Remove graphic elements

docs/docs/extraction/extraction-charts-infographics.md

@@ -1,6 +1,6 @@
 # Charts and infographics
 
-Charts and infographic regions are classified as graphic elements and processed with the corresponding NVIDIA NIM workflows (for example, **yolox-graphic-elements** in current releases). Outputs use the same metadata schema as other extracted objects.
+Chart and infographic regions are detected by page-element detection and processed by NeMo Retriever Library's OCR workflow. Outputs use the same metadata schema as other extracted objects.
 
 **Related**
 

nemo_retriever/README.md

@@ -151,7 +151,6 @@ python -m nemo_retriever.examples.graph_pipeline \
   /your-example-dir \
   --lancedb-uri lancedb \
   --page-elements-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3 \
-  --graphic-elements-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-graphic-elements-v1 \
   --ocr-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-ocr-v1 \
   --ocr-version v1 \
   --table-structure-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1 \
@@ -493,7 +492,6 @@ ingestor = (
   .extract(
     # for self hosted NIMs, your URLs will depend on your NIM container DNS settings
     page_elements_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3",
-    graphic_elements_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-graphic-elements-v1",
     ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-ocr-v1",
     table_structure_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1"
   )

nemo_retriever/pyproject.toml

@@ -82,7 +82,6 @@ local = [
   "timm==1.0.22",
   "albumentations==2.0.8",
   "nemotron-page-elements-v3>=0.dev0",
-  "nemotron-graphic-elements-v1>=0.dev0",
   "nemotron-table-structure-v1>=0.dev0",
   # Stay on the 1.0.2 OCR dev train and exclude older PyPI finals.
   "nemotron-ocr>=1.0.2.dev0,<1.0.2a0; sys_platform == 'linux'",
@@ -150,7 +149,6 @@ version = {attr = "nemo_retriever.version.get_build_version"}
 nv-ingest-api = { path = "../api/", editable = true }
 nv-ingest-client = { path = "../client/", editable = true }
 nemotron-page-elements-v3 = { index = "test-pypi" }
-nemotron-graphic-elements-v1 = { index = "test-pypi" }
 nemotron-table-structure-v1 = { index = "test-pypi" }
 nemotron-ocr = { index = "test-pypi" }
 # On Linux, resolve torch/torchvision from the CUDA wheel index.

nemo_retriever/src/nemo_retriever/application/pipeline/build_plan.py

@@ -52,7 +52,6 @@ def stage_names_from_flags(
     use_table_structure: bool = False,
     table_output_format: str = "pseudo_markdown",
     extract_charts: bool = False,
-    use_graphic_elements: bool = False,
     embed_text: bool = False,
 ) -> Iterable[str]:
     validate_table_structure_flags(use_table_structure, table_output_format)
@@ -62,9 +61,7 @@ def stage_names_from_flags(
         yield "enrich_table_structure"
     elif extract_tables:
         yield "enrich_table"
-    if extract_charts and use_graphic_elements:
-        yield "enrich_graphic_elements"
-    elif extract_charts:
+    if extract_charts:
         yield "enrich_chart"
     if embed_text:
         yield "embed_text"

nemo_retriever/src/nemo_retriever/application/pipeline/stage_registry.py

@@ -8,7 +8,6 @@
 
 import pandas as pd
 
-from nemo_retriever.chart.chart_detection import graphic_elements_ocr_page_elements
 from nemo_retriever.chart.processor import extract_chart_data_from_primitives_df
 from nemo_retriever.infographic.processor import extract_infographic_data_from_primitives_df
 from nemo_retriever.table.processor import extract_table_data_from_primitives_df
@@ -24,19 +23,12 @@ def _enrich_table_structure(df: pd.DataFrame, **kwargs: Any) -> tuple[pd.DataFra
     return out_df, {}
 
 
-def _enrich_graphic_elements(df: pd.DataFrame, **kwargs: Any) -> tuple[pd.DataFrame, Dict[str, Any]]:
-    """Thin wrapper so ``graphic_elements_ocr_page_elements`` matches the ``(df, info)`` contract."""
-    out_df = graphic_elements_ocr_page_elements(df, **kwargs)
-    return out_df, {}
-
-
 # Registry can be extended by future package integrations without changing
 # pipeline orchestration internals.
 STAGE_REGISTRY: Dict[str, StageHandler] = {
     "enrich_infographic": extract_infographic_data_from_primitives_df,
     "enrich_table": extract_table_data_from_primitives_df,
     "enrich_table_structure": _enrich_table_structure,
     "enrich_chart": extract_chart_data_from_primitives_df,
-    "enrich_graphic_elements": _enrich_graphic_elements,
     "embed_text": embed_text_from_primitives_df,
 }

nemo_retriever/src/nemo_retriever/chart/commands.py

@@ -72,66 +72,5 @@ def run(
     console.print(f"[green]Done[/green] wrote={output_path} rows={len(out_df)}")
 
 
-@app.command("graphic-elements")
-def render_graphic_elements(
-    config: Optional[Path] = typer.Option(
-        None,
-        "--config",
-        exists=True,
-        dir_okay=False,
-        file_okay=True,
-        help="Optional YAML config file. If set, values are loaded from YAML; "
-        "explicitly passed CLI flags override YAML.",
-    ),
-    input_dir: Optional[Path] = typer.Option(
-        None,
-        "--input-dir",
-        exists=True,
-        file_okay=False,
-        dir_okay=True,
-        help="Directory to scan recursively for *.pdf (can be provided via --config).",
-    ),
-    method: str = typer.Option("pdfium", "--method", help="PDF extraction method."),
-    auth_token: Optional[str] = typer.Option(None, "--auth-token", help="Auth token for NIM-backed services."),
-    yolox_grpc_endpoint: Optional[str] = typer.Option(None, "--yolox-grpc-endpoint"),
-    yolox_http_endpoint: Optional[str] = typer.Option(None, "--yolox-http-endpoint"),
-    nemotron_parse_grpc_endpoint: Optional[str] = typer.Option(None, "--nemotron-parse-grpc-endpoint"),
-    nemotron_parse_http_endpoint: Optional[str] = typer.Option(None, "--nemotron-parse-http-endpoint"),
-    nemotron_parse_model_name: Optional[str] = typer.Option(None, "--nemotron-parse-model-name"),
-    extract_text: bool = typer.Option(True, "--extract-text/--no-extract-text"),
-    extract_images: bool = typer.Option(False, "--extract-images/--no-extract-images"),
-    extract_tables: bool = typer.Option(False, "--extract-tables/--no-extract-tables"),
-    extract_charts: bool = typer.Option(False, "--extract-charts/--no-extract-charts"),
-    extract_infographics: bool = typer.Option(False, "--extract-infographics/--no-extract-infographics"),
-    extract_page_as_image: bool = typer.Option(False, "--extract-page-as-image/--no-extract-page-as-image"),
-    text_depth: str = typer.Option("page", "--text-depth"),
-    write_json_outputs: bool = typer.Option(True, "--write-json-outputs/--no-write-json-outputs"),
-    json_output_dir: Optional[Path] = typer.Option(None, "--json-output-dir", file_okay=False, dir_okay=True),
-    limit: Optional[int] = typer.Option(None, "--limit", help="Optionally limit number of PDFs processed."),
-) -> None:
-    _ = (
-        config,
-        input_dir,
-        method,
-        auth_token,
-        yolox_grpc_endpoint,
-        yolox_http_endpoint,
-        nemotron_parse_grpc_endpoint,
-        nemotron_parse_http_endpoint,
-        nemotron_parse_model_name,
-        extract_text,
-        extract_images,
-        extract_tables,
-        extract_charts,
-        extract_infographics,
-        extract_page_as_image,
-        text_depth,
-        write_json_outputs,
-        json_output_dir,
-        limit,
-    )
-    typer.echo("graphic-elements command is not implemented yet.")
-
-
 def main() -> None:
     app()

nemo_retriever/src/nemo_retriever/chart/config.py

@@ -29,21 +29,3 @@ def load_chart_extractor_schema_from_dict(cfg: Dict[str, Any]) -> ChartExtractor
 class ChartExtractionStageConfig:
     batch_size: int = 64
     stage_name: str = "chart_extraction"
-
-
-@dataclass(frozen=True)
-class GraphicElementsOCRStageConfig:
-    graphic_elements_invoke_url: str = ""
-    ocr_invoke_url: str = ""
-    api_key: str = ""
-    request_timeout_s: float = 60.0
-
-
-def load_graphic_elements_ocr_config_from_dict(cfg: Dict[str, Any]) -> GraphicElementsOCRStageConfig:
-    cfg = dict(cfg or {})
-    return GraphicElementsOCRStageConfig(
-        graphic_elements_invoke_url=str(cfg.get("graphic_elements_invoke_url") or ""),
-        ocr_invoke_url=str(cfg.get("ocr_invoke_url") or ""),
-        api_key=str(cfg.get("api_key") or ""),
-        request_timeout_s=float(cfg.get("request_timeout_s", 60.0)),
-    )