Usage

The API provides two endpoints: one for urls, one for files. This is necessary to send files directly in binary format instead of base64-encoded strings.

Common parameters

On top of the source of file (see below), both endpoints support the same parameters.

ConvertDocumentsOptions

CodeFormulaVlmOptions

VlmModelSpec

BaseVlmEngineOptions

PictureDescriptionVlmEngineOptions

VlmConvertOptions

VlmModelApi

VlmModelLocal

PictureDescriptionApi

PictureDescriptionLocal

Authentication

When authentication is activated (see the parameter DOCLING_SERVE_API_KEY in configuration.md), all the API requests must provide the header X-Api-Key with the correct secret key.

Convert endpoints

Source endpoint

The endpoint is /v1/convert/source, listening for POST requests of JSON payloads.

On top of the above parameters, you must send the URL(s) of the document you want process with either the http_sources or file_sources fields. The first is fetching URL(s) (optionally using with extra headers), the second allows to provide documents as base64-encoded strings. No options is required, they can be partially or completely omitted.

Simple payload example:

{
  "http_sources": [{"url": "https://arxiv.org/pdf/2206.01062"}]
}

{
  "options": {
    "from_formats": ["docx", "pptx", "html", "image", "pdf", "asciidoc", "md", "xlsx"],
    "to_formats": ["md", "json", "html", "text", "doctags"],
    "image_export_mode": "placeholder",
    "do_ocr": true,
    "force_ocr": false,
    "ocr_engine": "easyocr",
    "ocr_lang": ["en"],
    "pdf_backend": "dlparse_v2",
    "table_mode": "fast",
    "abort_on_error": false,
  },
  "http_sources": [{"url": "https://arxiv.org/pdf/2206.01062"}]
}

curl -X 'POST' \
  'http://localhost:5001/v1/convert/source' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "options": {
    "from_formats": [
      "docx",
      "pptx",
      "html",
      "image",
      "pdf",
      "asciidoc",
      "md",
      "xlsx"
    ],
    "to_formats": ["md", "json", "html", "text", "doctags"],
    "image_export_mode": "placeholder",
    "do_ocr": true,
    "force_ocr": false,
    "ocr_engine": "easyocr",
    "ocr_lang": [
      "fr",
      "de",
      "es",
      "en"
    ],
    "pdf_backend": "dlparse_v2",
    "table_mode": "fast",
    "abort_on_error": false,
    "do_table_structure": true,
    "include_images": true,
    "images_scale": 2
  },
  "http_sources": [{"url": "https://arxiv.org/pdf/2206.01062"}]
}'

import httpx

async_client = httpx.AsyncClient(timeout=60.0)
url = "http://localhost:5001/v1/convert/source"
payload = {
  "options": {
    "from_formats": ["docx", "pptx", "html", "image", "pdf", "asciidoc", "md", "xlsx"],
    "to_formats": ["md", "json", "html", "text", "doctags"],
    "image_export_mode": "placeholder",
    "do_ocr": True,
    "force_ocr": False,
    "ocr_engine": "easyocr",
    "ocr_lang": "en",
    "pdf_backend": "dlparse_v2",
    "table_mode": "fast",
    "abort_on_error": False,
  },
  "http_sources": [{"url": "https://arxiv.org/pdf/2206.01062"}]
}

response = await async_client_client.post(url, json=payload)

data = response.json()

File as base64

The file_sources argument in the endpoint allows to send files as base64-encoded strings. When your PDF or other file type is too large, encoding it and passing it inline to curl can lead to an “Argument list too long” error on some systems. To avoid this, we write the JSON request body to a file and have curl read from that file.

# 1. Base64-encode the file
B64_DATA=$(base64 -w 0 /path/to/file/pdf-to-convert.pdf)

# 2. Build the JSON with your options
cat <<EOF > /tmp/request_body.json
{
  "options": {
  },
  "file_sources": [{
    "base64_string": "${B64_DATA}",
    "filename": "pdf-to-convert.pdf"
  }]
}
EOF

# 3. POST the request to the docling service
curl -X POST "localhost:5001/v1/convert/source" \
     -H "Content-Type: application/json" \
     -d @/tmp/request_body.json

File endpoint

The endpoint is: /v1/convert/file, listening for POST requests of Form payloads (necessary as the files are sent as multipart/form data). You can send one or multiple files.

curl -X 'POST' \
  'http://127.0.0.1:5001/v1/convert/file' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'ocr_engine=easyocr' \
  -F 'pdf_backend=dlparse_v2' \
  -F 'from_formats=pdf' \
  -F 'from_formats=docx' \
  -F 'force_ocr=false' \
  -F 'image_export_mode=embedded' \
  -F 'ocr_lang=en' \
  -F 'ocr_lang=pl' \
  -F 'table_mode=fast' \
  -F 'files=@2206.01062v1.pdf;type=application/pdf' \
  -F 'abort_on_error=false' \
  -F 'to_formats=md' \
  -F 'to_formats=text' \
  -F 'do_ocr=true'

import httpx

async_client = httpx.AsyncClient(timeout=60.0)
url = "http://localhost:5001/v1/convert/file"
parameters = {
"from_formats": ["docx", "pptx", "html", "image", "pdf", "asciidoc", "md", "xlsx"],
"to_formats": ["md", "json", "html", "text", "doctags"],
"image_export_mode": "placeholder",
"do_ocr": True,
"force_ocr": False,
"ocr_engine": "easyocr",
"ocr_lang": ["en"],
"pdf_backend": "dlparse_v2",
"table_mode": "fast",
"abort_on_error": False,
}

current_dir = os.path.dirname(__file__)
file_path = os.path.join(current_dir, '2206.01062v1.pdf')

files = {
    'files': ('2206.01062v1.pdf', open(file_path, 'rb'), 'application/pdf'),
}

response = await async_client.post(url, files=files, data=parameters)
assert response.status_code == 200, "Response should be 200 OK"

data = response.json()

Picture description options

When the picture description enrichment is activated, users may specify which model and which execution mode to use for this task. There are two choices for the execution mode: local will run the vision-language model directly, api will invoke an external API endpoint.

The local option is specified with:

{
  "picture_description_local": {
    "repo_id": "",  // Repository id from the Hugging Face Hub.
    "generation_config": {"max_new_tokens": 200, "do_sample": false},  // HF generation config.
    "prompt": "Describe this image in a few sentences. ",  // Prompt used when calling the vision-language model.
  }
}

The possible values for generation_config are documented in the Hugging Face text generation docs.

The api option is specified with:

{
  "picture_description_api": {
    "url": "",  // Endpoint which accepts openai-api compatible requests.
    "headers": {},  // Headers used for calling the API endpoint. For example, it could include authentication headers.
    "params": {},  // Model parameters.
    "timeout": 20,  // Timeout for the API request.
    "prompt": "Describe this image in a few sentences. ",  // Prompt used when calling the vision-language model.
  }
}

Example URLs are:

• http://localhost:8000/v1/chat/completions for the local vllm api, with example picture_description_api:

  • the HuggingFaceTB/SmolVLM-256M-Instruct model

    {
      "url": "http://localhost:8000/v1/chat/completions",
      "params": {
        "model": "HuggingFaceTB/SmolVLM-256M-Instruct",
        "max_completion_tokens": 200,
      }
    }
    

  • the ibm-granite/granite-vision-3.2-2b model

    {
      "url": "http://localhost:8000/v1/chat/completions",
      "params": {
        "model": "ibm-granite/granite-vision-3.2-2b",
        "max_completion_tokens": 200,
      }
    }
    

• http://localhost:11434/v1/chat/completions for the local Ollama api, with example picture_description_api:

  • the granite3.2-vision:2b model

    {
      "url": "http://localhost:11434/v1/chat/completions",
      "params": {
        "model": "granite3.2-vision:2b"
      }
    }
    

Note that when using picture_description_api, the server must be launched with DOCLING_SERVE_ENABLE_REMOTE_SERVICES=true.

Response format

The response can be a JSON Document or a File.

• If you process only one file, the response will be a JSON document with the following format:

  {
    "document": {
      "md_content": "",
      "json_content": {},
      "html_content": "",
      "text_content": "",
      "doctags_content": ""
      },
    "status": "<success|partial_success|skipped|failure>",
    "processing_time": 0.0,
    "timings": {},
    "errors": []
  }
  

  Depending on the value you set in output_formats, the different items will be populated with their respective results or empty.

  processing_time is the Docling processing time in seconds, and timings (when enabled in the backend) provides the detailed timing of all the internal Docling components.

• If you set the parameter target to the zip mode, the response will be a zip file.

• If multiple files are generated (multiple inputs, or one input but multiple outputs with the zip target mode), the response will be a zip file.

Asynchronous API

Both /v1/convert/source and /v1/convert/file endpoints are available as asynchronous variants. The advantage of the asynchronous endpoints is the possible to interrupt the connection, check for the progress update and fetch the result. This approach is more resilient against network instabilities and allows the client application logic to easily interleave conversion with other tasks.

Launch an asynchronous conversion with:

• POST /v1/convert/source/async when providing the input as sources.
• POST /v1/convert/file/async when providing the input as multipart-form files.

The response format is a task detail:

{
  "task_id": "<task_id>",  // the task_id which can be used for the next operations
  "task_status": "pending|started|success|failure",  // the task status
  "task_position": 1,  // the position in the queue
  "task_meta": null,  // metadata e.g. how many documents are in the total job and how many have been converted
}

Polling status

For checking the progress of the conversion task and wait for its completion, use the endpoint:

• GET /v1/status/poll/{task_id}

import time
import httpx

# ...
# response from the async task submission
task = response.json()

while task["task_status"] not in ("success", "failure"):
    response = httpx.get(f"{base_url}/status/poll/{task['task_id']}")
    task = response.json()

    time.sleep(5)

{
  "message": "connection|update|error",  // type of message being sent
  "task": {},  // the same content of the task description
  "error": "",  // description of the error
}

from websockets.sync.client import connect

uri = f"ws://{base_url}/v1/status/ws/{task['task_id']}"
with connect(uri) as websocket:
    for message in websocket:
        try:
            payload = json.loads(message)
            if payload["message"] == "error":
                break
            if payload["message"] == "update" and payload["task"]["task_status"] in ("success", "failure"):
                break
        except:
          break