Move bbox.py from domain/ to infra/ and unify coordinate logic

domain/ must be pure with no external dependencies. bbox.py imports
docling_core and belongs in infra/. Also refactor ServeConverter to
use the canonical to_topleft_list via BoundingBox instead of
duplicated manual coordinate conversion. Move docling-core to base
requirements since it is now needed in both modes.

README.md

@@ -48,8 +48,8 @@ document-parser/
 ├── main.py                   # FastAPI app, CORS, lifespan
 ├── domain/                   # Pure domain — no HTTP, no DB
 │   ├── models.py             # Document, AnalysisJob dataclasses
-│   ├── parsing.py            # Docling conversion & page extraction
-│   └── bbox.py               # Bounding box coordinate normalization
+│   ├── ports.py              # Abstract protocols (converter, chunker)
+│   └── value_objects.py      # ConversionResult, PageDetail, ChunkResult
 ├── api/                      # HTTP layer (FastAPI routers)
 │   ├── schemas.py            # Pydantic DTOs (camelCase serialization)
 │   ├── documents.py          # /api/documents endpoints

docs/architecture.md

@@ -68,7 +68,7 @@ document-parser/
 
 | Layer | Role | Depends on |
 |-------|------|------------|
-| **domain** | Dataclasses, bbox math, Docling conversion | Nothing (pure Python) |
+| **domain** | Dataclasses, value objects, ports | Nothing (pure Python) |
 | **persistence** | SQLite CRUD, aiosqlite | domain (models) |
 | **services** | Orchestrate use cases, call Docling | domain + persistence |
 | **api** | HTTP endpoints, Pydantic DTOs, error handling | services |

docs/bbox-pipeline.md

@@ -38,7 +38,7 @@ The frontend converts PDF points to CSS pixels, then the canvas renders at `devi
 
 ## Transformation 1 — `to_topleft_list()`
 
-**File:** `document-parser/domain/bbox.py`
+**File:** `document-parser/infra/bbox.py`
 
 Normalizes any Docling bbox to `[left, top, right, bottom]` in TOPLEFT coordinates.
 

document-parser/infra/local_chunker.py

@@ -15,7 +15,7 @@ from docling_core.transforms.chunker import HierarchicalChunker
 from docling_core.transforms.chunker.hybrid_chunker import HybridChunker
 from docling_core.types.doc.document import DoclingDocument
 
-from domain.bbox import EMPTY_BBOX, to_topleft_list
+from infra.bbox import EMPTY_BBOX, to_topleft_list
 from domain.value_objects import ChunkBbox, ChunkingOptions, ChunkResult
 
 logger = logging.getLogger(__name__)

document-parser/infra/local_converter.py

@@ -35,7 +35,7 @@ from docling_core.types.doc import (
     TitleItem,
 )
 
-from domain.bbox import to_topleft_list
+from infra.bbox import to_topleft_list
 from domain.value_objects import (
     ConversionOptions,
     ConversionResult,

document-parser/infra/serve_converter.py

@@ -18,6 +18,7 @@ import mimetypes
 from pathlib import Path
 
 import httpx
+from docling_core.types.doc.base import BoundingBox, CoordOrigin
 
 from domain.value_objects import (
     ConversionOptions,
@@ -25,6 +26,7 @@ from domain.value_objects import (
     PageDetail,
     PageElement,
 )
+from infra.bbox import to_topleft_list
 
 logger = logging.getLogger(__name__)
 
@@ -222,21 +224,22 @@ def _add_element(item: dict, pages: dict[int, PageDetail]) -> None:
 
 
 def _extract_bbox(bbox_data: dict, page_height: float) -> list[float]:
-    """Extract and normalize bbox to TOPLEFT [l, t, r, b] format."""
+    """Extract and normalize bbox to TOPLEFT [l, t, r, b] format.
+
+    Delegates to the canonical to_topleft_list function via a docling-core
+    BoundingBox, ensuring consistent coordinate handling across all converters.
+    """
     if not isinstance(bbox_data, dict):
         return [0.0, 0.0, 0.0, 0.0]
 
-    left = bbox_data.get("l", 0.0)
-    top = bbox_data.get("t", 0.0)
-    right = bbox_data.get("r", 0.0)
-    bottom = bbox_data.get("b", 0.0)
-    coord_origin = bbox_data.get("coord_origin", "TOPLEFT")
+    origin_str = bbox_data.get("coord_origin", "TOPLEFT")
+    origin = CoordOrigin.BOTTOMLEFT if origin_str == "BOTTOMLEFT" else CoordOrigin.TOPLEFT
 
-    if coord_origin == "BOTTOMLEFT":
-        # In BOTTOMLEFT: top has higher y, bottom has lower y
-        # In TOPLEFT: flip both — new_top = page_height - old_top
-        new_top = page_height - top
-        new_bottom = page_height - bottom
-        top, bottom = new_top, new_bottom
-
-    return [left, top, right, bottom]
+    bbox = BoundingBox(
+        l=bbox_data.get("l", 0.0),
+        t=bbox_data.get("t", 0.0),
+        r=bbox_data.get("r", 0.0),
+        b=bbox_data.get("b", 0.0),
+        coord_origin=origin,
+    )
+    return to_topleft_list(bbox, page_height)

document-parser/requirements-local.txt

@@ -1,3 +1,2 @@
 -r requirements.txt
 docling>=2.80.0,<3.0.0
-docling-core>=2.0.0,<3.0.0

document-parser/requirements.txt

@@ -1,3 +1,4 @@
+docling-core>=2.0.0,<3.0.0
 fastapi>=0.115.0,<1.0.0
 uvicorn[standard]>=0.32.0,<1.0.0
 python-multipart>=0.0.12

document-parser/tests/test_bbox.py

@@ -8,7 +8,7 @@ misaligned overlays in the UI.
 import pytest
 from docling_core.types.doc.base import BoundingBox, CoordOrigin
 
-from domain.bbox import EMPTY_BBOX, to_topleft_list
+from infra.bbox import EMPTY_BBOX, to_topleft_list
 
 # ---------------------------------------------------------------------------
 # Standard conversions