From 72dd815195909db67088eb268263c4b3294db726 Mon Sep 17 00:00:00 2001
From: Gabe Goodhart <ghart@us.ibm.com>
Date: Tue, 8 Apr 2025 13:29:50 -0600
Subject: [PATCH] feat: Connect "granite_vision_ollama" pipeline option to CLI

Branch: OllamaVlmModel

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
---
 docling/cli/main.py              |  3 +++
 docling/cli/models.py            |  1 +
 docling/pipeline/vlm_pipeline.py | 14 ++++++++++++--
 3 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/docling/cli/main.py b/docling/cli/main.py
index e0f0cbd8..71527b92 100644
--- a/docling/cli/main.py
+++ b/docling/cli/main.py
@@ -40,6 +40,7 @@ from docling.datamodel.pipeline_options import (
     VlmModelType,
     VlmPipelineOptions,
     granite_vision_vlm_conversion_options,
+    granite_vision_vlm_ollama_conversion_options,
     smoldocling_vlm_conversion_options,
     smoldocling_vlm_mlx_conversion_options,
 )
@@ -535,6 +536,8 @@ def convert(
 
             if vlm_model == VlmModelType.GRANITE_VISION:
                 pipeline_options.vlm_options = granite_vision_vlm_conversion_options
+            elif vlm_model == VlmModelType.GRANITE_VISION_OLLAMA:
+                pipeline_options.vlm_options = granite_vision_vlm_ollama_conversion_options
             elif vlm_model == VlmModelType.SMOLDOCLING:
                 pipeline_options.vlm_options = smoldocling_vlm_conversion_options
                 if sys.platform == "darwin":
diff --git a/docling/cli/models.py b/docling/cli/models.py
index 7bc313c1..5d38fe50 100644
--- a/docling/cli/models.py
+++ b/docling/cli/models.py
@@ -33,6 +33,7 @@ class _AvailableModels(str, Enum):
     PICTURE_CLASSIFIER = "picture_classifier"
     SMOLVLM = "smolvlm"
     GRANITE_VISION = "granite_vision"
+    GRANITE_VISION_OLLAMA = "granite_vision_ollama"
     EASYOCR = "easyocr"
 
 
diff --git a/docling/pipeline/vlm_pipeline.py b/docling/pipeline/vlm_pipeline.py
index d4defa89..3590d068 100644
--- a/docling/pipeline/vlm_pipeline.py
+++ b/docling/pipeline/vlm_pipeline.py
@@ -15,13 +15,16 @@ from docling.backend.pdf_backend import PdfDocumentBackend
 from docling.datamodel.base_models import InputFormat, Page
 from docling.datamodel.document import ConversionResult, InputDocument
 from docling.datamodel.pipeline_options import (
+    HuggingFaceVlmOptions,
     InferenceFramework,
+    OllamaVlmOptions,
     ResponseFormat,
     VlmPipelineOptions,
 )
 from docling.datamodel.settings import settings
 from docling.models.hf_mlx_model import HuggingFaceMlxModel
 from docling.models.hf_vlm_model import HuggingFaceVlmModel
+from docling.models.ollama_vlm_model import OllamaVlmModel
 from docling.pipeline.base_pipeline import PaginatedPipeline
 from docling.utils.profiling import ProfilingScope, TimeRecorder
 
@@ -57,7 +60,14 @@ class VlmPipeline(PaginatedPipeline):
 
         self.keep_images = self.pipeline_options.generate_page_images
 
-        if (
+        if isinstance(pipeline_options.vlm_options, OllamaVlmOptions):
+            self.build_pipe = [
+                OllamaVlmModel(
+                    enabled=True,  # must be always enabled for this pipeline to make sense.
+                    vlm_options=self.pipeline_options.vlm_options,
+                ),
+            ]
+        elif (
             self.pipeline_options.vlm_options.inference_framework
             == InferenceFramework.MLX
         ):
@@ -69,7 +79,7 @@ class VlmPipeline(PaginatedPipeline):
                     vlm_options=self.pipeline_options.vlm_options,
                 ),
             ]
-        else:
+        elif isinstance(pipeline_options.vlm_options, HuggingFaceVlmOptions):
             self.build_pipe = [
                 HuggingFaceVlmModel(
                     enabled=True,  # must be always enabled for this pipeline to make sense.