From 8a6ad9d23269e6355c7a15e19703b74dfa01316e Mon Sep 17 00:00:00 2001
From: Mahesh Rajamani <99678631+maheshrajamani@users.noreply.github.com>
Date: Mon, 9 Jun 2025 10:34:46 -0400
Subject: [PATCH] feat: NV-ingest high resolution extraction and infographics
 content extraction (#8315)

* Added support for high resolution and extracting infographics content

* [autofix.ci] apply automated fixes

* Added validation for high resolution. Made image filter option value as false by default.

* [autofix.ci] apply automated fixes

* Fixed as per linters suggestion

* [autofix.ci] apply automated fixes

* Updated the file based on review for improved description

* Update src/backend/base/langflow/components/nvidia/nvidia_ingest.py

Accepted code change suggestion

Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>

* Addressed suggestion from coderabbitai

* [autofix.ci] apply automated fixes

* Addressed suggestion from Ruff

* [autofix.ci] apply automated fixes

---------

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Co-authored-by: Jordan Frazier <122494242+jordanrfrazier@users.noreply.github.com>
---
 .../components/nvidia/nvidia_ingest.py        | 31 ++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)

diff --git a/src/backend/base/langflow/components/nvidia/nvidia_ingest.py b/src/backend/base/langflow/components/nvidia/nvidia_ingest.py
index a29ac1324..7ee8f2227 100644
--- a/src/backend/base/langflow/components/nvidia/nvidia_ingest.py
+++ b/src/backend/base/langflow/components/nvidia/nvidia_ingest.py
@@ -1,5 +1,7 @@
 from urllib.parse import urlparse
 
+from pypdf import PdfReader
+
 from langflow.base.data import BaseFileComponent
 from langflow.io import (
     BoolInput,
@@ -67,6 +69,13 @@ class NvidiaIngestComponent(BaseFileComponent):
             info="Extract images from document",
             value=True,
         ),
+        BoolInput(
+            name="extract_infographics",
+            display_name="Extract Infographics",
+            info="Extract infographics from document",
+            value=False,
+            advanced=True,
+        ),
         DropdownInput(
             name="text_depth",
             display_name="Text Depth",
@@ -104,7 +113,7 @@ class NvidiaIngestComponent(BaseFileComponent):
             display_name="Filter Images",
             info="Filter images (see advanced options for filtering criteria).",
             advanced=True,
-            value=True,
+            value=False,
         ),
         IntInput(
             name="min_image_size",
@@ -141,6 +150,13 @@ class NvidiaIngestComponent(BaseFileComponent):
             advanced=True,
             value=True,
         ),
+        BoolInput(
+            name="high_resolution",
+            display_name="High Resolution (PDF only)",
+            info=("Process pdf in high-resolution mode for better quality extraction from scanned pdf."),
+            advanced=True,
+            value=False,
+        ),
     ]
 
     outputs = [
@@ -162,6 +178,17 @@ class NvidiaIngestComponent(BaseFileComponent):
             self.log(err_msg)
             raise ValueError(err_msg)
 
+        # Check if all files are PDFs when high resolution mode is enabled
+        if self.high_resolution:
+            for file in file_list:
+                try:
+                    with file.path.open("rb") as f:
+                        PdfReader(f)
+                except Exception as exc:
+                    error_msg = "High-resolution mode only supports valid PDF files."
+                    self.log(error_msg)
+                    raise ValueError(error_msg) from exc
+
         file_paths = [str(file.path) for file in file_list]
 
         self.base_url: str | None = self.base_url.strip() if self.base_url else None
@@ -196,7 +223,9 @@ class NvidiaIngestComponent(BaseFileComponent):
                     extract_tables=self.extract_tables,
                     extract_charts=self.extract_charts,
                     extract_images=self.extract_images,
+                    extract_infographics=self.extract_infographics,
                     text_depth=self.text_depth,
+                    **({"extract_method": "nemoretriever_parse"} if self.high_resolution else {}),
                 )
             )