Upload inference_server.py with huggingface_hub

inference_server.py

@@ -238,14 +238,17 @@ def predict_bottleneck_risk(req: BottleneckRequest):
     complexity_score = req.constraint_complexity * tc_log
     hours_ratio = req.hours_logged / max(req.estimated_hours, 1)
     stage_idx = STAGE_IDX.get(req.current_stage, 1)
+    hours_budget_pct = req.hours_logged / max(req.estimated_hours, 1) * 100
+    stage_velocity = req.hours_logged / max(stage_idx, 1)
     
     features = np.array([[
         tech_enc, type_enc, priority_enc, tc_log,
         int(req.has_dependencies), req.num_dependencies,
         req.constraint_complexity, req.estimated_hours, req.hours_logged,
-        hours_ratio, req.drc_iterations, req.drc_violations_total,
-        req.lvs_mismatches_total, stage_idx, req.days_in_current_stage,
-        req.engineer_skill_factor, int(req.is_overdue), complexity_score
+        req.drc_iterations, req.drc_violations_total,
+        req.lvs_mismatches_total, stage_idx,
+        req.engineer_skill_factor, complexity_score,
+        hours_budget_pct, stage_velocity
     ]])
     
     risk_idx = bottleneck_model.predict(features)[0]