Update app.py

app.py

@@ -1,17 +1,22 @@
-import os
 import json
 import numpy as np
 from fastapi import FastAPI
 from pydantic import BaseModel
 from sentence_transformers import SentenceTransformer
 from huggingface_hub import hf_hub_download
+import os
 
 # ============================================================
-# CONFIG
+# CONFIG — UPDATE ONLY IF YOU CHANGE REPO NAMES
 # ============================================================
 
-REPO_USER = "ClergeF"
+# Your username
+HF_USER = "ClergeF"
+
+# Your embedder repo
+EMBEDDER_REPO = "MVT-embedder"
 
+# Your 12 model repos
 MODEL_REPOS = {
     "value_impact": "value-impact-model",
     "impact": "impact-model",
@@ -27,60 +32,63 @@ MODEL_REPOS = {
     "innovation": "innovation-model",
 }
 
-# Embedder location in your HF repo:
-EMBEDDER_REPO = "MVT-models"
-EMBEDDER_SUBFOLDER = "universal_embedder"
-
-HF_TOKEN = os.environ.get("HF_TOKEN", None)
+# Files inside each repo
+MODEL_FILES = {
+    "value_impact": "value_impact.json",
+    "impact": "impact.json",
+    "family": "family_level.json",
+    "community": "community_level.json",
+    "education": "education_level.json",
+    "health": "health_level.json",
+    "environment": "environment_level.json",
+    "business": "business_level.json",
+    "finance": "finance_level.json",
+    "history": "history_level.json",
+    "spirituality": "spirituality_level.json",
+    "innovation": "innovation_level.json",
+}
 
 # ============================================================
-# LOAD UNIVERSAL EMBEDDER
+# LOAD EMBEDDER (automatically works in Spaces)
 # ============================================================
 
-print("Loading universal embedder from HuggingFace Hub...")
-
-embedder = SentenceTransformer(
-    f"{REPO_USER}/{EMBEDDER_REPO}",
-    subfolder=EMBEDDER_SUBFOLDER,
-    use_auth_token=HF_TOKEN
-)
+print("Loading SentenceTransformer embedder...")
+embedder = SentenceTransformer(f"{HF_USER}/{EMBEDDER_REPO}")
 
 # ============================================================
-# MODEL LOADING HELPER
+# MODEL LOADING HELPERS
 # ============================================================
 
-def load_model(repo_name, filename):
-    """Download & load .json linear regression model from HF Hub."""
-    model_path = hf_hub_download(
-        repo_id=f"{REPO_USER}/{repo_name}",
-        filename=filename,
-        token=HF_TOKEN,
+def load_json_model(repo_name: str, filename: str):
+    """Download and load model JSON from HuggingFace Hub."""
+    print(f"Loading {repo_name}/{filename} ...")
+
+    path = hf_hub_download(
+        repo_id=f"{HF_USER}/{repo_name}",
+        filename=filename
     )
-    with open(model_path, "r") as f:
+
+    with open(path, "r") as f:
         return json.load(f)
 
-print("Loading all 12 models from Hugging Face Hub...")
-
-models = {
-    "value_impact": load_model(MODEL_REPOS["value_impact"], "value_impact.json"),
-    "impact": load_model(MODEL_REPOS["impact"], "impact.json"),
-    "family": load_model(MODEL_REPOS["family"], "family_level.json"),
-    "community": load_model(MODEL_REPOS["community"], "community_level.json"),
-    "education": load_model(MODEL_REPOS["education"], "education_level.json"),
-    "health": load_model(MODEL_REPOS["health"], "health_level.json"),
-    "environment": load_model(MODEL_REPOS["environment"], "environment_level.json"),
-    "business": load_model(MODEL_REPOS["business"], "business_level.json"),
-    "finance": load_model(MODEL_REPOS["finance"], "finance_level.json"),
-    "history": load_model(MODEL_REPOS["history"], "history_level.json"),
-    "spirituality": load_model(MODEL_REPOS["spirituality"], "spirituality_level.json"),
-    "innovation": load_model(MODEL_REPOS["innovation"], "innovation_level.json"),
-}
+# ============================================================
+# LOAD ALL 12 MODELS ONCE
+# ============================================================
+
+print("Loading linear regression weight files...")
+
+models = {}
+
+for key, repo in MODEL_REPOS.items():
+    models[key] = load_json_model(repo, MODEL_FILES[key])
+
+print("All models successfully loaded!")
 
 # ============================================================
-# FASTAPI
+# FASTAPI SETUP
 # ============================================================
 
-app = FastAPI(title="MVT Community Value Model API")
+app = FastAPI(title="MVT Category Scoring API")
 
 class InputText(BaseModel):
     text: str
@@ -90,21 +98,22 @@ class InputText(BaseModel):
 # ============================================================
 
 def embed(text: str):
+    """Convert text → embedding."""
     return embedder.encode([text])[0]
 
 def linear_predict(model_json, vec):
+    """Apply manual linear regression: dot(coef, x) + intercept."""
     coef = np.array(model_json["coef"])
-    intercept = np.array(model_json["intercept"])
+    intercept = model_json["intercept"]
 
-    # Multi-output (value + impact)
-    if coef.ndim == 2:
-        return coef @ vec + intercept
+    if isinstance(intercept, list):
+        intercept = np.array(intercept)
 
-    # Single scalar output
-    return float(np.dot(coef, vec) + intercept)
+    pred = np.dot(coef, vec) + intercept
+    return float(pred)
 
 # ============================================================
-# API ROUTE
+# MAIN API ROUTE
 # ============================================================
 
 @app.post("/predict")
@@ -114,20 +123,26 @@ def predict(payload: InputText):
 
     result = {}
 
-    # Two-output regression model
-    value_pred, impact_pred = linear_predict(models["value_impact"], vec)
+    # --- Value + Impact model (2 outputs) ---
+    value_impact_model = models["value_impact"]
+    coef = np.array(value_impact_model["coef"])
+    intercept = np.array(value_impact_model["intercept"])
+
+    value_pred, impact_pred = coef @ vec + intercept
     result["estimated_value"] = float(value_pred)
     result["impact_level"] = float(impact_pred)
 
-    # Individual category models
-    for key in [
-        "impact", "family", "community", "education", "health",
-        "environment", "business", "finance", "history",
-        "spirituality", "innovation"
+    # --- 10 category models ---
+    for name in [
+        "family", "community", "education", "health", "environment",
+        "business", "finance", "history", "spirituality", "innovation"
     ]:
-        result[key] = float(linear_predict(models[key], vec))
+        result[f"{name}_score"] = linear_predict(models[name], vec)
+
+    # --- Standalone impact model ---
+    result["impact_model_score"] = linear_predict(models["impact"], vec)
 
     return {
         "input": text,
         "predictions": result
-    }
+    }