Spaces:

10gen
/

deepsearchitv2

Running

App Files Files Community

Guiyom commited on Feb 27, 2025

Commit

db66925

verified ·

1 Parent(s): 3844fc5

Update app.py

Browse files

Files changed (1) hide show

app.py +37 -21

app.py CHANGED Viewed

@@ -96,30 +96,46 @@ def clean_llm_response(response: str) -> str:
 def expand_snippet_area(full_html: str, snippet: str) -> str:
     """
-    Given the full HTML and a snippet of text, this function finds the element containing the snippet
-    and traverses upward through the DOM to identify the first container tag that matches the allowed list.
-    Allowed container tags include div, table, and iframe.
-    The traversal stops as soon as a parent in the allowed list is found, or before reaching <body>.
-    Returns the outer HTML of the found container.
     """
-    allowed_tags = {"div", "table", "iframe"}
     soup = BeautifulSoup(full_html, "html.parser")
-    # Find the tag containing the snippet
-    candidate = soup.find(lambda tag: snippet in tag.get_text() if tag.get_text() else False)
-    if candidate:
-        current = candidate
-        # Traverse upward to find the first matching container in the allowed list
-        while current.parent is not None and current.parent.name.lower() != "body" and current.parent.name.lower() != "html":
-            if current.parent.name.lower() in allowed_tags:
-                current = current.parent
-                break  # Stop as soon as a match is found
-            else:
-                current = current.parent
-        return str(current)  # Return the outer HTML of the found container
     else:
-        return snippet  # If snippet not found, return the snippet itself
 def fine_tune_report(adjustment_request: str, openai_api_key: str, serpapi_api_key: str, report_html: str,
@@ -209,7 +225,7 @@ Only output valid JSON."""
 You are a technical editor.
 Given the following HTML container (with its outer tags) extracted from a larger report and based on the user adjustment request,
 produce a corrected version by making only the necessary changes. Preserve inline citations, formatting, and context.
-The updated version will put back precisely in the same location.
 // Context
 - Overall Report HTML:

 def expand_snippet_area(full_html: str, snippet: str) -> str:
     """
+    Given the full HTML and a snippet of text, this function finds the element that contains the snippet.
+    It then uses an iterative while loop to traverse upward (from the immediate parent to the top)
+    until the highest level iframe is reached or (if no iframe is present) until a div or table is
+    encountered—the first allowed container (div or table) found is used. If neither an iframe nor
+    an allowed container is found, it returns the candidate element itself.
     """
+    # Define allowed container tags for the non-iframe case.
+    allowed_tags = {"div", "table"}
     soup = BeautifulSoup(full_html, "html.parser")
+    # Find the candidate element that contains the snippet.
+    candidate = soup.find(lambda tag: tag.get_text() and snippet in tag.get_text())
+    if not candidate:
+        return snippet
+    iframe_candidate = None
+    allowed_candidate = None
+    # Start with the candidate's direct parent.
+    current = candidate.parent
+    while current is not None and current.name.lower() != "body":
+        tag_name = current.name.lower()
+        if tag_name == "iframe":
+            # Whenever we find an iframe, update the iframe candidate.
+            iframe_candidate = current
+        elif tag_name in allowed_tags and allowed_candidate is None:
+            # Record the first encountered allowed container if no iframe was encountered yet.
+            allowed_candidate = current
+        # Move upward.
+        current = current.parent
+    # If an iframe was encountered, return the outermost iframe.
+    if iframe_candidate is not None:
+        return str(iframe_candidate)
+    # Otherwise, if a div or table was found, return that.
+    elif allowed_candidate is not None:
+        return str(allowed_candidate)
+    # Otherwise, return the candidate element itself.
     else:
+        return str(candidate)
 def fine_tune_report(adjustment_request: str, openai_api_key: str, serpapi_api_key: str, report_html: str,
 You are a technical editor.
 Given the following HTML container (with its outer tags) extracted from a larger report and based on the user adjustment request,
 produce a corrected version by making only the necessary changes. Preserve inline citations, formatting, and context.
+The updated version will put back precisely in the same location, the output should have the same outer tags.
 // Context
 - Overall Report HTML: