Use live retrieval result instead of Neuroxen fallback

gradio_pharmacopilot_demo.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import json
 import os
+import unicodedata
 import time
 from difflib import SequenceMatcher, get_close_matches
 from pathlib import Path
@@ -94,7 +95,9 @@ SESSION_SEARCHES = 0
 
 
 def normalize(text: str) -> str:
-    return " ".join((text or "").strip().lower().split())
+    text = unicodedata.normalize("NFKD", str(text or ""))
+    text = "".join(ch for ch in text if not unicodedata.combining(ch))
+    return " ".join(text.strip().lower().split())
 
 
 def clean_prediction(raw_prediction: str) -> str:
@@ -128,12 +131,11 @@ def label_for_medicine(ocr_text: str, medicine: dict[str, Any]) -> str:
     return brands[0] if brands else medicine["name"]
 
 
-def find_medicine_from_ocr(ocr_text: str) -> tuple[dict[str, Any], list[dict[str, Any]], str]:
+def find_medicine_from_ocr(ocr_text: str) -> tuple[dict[str, Any], list[dict[str, Any]], str, int]:
     query = normalize(ocr_text)
     corrected_query = query
     canonical = BD_BRAND_TO_GENERIC.get(corrected_query, corrected_query)
-    medicine = MED_BY_NAME.get(normalize(canonical)) or MED_BY_NAME.get("naproxen") or MEDICINES[0]
-    display_name = label_for_medicine(ocr_text, medicine)
+    direct_medicine = MED_BY_NAME.get(normalize(canonical))
 
     candidate_names = set()
     for med in MEDICINES:
@@ -152,7 +154,21 @@ def find_medicine_from_ocr(ocr_text: str) -> tuple[dict[str, Any], list[dict[str
                 scored.append({"label": name, "medicine": med, "score": score})
 
     scored.sort(key=lambda item: item["score"], reverse=True)
-    top = [{"label": display_name, "medicine": medicine, "score": 0.97}]
+    if direct_medicine:
+        medicine = direct_medicine
+        display_name = label_for_medicine(ocr_text, medicine)
+        primary_score = 0.97
+    elif scored:
+        best = scored[0]
+        medicine = best["medicine"]
+        display_name = best["label"]
+        primary_score = best["score"]
+    else:
+        medicine = MEDICINES[0]
+        display_name = clean_prediction(ocr_text) or "Needs review"
+        primary_score = 0.0
+
+    top = [{"label": display_name, "medicine": medicine, "score": primary_score}]
     seen_ids = {medicine["id"]}
     for item in scored:
         if item["medicine"]["id"] in seen_ids:
@@ -173,7 +189,8 @@ def find_medicine_from_ocr(ocr_text: str) -> tuple[dict[str, Any], list[dict[str
                 continue
         break
 
-    return medicine, top, display_name
+    confidence = max(0, min(99, round(primary_score * 100)))
+    return medicine, top, display_name, confidence
 
 
 def get_inventory(medicine: dict[str, Any]) -> dict[str, Any]:
@@ -291,6 +308,7 @@ def medicine_details_html(
     inventory: dict[str, Any],
     ocr_text: str,
     display_name: str,
+    confidence: int,
 ) -> str:
     return f"""
     <div class="result-card">
@@ -300,7 +318,7 @@ def medicine_details_html(
         <dt>Generic</dt><dd>{medicine.get('name', 'Unknown')}</dd>
         <dt>Strength</dt><dd>{first_strength(medicine.get('strength', ''))}</dd>
         <dt>Manufacturer</dt><dd>{medicine.get('manufacturer') or 'Not listed'}</dd>
-        <dt>Confidence</dt><dd>97%</dd>
+        <dt>Confidence</dt><dd>{confidence}%</dd>
         <dt>Category</dt><dd>{medicine.get('category', 'General')}</dd>
         <dt>Price</dt><dd>PKR 145</dd>
       </dl>
@@ -421,7 +439,7 @@ def analyze_prescription(image, progress=gr.Progress()):
         progress(pct, desc=label)
         time.sleep(0.25)
 
-    medicine, candidates, display_name = find_medicine_from_ocr(ocr_text)
+    medicine, candidates, display_name, confidence = find_medicine_from_ocr(ocr_text)
     inventory = get_inventory(medicine)
     image_path = resolve_asset_path(medicine.get("image_path"))
     package_image = str(image_path) if image_path else None
@@ -438,10 +456,10 @@ def analyze_prescription(image, progress=gr.Progress()):
     return (
         load_kpi_metrics(SESSION_SEARCHES),
         pipeline_html(5),
-        medicine_details_html(medicine, inventory, ocr_text, display_name),
+        medicine_details_html(medicine, inventory, ocr_text, display_name, confidence),
         package_image,
         package_status_html(inventory),
-        confidence_gauge(97),
+        confidence_gauge(confidence),
         candidates_html(candidates),
         ocr_compare_html(medicine, ocr_text, display_name),
         gr.update(visible=True),