fix: fix duplicate images

source/lambda/job/dep/llm_bot_dep/figure_llm.py

@@ -92,7 +92,14 @@ def invoke_llm(self, img, prompt, prefix="<output>", stop="</output>"):
             {
                 "role": "user",
                 "content": [
-                    {"type": "image", "source": {"type": "base64", "media_type": "image/jpeg", "data": base64_encoded}},
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/jpeg",
+                            "data": base64_encoded,
+                        },
+                    },
                     {"type": "text", "text": prompt},
                 ],
             },
@@ -132,7 +139,10 @@ def get_description(self, img, context, tag):
     def get_mermaid(self, img, classification):
         with open(MERMAID_TEMPLATE_PATH) as f:
             mermaid_prompt = f.read()
-        prompt = mermaid_prompt.format(diagram_type=classification, diagram_example=self.mermaid_prompt[classification])
+        prompt = mermaid_prompt.format(
+            diagram_type=classification,
+            diagram_example=self.mermaid_prompt[classification],
+        )
         output = self.invoke_llm(img, prompt, prefix="<description>", stop="</mermaid>")
         return output
 
@@ -192,7 +202,12 @@ def encode_image_to_base64(image_path: str) -> str:
 
 
 def upload_image_to_s3(
-    image_data: Union[str, bytes], bucket: str, file_name: str, splitting_type: str, idx: int, is_bytes: bool = False
+    image_data: Union[str, bytes],
+    bucket: str,
+    file_name: str,
+    splitting_type: str,
+    idx: int,
+    is_bytes: bool = False,
 ):
     """Upload image to S3 from either a file path or binary data.
 
@@ -237,7 +252,13 @@ def download_image_from_url(img_url: str) -> str:
 
 
 def process_single_image(
-    img_path: str, context: str, image_tag: str, bucket_name: str, file_name: str, idx: int
+    img_path: str,
+    context: str,
+    image_tag: str,
+    bucket_name: str,
+    file_name: str,
+    idx: int,
+    s3_link: str = None,
 ) -> str:
     """Process a single image and return its understanding text.
 
@@ -268,10 +289,12 @@ def process_single_image(
     understanding = figure_llm.figure_understand(image_base64, context, image_tag, s3_link=f"{idx:05d}.jpg")
 
     # Update S3 link
-    updated_s3_link = upload_image_to_s3(img_path, bucket_name, file_name, "image", idx)
-    understanding = understanding.replace(f"<link>{idx:05d}.jpg</link>", f"<link>{updated_s3_link}</link>")
+    if not s3_link:
+        s3_link = upload_image_to_s3(img_path, bucket_name, file_name, "image", idx)
+
+    understanding = understanding.replace(f"<link>{idx:05d}.jpg</link>", f"<link>{s3_link}</link>")
 
-    return understanding
+    return understanding, s3_link
 
 
 def process_markdown_images_with_llm(content: str, bucket_name: str, file_name: str) -> str:
@@ -296,6 +319,8 @@ def process_markdown_images_with_llm(content: str, bucket_name: str, file_name:
     image_pattern = r"!\[([^\]]*)\]\(([^)]+)\)"
     last_end = 0
     result = ""
+    # Add mapping to track image paths and their S3 objects
+    image_s3_mapping = {}
 
     for idx, match in enumerate(re.finditer(image_pattern, content), 1):
         start, end = match.span()
@@ -309,20 +334,40 @@ def process_markdown_images_with_llm(content: str, bucket_name: str, file_name:
             # Handle URL images
             if img_path.startswith(("http://", "https://")):
                 try:
-                    img_path = download_image_from_url(img_path)
+                    local_img_path = download_image_from_url(img_path)
                 except Exception as e:
                     logger.error(f"Error downloading image from URL {img_path}: {e}")
                     result += match.group(1)
                     last_end = end
                     continue
+            else:
+                logger.error(f"Image path {img_path} is not a URL")
+                result += match.group(1)
+                last_end = end
+                continue
 
             # Get context
             context_start = max(0, start - 200)
             context_end = min(len(content), end + 200)
             context = f"{content[context_start:start]}\n<image>\n{image_tag}\n</image>\n{content[end:context_end]}"
 
+            # Check if image was already processed
+            s3_link = image_s3_mapping.get(img_path)
+
             # Process the image
-            understanding = process_single_image(img_path, context, image_tag, bucket_name, file_name, idx)
+            understanding, updated_s3_link = process_single_image(
+                local_img_path,
+                context,
+                image_tag,
+                bucket_name,
+                file_name,
+                idx,
+                s3_link,
+            )
+
+            # If this is a new image path, store its S3 object name
+            if not s3_link and understanding:
+                image_s3_mapping[img_path] = updated_s3_link
 
             if understanding:
                 result += f"\n\n{understanding}\n\n"