf_hub_download(
-        repo_id=HF_MODEL_ID,
-        filename=fname,
-        repo_type="model",
-        token=HF_TOKEN  # None is fine for public repos; required for private
-    )
+def _dl(name: str) -> str:
+    return hf_hub_download(repo_id=HF_MODEL_ID, filename=name, repo_type="model", token=HF_TOKEN)
 
-# ---- Download artifacts (expects model.pkl at repo root) ----
-model_path = _download("model.pkl")
-feat_path  = _download("feature_names.json")
+# --- load artifacts ---
+model_path = _dl("model.pkl")
+features_path = _dl("feature_names.json")
 
-# ---- Load artifacts ----
 with open(model_path, "rb") as f:
     reg = pickle.load(f)
 
-with open(feat_path, "r") as f:
-    FEATURES = json.load(f)  # list of feature names in the trained order
-
-# ---- Prediction ----
-def predict(
-    savings_amount,
-    monthly_income,
-    monthly_expenses,
-    monthly_entertainment,
-    sales_skills_1to10,
-    age,
-    dependents,
-    assets_count,
-    risk_tolerance_1to10,
-    confidence_1to10,
-    business_idea_difficulty_1to10,
-    income_stability_1to10,
-    prior_experience_years,
-    credit_score
-):
-    sample = {
-        "savings_amount": savings_amount,
-        "monthly_income": monthly_income,
-        "monthly_expenses": monthly_expenses,
-        "monthly_entertainment": monthly_entertainment,
-        "sales_skills_1to10": sales_skills_1to10,
-        "age": age,
-        "dependents": dependents,
-        "assets_count": assets_count,
-        "risk_tolerance_1to10": risk_tolerance_1to10,
-        "confidence_1to10": confidence_1to10,
-        "business_idea_difficulty_1to10": business_idea_difficulty_1to10,
-        "income_stability_1to10": income_stability_1to10,
-        "prior_experience_years": prior_experience_years,
-        "credit_score": credit_score,
-    }
+with open(features_path, "r") as f:
+    FEATURES = json.load(f)  # list of feature names in training order
+
+print("[INFO] n_features_in_:", getattr(reg, "n_features_in_", None))
+print("[INFO] FEATURES:", FEATURES)
+
+def _validate_and_vectorize(values):
+    """values come in the same order as the UI components we generate from FEATURES."""
+    if len(values) != len(FEATURES):
+        return None, f"Feature count mismatch: model expects {len(FEATURES)} but got {len(values)} from UI."
+
+    try:
+        x = [float(v) for v in values]
+    except Exception as e:
+        return None, f"Input formatting error: {type(e).__name__}: {e}"
+
+    n_expected = getattr(reg, "n_features_in_", None)
+    if n_expected is not None and len(x) != int(n_expected):
+        return None, (
+            f"Model expects {n_expected} features but UI provided {len(x)}. "
+            "Double-check feature_names.json vs the trained model."
+        )
+    return np.array(x, dtype=float).reshape(1, -1), None
+
+# --- prediction: accepts a variable number of inputs ---
+def predict(*values):
+    X, err = _validate_and_vectorize(values)
+    if err:
+        return None, err
     try:
-        x = [sample[f] for f in FEATURES]  # keep the trained column order
-    except KeyError as e:
-        return f"Missing feature in UI: {e}"
-    pred = float(reg.predict(np.array(x, dtype=float).reshape(1, -1))[0])
-    return round(pred, 2)
+        y = float(reg.predict(X)[0])
+        return round(y, 2), "OK"
+    except Exception as e:
+        import traceback
+        return None, f"Model error: {type(e).__name__}: {e}\n{traceback.format_exc()}"
+
+# --- UI: build inputs dynamically from FEATURES ---
+def pretty(label: str) -> str:
+    return label.replace("_", " ").title()
 
-# ---- UI ----
 with gr.Blocks(title="Entrepreneur Readiness Regressor") as iface:
     gr.Markdown("## Entrepreneur Readiness — Prediction Demo")
 
+    # Build inputs in the SAME ORDER as FEATURES
+    inputs = []
     with gr.Row():
         with gr.Column():
-            savings_amount = gr.Number(label="Savings Amount", value=25000)
-            monthly_income = gr.Number(label="Monthly Income", value=5000)
-            monthly_expenses = gr.Number(label="Monthly Expenses", value=3000)
-            monthly_entertainment = gr.Number(label="Monthly Entertainment", value=200)
-            sales_skills_1to10 = gr.Slider(1, 10, value=7, step=1, label="Sales Skills (1-10)")
-            age = gr.Number(label="Age", value=32)
-            dependents = gr.Number(label="Dependents", value=1, precision=0)
-
+            half = (len(FEATURES) + 1) // 2
+            left = FEATURES[:half]
+            right = FEATURES[half:]
+            for name in left:
+                # You can customize types here; Number for simplicity
+                inputs.append(gr.Number(label=pretty(name), value=0))
         with gr.Column():
-            assets_count = gr.Number(label="Assets Count", value=2, precision=0)
-            risk_tolerance_1to10 = gr.Slider(1, 10, value=6, step=1, label="Risk Tolerance (1-10)")
-            confidence_1to10 = gr.Slider(1, 10, value=8, step=1, label="Confidence (1-10)")
-            business_idea_difficulty_1to10 = gr.Slider(1, 10, value=5, step=1, label="Idea Difficulty (1-10)")
-            income_stability_1to10 = gr.Slider(1, 10, value=7, step=1, label="Income Stability (1-10)")
-            prior_experience_years = gr.Number(label="Prior Experience (years)", value=4)
-            credit_score = gr.Number(label="Credit Score", value=710, precision=0)
-
-    output = gr.Number(label="Predicted Readiness Score", precision=2)
+            for name in right:
+                inputs.append(gr.Number(label=pretty(name), value=0))
+
+    score_out = gr.Number(label="Predicted Readiness Score", precision=2)
+    msg_out   = gr.Textbox(label="Message", interactive=False)
+
     gr.Button("Predict Readiness").click(
         predict,
-        inputs=[
-            savings_amount, monthly_income, monthly_expenses, monthly_entertainment,
-            sales_skills_1to10, age, dependents, assets_count, risk_tolerance_1to10,
-            confidence_1to10, business_idea_difficulty_1to10, income_stability_1to10, prior_experience_years, credit_score
-        ],
-        outputs=output
+        inputs=inputs,                 # pass list of components (order matches FEATURES)
+        outputs=[score_out, msg_out],  # number + message
     )
 
 if __name__ == "__main__":
-    iface.launch()
+    iface.launch()