Upload 2 files

backend/media_analysis/public_routes.py

@@ -10,6 +10,8 @@ import time
 from collections import defaultdict, deque
 from threading import Lock, BoundedSemaphore
 from flask import Blueprint, jsonify, request
+from pathlib import PurePosixPath
+from urllib.parse import unquote
 from .figure_analysis import analyze_media_question, MediaSelectionError
 from .vision_client import VisionConfigurationError
 
@@ -118,12 +120,30 @@ def media_analyze_public():
         sources = result.get("sources") or result.get("source_identifiers") or []
         if len(sources) != 1:
             return jsonify({"error": "Media source mismatch: expected exactly one source"}), 502
-        src_label = str(sources[0].get("label") or "")
-        src_match = re.search(r"\b(fig(?:ure)?|table)[\s_.-]*0*(\d+)\b", src_label, re.I)
-        if (not src_match or
-            ("table" if src_match.group(1).lower() == "table" else "figure") != requested_kind or
-            int(src_match.group(2)) != requested_number):
-            return jsonify({"error": "Media source mismatch: requested and returned labels differ"}), 502
+        # Gallery selections are authorized by a precise image filename which
+        # prepare_media_evidence already resolved against this paper's manifest.
+        # Preserve its original label even if it differs from the gallery's
+        # display number.  This is especially important when source PDFs have
+        # duplicated printed table numbers or inconsistent extraction labels.
+        source = sources[0]
+        if media_image:
+            source_url = str(source.get("image_url") or "")
+            image_returned = PurePosixPath(unquote(source_url).split("?", 1)[0]).name
+            if image_returned != media_image:
+                return jsonify({"error": "Media source mismatch: selected image differs from the returned image"}), 502
+        else:
+            # Typed requests have no unique image: require the returned
+            # manifest label to match the explicitly requested number.
+            src_label = str(source.get("label") or "")
+            src_match = re.search(r"\b(fig(?:ure)?|table)[\s_.-]*0*(\d+)\b", src_label, re.I)
+            if (not src_match or
+                ("table" if src_match.group(1).lower() == "table" else "figure") != requested_kind or
+                int(src_match.group(2)) != requested_number):
+                return jsonify({"error": "Media source mismatch: requested and returned labels differ"}), 502
+        # Tell the frontend which image was validated. It should not re-apply
+        # a display-label check for exact gallery selections.
+        if media_image:
+            result["selected_media_image"] = media_image
         # Do not send large duplicate tabular records: CSV already carries the data.
         result.pop("structured_data", None)
         if "answer" in result: