Reverse merge V0.176.4 hotfix (#2112)

* UN-3621 [HOTFIX] Structure tool no longer crashes when single-pass extraction returns a non-object output (#2110)

UN-3621 [FIX] Guard structure pipeline against non-dict single-pass output

Single-pass extraction can return a top-level JSON array (e.g. a truncated/
runaway LLM response that hit its output-token cap). The parsed `output` is
then a list, and _handle_structure_pipeline called `.values()` on it
unconditionally, raising an opaque `AttributeError: 'list' object has no
attribute 'values'` that failed the file with no actionable signal.

Guard the output shape: if it isn't a dict, return a clear ExecutionResult
failure naming the likely cause instead of crashing, and stop the malformed
payload from flowing downstream as a success.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* [HOTFIX] Raise URLValidator max_length to 8192 for long S3 presigned URLs (#2111)

---------

Co-authored-by: Kirtiman Mishra <110175055+kirtimanmishrazipstack@users.noreply.github.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: 3 people

backend/backend/settings/base.py

@@ -15,10 +15,17 @@
 from urllib.parse import quote
 
 import httpx
+from django.core.validators import URLValidator
 from dotenv import find_dotenv, load_dotenv
 from utils.common_utils import CommonUtils
 from utils.cors_origin import normalize_web_app_origin
 
+# Django 5.0+ caps URLValidator at 2048 chars. S3 pre-signed URLs signed with
+# temporary/STS credentials (carrying X-Amz-Security-Token) routinely exceed this,
+# causing "Enter a valid URL." on the API deployment `presigned_urls` field.
+# Raise the cap globally. No-op on Django 4.2.x (no such attribute is checked).
+URLValidator.max_length = 8192
+
 missing_settings = []
 
 

workers/executor/executors/legacy_executor.py

@@ -703,6 +703,19 @@ def _failure(child_result: ExecutionResult) -> ExecutionResult:
         )
 
         output_map = structured_output.get(PSKeys.OUTPUT, {}) or {}
+        if not isinstance(output_map, dict):
+            # Single-pass can return a non-dict (e.g. a top-level JSON array) when the
+            # LLM emits a truncated/runaway response. Fail clearly instead of crashing
+            # on .values(), and keep the malformed shape from flowing downstream as a
+            # success.
+            return ExecutionResult.failure(
+                error=(
+                    f"Single-pass extraction returned a {type(output_map).__name__}, "
+                    "expected a field map; the LLM likely produced a truncated or "
+                    "runaway response (check output-token usage)."
+                ),
+                metadata={"usage_records": pipeline_records},
+            )
         answered = sum(1 for v in output_map.values() if v not in (None, "", [], {}))
         shim.stream_log(f"Pipeline completed: {answered}/{len(outputs)} prompts answered")
         out_metadata = {