Add support for kreuzberg

.env.sample

@@ -4,6 +4,7 @@ LOG_LEVEL=INFO
 
 HOST_DATA_FOLDER=
 CONTAINER_DATA_FOLDER="/data"
+DATA_FOLDER=
 
 http_proxy=
 https_proxy=
@@ -16,6 +17,7 @@ MARKER_API_PORT=
 SPARROW_API_PORT=8001 # hard-coded in sparrow API
 DOCLING_API_PORT=
 PADDLEOCR_API_PORT=
+KREUZBERG_API_PORT=
 
 # Source PostgreSQL instance
 POSTGRES_SOURCE_USER=postgres

docker-compose.yml

@@ -176,6 +176,40 @@ services:
       retries: 3
       start_period: 30s
 
+  kreuzberg:
+    profiles: [kreuzberg]
+    build:
+      context: packages/ocr/kreuzberg
+      dockerfile: Dockerfile
+      args:
+        <<: *build-args-common
+        KREUZBERG_API_PORT: ${KREUZBERG_API_PORT}
+    environment:
+      <<: [*proxy-common, *common-env]
+      DATA_FOLDER: /data
+      KREUZBERG_API_PORT: ${KREUZBERG_API_PORT}
+    env_file:
+      - ./.env
+    ports:
+      - "${KREUZBERG_API_PORT}:${KREUZBERG_API_PORT}"
+    volumes:
+      - ${PWD}/${DATA_FOLDER}:/data
+    networks:
+      - pyonb_ocr_api
+    healthcheck:
+      test:
+        [
+          "CMD",
+          "curl",
+          "-X",
+          "GET",
+          "http://localhost:${KREUZBERG_API_PORT}/health",
+        ]
+      interval: 10s
+      timeout: 3s
+      retries: 3
+      start_period: 30s
+
   ocr-forwarding-api:
     build:
       context: src/api

packages/ocr/kreuzberg/Dockerfile

@@ -0,0 +1,30 @@
+FROM ghcr.io/astral-sh/uv:python3.13-bookworm AS app
+
+SHELL ["/bin/bash", "-o", "pipefail", "-e", "-u", "-x", "-c"]
+
+WORKDIR /app
+ENV PYTHONDONTWRITEBYTECODE=1
+ENV PYTHONUNBUFFERED=1
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    libgl1 \
+    libglib2.0-0 \
+    libgomp1 \
+    libsm6 \
+    libxext6 \
+    libxrender-dev \
+    pandoc \
+    tesseract-ocr \
+    tesseract-ocr-eng \
+    && apt-get clean && rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
+
+COPY ./pyproject.toml ./README.md .
+COPY ./src src/
+
+RUN uv venv
+RUN --mount=type=cache,target=/root/.cache/uv,sharing=locked uv sync --no-editable --no-dev
+
+# make uvicorn etc available
+ENV PATH="/app/.venv/bin:$PATH"
+
+CMD uvicorn pyonb_kreuzberg.api:app --host 0.0.0.0 --port "$KREUZBERG_API_PORT" --workers 4 --reload --use-colors

packages/ocr/kreuzberg/README.md

@@ -0,0 +1,52 @@
+# Instructions
+
+Before using the `kreuzberg` API for OCR, you will need to set the `KREUZBERG_API_PORT`
+environment variable in the top-level `.env` file.
+
+## Python
+
+First install the `kreuzberg` API. From the top-level `pyonb` directory:
+
+```shell
+uv sync -extra kreuzberg
+```
+
+Then start the `kreuzberg` API:
+
+```shell
+python src/pyonb_kreuzberg/api.py
+```
+
+You can then use `curl` to send a PDF to the API:
+
+```shell
+curl -v -X POST http://127.0.0.1:8116/extract \
+  -F "file_upload=@document.pdf" \
+  -H "accept: application/json"
+```
+
+Note, this assumes you have set `KREUZBERG_API_PORT=8116`.
+
+Currently, this returns the response from the
+[`kreuzberg` API](https://kreuzberg.dev/user-guide/api-server/#extract-files)
+directly, rather than the standard `pyonb` response.
+
+## Docker Compose
+
+You will need to define the `OCR_FORWARDING_API_PORT` in the `.env` file.
+
+Then, spin up the `ocr-forwarding-api` and `kreuzberg` services:
+
+```shell
+docker-compose --profile kreuzberg up --build --detach
+```
+
+You can then use `curl` to send a PDF to the forwarding API:
+
+```shell
+curl -v -X POST http://127.0.0.1:8110/kreuzberg-ocr/inference_single \
+  -F "file_upload=@document.pdf" \
+  -H "accept: application/json"
+```
+
+Note, this assumes you have set `OCR_FORWARDING_API_PORT` to `8110`.

packages/ocr/kreuzberg/pyproject.toml

@@ -0,0 +1,14 @@
+[build-system]
+build-backend = "hatchling.build"
+requires = ["hatchling"]
+
+[project]
+dependencies = [
+    "kreuzberg[api]==3.13.3",
+    "uvicorn",
+]
+description = "pyonb wrapper around kreuzberg"
+name = "pyonb-kreuzberg"
+readme = "README.md"
+requires-python = ">=3.11"
+version = "0.1.0"

packages/ocr/kreuzberg/src/pyonb_kreuzberg/__init__.py

@@ -0,0 +1 @@
+"""Package for converting PDFs to structured text using Kreuzberg OCR."""

packages/ocr/kreuzberg/src/pyonb_kreuzberg/api.py

@@ -0,0 +1,40 @@
+"""API for Kreuzberg OCR."""
+
+import os
+
+import uvicorn
+from kreuzberg._api.main import (
+    KreuzbergError,
+    Litestar,
+    OpenTelemetryConfig,
+    OpenTelemetryPlugin,
+    StructLoggingConfig,
+    exception_handler,
+    general_exception_handler,
+    get_configuration,
+    handle_files_upload,
+    health_check,
+)
+
+KREUZBERG_API_PORT = int(os.getenv("KREUZBERG_API_PORT", default="8116"))
+
+app = Litestar(
+    route_handlers=[handle_files_upload, health_check, get_configuration],
+    request_max_body_size=100_000_000,
+    plugins=[OpenTelemetryPlugin(OpenTelemetryConfig())],
+    logging_config=StructLoggingConfig(),
+    exception_handlers={
+        KreuzbergError: exception_handler,
+        Exception: general_exception_handler,
+    },
+)
+
+if __name__ == "__main__":
+    uvicorn.run(
+        app,
+        host="127.0.0.1",
+        port=KREUZBERG_API_PORT,
+        workers=4,
+        reload=True,
+        use_colors=True,
+    )

pyproject.toml

@@ -49,6 +49,8 @@ optional-dependencies = {dev = [
     "mkdocs-material",
     "mkdocstrings",
     "mkdocstrings-python",
+], kreuzberg = [
+    "pyonb-kreuzberg",
 ], test = [
     "pytest",
     "pytest-cov",
@@ -149,3 +151,11 @@ env.docs = {commands = [
 gh.python."3.11" = ["py311"]
 gh.python."3.12" = ["py312"]
 gh.python."3.13" = ["py313"]
+
+[tool.uv.sources]
+pyonb-kreuzberg = {workspace = true}
+
+[tool.uv.workspace]
+members = [
+    "packages/ocr/*",
+]

src/api/app/main.py

@@ -6,10 +6,11 @@
 from fastapi import FastAPI, status
 from fastapi.responses import JSONResponse, RedirectResponse
 
-from .routers import docling, marker, paddleocr, sparrow
+from .routers import docling, kreuzberg, marker, paddleocr, sparrow
 
+_today = datetime.datetime.now(datetime.UTC).strftime("%Y_%m_%d")  # type: ignore[attr-defined] # mypy complains that 'Module has no attribute "UTC"'
 logging.basicConfig(
-    filename="pyonb-" + datetime.datetime.now(datetime.UTC).strftime("%Y_%m_%d") + ".log",
+    filename=f"pyonb-{_today}.log",
     format="%(asctime)s %(message)s",
     filemode="a",
 )
@@ -23,6 +24,7 @@
 app.include_router(marker.router)
 app.include_router(paddleocr.router)
 app.include_router(docling.router)
+app.include_router(kreuzberg.router)
 
 
 @app.get("/", include_in_schema=False)

src/api/app/routers/kreuzberg.py

@@ -0,0 +1,84 @@
+"""Routers for Kreuzberg OCR."""
+
+import logging
+import os
+import time
+from typing import Annotated, Any
+
+import aiohttp
+from fastapi import APIRouter, File, UploadFile, status
+from fastapi.responses import JSONResponse
+
+# Creating an object
+logger = logging.getLogger()
+
+router = APIRouter()
+
+KREUZBERG_API_PORT = os.getenv("KREUZBERG_API_PORT")
+
+
+@router.get("/kreuzberg/health")
+async def healthcheck() -> dict[str, Any]:
+    """Test aliveness endpoint for Kreuzberg."""
+    logger.info("[GET] /kreuzberg/health")
+    url = f"http://kreuzberg:{KREUZBERG_API_PORT}/health"
+
+    try:
+        async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=60 * 60)) as session:  # noqa: SIM117
+            async with session.get(url) as response:
+                response.raise_for_status()
+    except aiohttp.ClientError:
+        logger.exception("Failed to connect to kreuzberg service")
+        raise
+
+    return JSONResponse(
+        status_code=status.HTTP_200_OK,
+        content={"service": "kreuzberg", "status": "healthy"},
+    )
+
+
+@router.post("/kreuzberg-ocr/inference_single", status_code=status.HTTP_200_OK)
+async def inference_single_doc(file_upload: Annotated[UploadFile, File()] = None) -> JSONResponse:
+    """
+    Runs Kreuzberg OCR inference on a single document.
+
+    UploadFile object forwarded onto inference API.
+    """
+    logger.info("[POST] /kreuzberg-ocr/extract")
+    url = f"http://kreuzberg:{KREUZBERG_API_PORT}/extract"  # fwd request to kreuzberg service
+
+    data = aiohttp.FormData()
+    data.add_field(
+        "data",  # field name expected by Kreuzberg's /extract API
+        file_upload.file,
+        filename=file_upload.filename,
+        content_type=file_upload.content_type,
+    )
+    headers = {"accept": "application/json"}
+
+    logger.info("post request - url: %s", url)
+    logger.info("post request - data: %s", data)
+    logger.info("post request - headers: %s", headers)
+
+    t1 = time.perf_counter()
+    try:
+        async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=60 * 60)) as session:  # noqa: SIM117
+            async with session.post(url, data=data, headers=headers) as response:
+                response.raise_for_status()
+                ocr_results = await response.json()
+    except aiohttp.ClientError:
+        logger.exception("Request Exception")
+        raise
+    t2 = time.perf_counter()
+
+    # Kreuzberg's /extract API expects a list of documents and always returns a list of extracted text
+    # We only ever extract and return content for a single document
+    ocr_result = ocr_results[0]["content"]
+
+    response_json = {
+        "filename": str(file_upload.filename),
+        "duration_in_second": t2 - t1,
+        "ocr-result": ocr_result,
+    }
+
+    return JSONResponse(status_code=status.HTTP_200_OK, content=response_json)

src/api/requirements.txt

@@ -1,3 +1,4 @@
 fastapi[standard]
 uvicorn
 requests
+aiohttp

tests/.env.tests

@@ -4,6 +4,7 @@ LOG_LEVEL=INFO
 
 HOST_DATA_FOLDER="/absolute/path/to/tests/data/single_synthetic_doc"
 CONTAINER_DATA_FOLDER="/data"
+DATA_FOLDER="tests/data/single_synthetic_doc"
 
 http_proxy=
 https_proxy=
@@ -16,6 +17,7 @@ MARKER_API_PORT=8112
 SPARROW_API_PORT=8001 # hard-coded in sparrow API
 PADDLEOCR_API_PORT=8114
 DOCLING_API_PORT=8115
+KREUZBERG_API_PORT=8116
 
 # Source PostgreSQL instance
 POSTGRES_SOURCE_USER=postgres

tests/api/test_app.py

@@ -17,7 +17,14 @@ def test_local_start_api_and_healthy(ocr_forwarding_api_port: str) -> None:
 
 @pytest.mark.parametrize(
     "check_container_healthy",
-    ["pyonb-ocr-forwarding-api-1", "pyonb-marker-1", "pyonb-sparrow-1", "pyonb-paddleocr-1", "pyonb-docling-1"],
+    [
+        "pyonb-ocr-forwarding-api-1",
+        "pyonb-marker-1",
+        "pyonb-sparrow-1",
+        "pyonb-paddleocr-1",
+        "pyonb-docling-1",
+        "pyonb-kreuzberg-1",
+    ],
     indirect=True,
 )
 def test_check_services(check_container_healthy: bool) -> None:

tests/api/test_routers.py

@@ -41,6 +41,13 @@ def test_router_paddleocr(ocr_forwarding_api_port: str) -> None:
     assert response.json() == {"service": "paddleocr", "status": "healthy"}
 
 
+def test_router_kreuzberg(ocr_forwarding_api_port: str) -> None:
+    """Test healthcheck for kreuzberg."""
+    response = requests.get(f"http://127.0.0.1:{ocr_forwarding_api_port}/kreuzberg/health", timeout=5)
+    assert response.status_code == requests.codes.ok
+    assert response.json() == {"service": "kreuzberg", "status": "healthy"}
+
+
 def test_inference_single_file_upload_marker(ocr_forwarding_api_port: str, single_pdf_filepath: Path) -> None:
     """Test PDF conversion using marker with single file endpoint."""
     url = f"http://127.0.0.1:{ocr_forwarding_api_port}/marker/inference_single"
@@ -104,6 +111,21 @@ def test_inference_single_file_upload_paddleocr(ocr_forwarding_api_port: str, si
     assert response.json()["filename"] in single_pdf_filename
 
 
+def test_inference_single_file_upload_kreuzberg(ocr_forwarding_api_port: str, single_pdf_filepath: Path) -> None:
+    """Test PDF conversion using kreuzberg with single file endpoint."""
+    url = f"http://127.0.0.1:{ocr_forwarding_api_port}/kreuzberg-ocr/inference_single"
+
+    single_pdf_filename = single_pdf_filepath.name
+
+    with Path.open(single_pdf_filepath, "rb") as f:
+        files = {"file_upload": (single_pdf_filename, f, "application/pdf")}
+        response = requests.post(url, files=files, timeout=60 * 60)
+
+    assert response.status_code == requests.codes.ok
+    assert response.json()["duration_in_second"] >= 0
+    assert response.json()["filename"] == single_pdf_filename
+
+
 def test_inference_on_folder_marker(ocr_forwarding_api_port: str) -> None:
     """
     Test PDF conversion using marker pointed at a folder of files.