fix: (EL-4040, 4041) improve image processing and add toggle for proc…

alita_sdk/tools/ado/wiki/ado_wrapper.py

@@ -508,6 +508,11 @@ def _process_images(self, page_content: str, wiki_identified: str, image_descrip
                 logger.error(f"Failed to initialize repos wrapper for wiki '{wiki_identified}': {str(e)}")
 
         for image_name, image_url in matches:
+            # BUG 1 fix: skip images with empty URLs — cannot fetch or resolve them
+            if not image_url:
+                logger.warning(f"Skipping image '{image_name}': empty URL, leaving original markdown unchanged.")
+                continue
+
             if image_url.startswith("/.attachments/"):
                 try:
                     if repos_wrapper is None:
@@ -517,26 +522,24 @@ def _process_images(self, page_content: str, wiki_identified: str, image_descrip
                                                           image_description_prompt=image_description_prompt,
                                                           repos_wrapper=repos_wrapper)
                 except Exception as e:
-                    logger.error(f"Error parsing attachment: {str(e)}")
-                    description = f"Error parsing attachment: {image_url}"
+                    # Skip rather than replace with a corrupted description
+                    logger.warning(f"Skipping image '{image_name}': error parsing attachment '{image_url}': {str(e)}")
+                    continue
             else:
-                if not image_url:
-                    logger.warning(f"Skipping image '{image_name}' with empty URL")
-                    description = "[Image could not be processed: empty URL]"
-                else:
-                    try:
-                        response = requests.get(image_url)
-                        response.raise_for_status()
-                        file_content = response.content
-                        description = parse_file_content(
-                            file_content=file_content,
-                            file_name="image.png",
-                            llm=self.llm,
-                            prompt=image_description_prompt
-                        )
-                    except Exception as e:
-                        logger.error(f"Error fetching external image: {str(e)}")
-                        description = f"Error fetching external image: {image_url}"
+                try:
+                    response = requests.get(image_url)
+                    response.raise_for_status()
+                    file_content = response.content
+                    description = parse_file_content(
+                        file_content=file_content,
+                        file_name="image.png",
+                        llm=self.llm,
+                        prompt=image_description_prompt
+                    )
+                except Exception as e:
+                    # BUG 2 fix: skip rather than replace with a corrupted literal description
+                    logger.warning(f"Skipping image '{image_name}': error fetching external image '{image_url}': {str(e)}")
+                    continue
 
             new_image_markdown = f"![{image_name}]({description})"
             page_content = page_content.replace(f"![{image_name}]({image_url})", new_image_markdown)

alita_sdk/tools/jira/api_wrapper.py

@@ -120,14 +120,16 @@
     jira_issue_key=(str, Field(description="Jira issue key from which field with images will be extracted, e.g. TEST-1234")),
     field_name=(str, Field(description="Field name containing images to be processed. Common values are 'description', 'comment', or custom fields like 'customfield_10300'")),
     prompt=(Optional[str], Field(description="Custom prompt to use for image description generation. If not provided, a default prompt will be used", default=None)),
-    context_radius=(Optional[int], Field(description="Number of characters to include before and after each image for context. Default is 500", default=500))
+    context_radius=(Optional[int], Field(description="Number of characters to include before and after each image for context. Default is 500", default=500)),
+    process_images=(Optional[bool], Field(description="Whether to process images with LLM and replace references with descriptions. Set to False to return raw field content without image processing. Default is True", default=True))
 )
 
 GetCommentsWithImageDescriptions = create_model(
     "GetCommentsWithImageDescriptionsModel",
     jira_issue_key=(str, Field(description="Jira issue key from which comments with images will be extracted, e.g. TEST-1234")),
     prompt=(Optional[str], Field(description="Custom prompt to use for image description generation. If not provided, a default prompt will be used", default=None)),
-    context_radius=(Optional[int], Field(description="Number of characters to include before and after each image for context. Default is 500", default=500))
+    context_radius=(Optional[int], Field(description="Number of characters to include before and after each image for context. Default is 500", default=500)),
+    process_images=(Optional[bool], Field(description="Whether to process images with LLM and replace references with descriptions. Set to False to return raw comments without image processing. Default is True", default=True))
 )
 
 GetRemoteLinks = create_model(
@@ -1270,7 +1272,9 @@ def _process_image_with_llm(self, image_data, image_name: str = "", context_text
             # Get the LLM instance
             llm = self.llm
             if not llm:
-                return "[LLM not available for image processing]"
+                raise ToolException(
+                    "LLM is required for image processing but is not configured in this toolkit instance."
+                )
 
             # Try to load and validate the image with PIL instead of using imghdr
             try:
@@ -1451,25 +1455,28 @@ def _extract_image_data(self, field_data):
         return f"Unsupported field content type: {type(field_data)}. Expected a string, list, or dict."
 
     def get_field_with_image_descriptions(self, jira_issue_key: str, field_name: str, prompt: Optional[str] = None,
-                                          context_radius: int = 500):
+                                          context_radius: int = 500, process_images: bool = True):
         """
         Get a field from Jira issue and augment any images in it with textual descriptions that include
         image names and contextual information from surrounding text.
 
         This method will:
         1. Extract the specified field content from Jira
         2. Detect images in the content
-        3. Retrieve and process each image with an LLM, providing surrounding context
-        4. Replace image references with the generated text descriptions
+        3. Retrieve and process each image with an LLM, providing surrounding context (if process_images=True)
+        4. Replace image references with the generated text descriptions (if process_images=True)
 
         Args:
             jira_issue_key: The Jira issue key to retrieve field from (e.g., 'TEST-1234')
             field_name: The field containing images (e.g., 'description')
             prompt: Custom prompt for the LLM when analyzing images. If None, a default prompt will be used.
             context_radius: Number of characters to include before and after each image for context. Default is 500.
+            process_images: Whether to process images with LLM. Set to False to return raw field content
+                            without image processing. Default is True.
 
         Returns:
-            The field content with image references replaced with contextual descriptions
+            The field content with image references replaced with contextual descriptions,
+            or raw field content when process_images=False
         """
         try:
             # Get the specified field from the Jira issue
@@ -1485,6 +1492,16 @@ def get_field_with_image_descriptions(self, jira_issue_key: str, field_name: str
             # Handle multiple images or non-string content
             field_content = self._extract_image_data(field_content)
 
+            if not process_images:
+                logger.info(f"Image processing skipped for field '{field_name}' of issue '{jira_issue_key}' (process_images=False)")
+                return f"Field '{field_name}' from issue '{jira_issue_key}':\n\n{field_content}"
+
+            if not self.llm:
+                raise ToolException(
+                    "LLM is required for image processing but is not configured in this toolkit instance. "
+                    "Please configure an LLM in the toolkit settings or set process_images=False to skip image processing."
+                )
+
             # Regular expression to find image references in Jira markup
             image_pattern = r'!([^!|]+)(?:\|[^!]*)?!'
 
@@ -1594,7 +1611,7 @@ def process_image_match(self, match, body, attachment_resolver, context_radius=5
             logger.error(f"Error retrieving attachment {image_ref}: {str(e)}")
             return f"[Image: {image_ref} - Error: {str(e)}]"
 
-    def get_processed_comments_list_with_image_description(self, jira_issue_key: str, prompt: Optional[str] = None, context_radius: int = 500):
+    def get_processed_comments_list_with_image_description(self, jira_issue_key: str, prompt: Optional[str] = None, context_radius: int = 500, process_images: bool = True):
         # Retrieve all comments for the issue
         comments = self._client.issue_get_comments(jira_issue_key)
 
@@ -1619,10 +1636,13 @@ def get_processed_comments_list_with_image_description(self, jira_issue_key: str
             comment_created = comment.get('created', 'Unknown date')
             comment_body = self._extract_image_data(comment_body)
 
-            # Process the comment body by replacing image references with descriptions
-            processed_body = re.sub(image_pattern,
-                                    lambda match: self.process_image_match(match, comment_body, attachment_resolver, context_radius, prompt),
-                                    comment_body)
+            if process_images:
+                # Process the comment body by replacing image references with descriptions
+                processed_body = re.sub(image_pattern,
+                                        lambda match: self.process_image_match(match, comment_body, attachment_resolver, context_radius, prompt),
+                                        comment_body)
+            else:
+                processed_body = comment_body
 
             # Add the processed comment to our results
             processed_comments.append({
@@ -1635,28 +1655,37 @@ def get_processed_comments_list_with_image_description(self, jira_issue_key: str
 
         return processed_comments
 
-    def get_comments_with_image_descriptions(self, jira_issue_key: str, prompt: Optional[str] = None, context_radius: int = 500):
+    def get_comments_with_image_descriptions(self, jira_issue_key: str, prompt: Optional[str] = None, context_radius: int = 500, process_images: bool = True):
         """
         Get all comments from Jira issue and augment any images in them with textual descriptions.
 
         This method will:
         1. Extract all comments from the specified Jira issue
         2. Detect images in each comment
-        3. Retrieve and process each image with an LLM, providing surrounding context
-        4. Replace image references with the generated text descriptions
+        3. Retrieve and process each image with an LLM, providing surrounding context (if process_images=True)
+        4. Replace image references with the generated text descriptions (if process_images=True)
 
         Args:
             jira_issue_key: The Jira issue key to retrieve comments from (e.g., 'TEST-1234')
             prompt: Custom prompt for the LLM when analyzing images. If None, a default prompt will be used.
             context_radius: Number of characters to include before and after each image for context. Default is 500.
+            process_images: Whether to process images with LLM. Set to False to return raw comments without
+                            image processing. Default is True.
 
         Returns:
-            The comments with image references replaced with contextual descriptions
+            The comments with image references replaced with contextual descriptions,
+            or raw comments when process_images=False
         """
         try:
+            if process_images and not self.llm:
+                raise ToolException(
+                    "LLM is required for image processing but is not configured in this toolkit instance. "
+                    "Please configure an LLM in the toolkit settings or set process_images=False to skip image processing."
+                )
             processed_comments = self.get_processed_comments_list_with_image_description(jira_issue_key=jira_issue_key,
                                                                                          prompt=prompt,
-                                                                                         context_radius=context_radius)
+                                                                                         context_radius=context_radius,
+                                                                                         process_images=process_images)
             if not processed_comments:
                 return f"No comments found for issue '{jira_issue_key}'"
             # Format the output