handle empty pages in doc-extract

GooeyAI · Feb 12, 2024 · 7f5adb9 · 7f5adb9
1 parent 620f873
commit 7f5adb9
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 1 deletion.
diff --git a/daras_ai_v2/azure_doc_extract.py b/daras_ai_v2/azure_doc_extract.py
@@ -67,6 +67,8 @@ def azure_form_recognizer(url: str, model_id: str, params: dict = None):
 
 
 def extract_records(result: dict, page_num: int) -> list[dict]:
+    if not result:
+        return []
     table_polys = extract_tables(result, page_num)
     records = []
     for para in result["paragraphs"]:

diff --git a/recipes/DocExtract.py b/recipes/DocExtract.py
@@ -424,7 +424,11 @@ def process_source(
                 params = dict(pages=str(page_num))
             else:
                 params = None
-            transcript = str(azure_doc_extract_pages(content_url, params=params)[0])
+            pages = azure_doc_extract_pages(content_url, params=params)
+            if pages and pages[0]:
+                transcript = str(pages[0])
+            else:
+                transcript = ""
         else:
             raise NotImplementedError(
                 f"Unsupported type {doc_meta and doc_meta.mime_type} for {webpage_url}"