Methods

Some models expose more than one entrypoint for visual understanding. One set of weights can read a page as Markdown, return a layout, or segment an object from a prompt. A chat API has no field to choose between them, so the Gateway adds one: method. It is a request kwarg that gives you direct access to each entrypoint.

Model	Methods	One model, different outputs
rednote-hilab/dots.mocr	`markdown`, `parse_layout`, `parse_layout_only`, `ocr`	Page text as Markdown, layout blocks with text, or layout regions without text.
lightonai/lightonocr-3-4b	`markdown`, `grounding`, `layout`	Page text as Markdown, every block with its label and box, or the boxes without text.
facebook/sam3.1	`segment`, `segment_box`, `track`	Masks from a text prompt, masks from a box, or masks tracked across video frames.
microsoft/florence-2-base-ft	`od`, `caption`, `dense_region_caption`, `ocr_with_region`, …	Boxes, captions, or text with regions.

Default: omit method and the model’s default_method applies, so a plain call still works.
Pick the entrypoint, not the model: to change the output, change method. Keep the model.
Output shape: the method fixes the payload type, for example doc.ocr.lines or img.detect.bboxes. See content object tags.
All kwargs: the full list of Gateway request fields is on Extra Kwargs.
Discovery: each model lists its methods on GET /v1/openai/models and on its catalog card.

response_format is independent. It selects how the result is serialized, and any method combines with any format.

The `method` field

Pass method (and optional method_params) at the top level of the request body, or via extra_body in the OpenAI Python SDK. Omit method and the model’s default_method applies, on an image and on every page of a document_url. Each model card lists its methods, its default, and what each one returns.

response = client.chat.completions.create(
    model="paddleocr/pp-ocrv6",
    messages=[{"role": "user", "content": [{"type": "image_url", "image_url": {"url": "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg"}}]}],
    extra_body={"method": "ocr"},
)

const response = await client.chat.completions.create({
  model: "paddleocr/pp-ocrv6",
  messages: [{ role: "user", content: [{ type: "image_url", image_url: { url: "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg" } }] }],
  method: "ocr",
});

vlmrun gw chat https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg \
  -m paddleocr/pp-ocrv6 --method ocr

curl https://gateway.vlm.run/v1/openai/chat/completions \
  -X POST \
  -H "Authorization: Bearer $VLMRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "paddleocr/pp-ocrv6",
    "method": "ocr",
    "messages": [{"role": "user", "content": [{"type": "image_url", "image_url": {"url": "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg"}}]}]
  }'

Live methods, default_method, and extra_body_help are on GET /v1/openai/models. A method a model does not advertise is a 400.

Method Parameters

method_params sits beside method, or in extra_body. An unknown key is a 400. Keys are on the catalog card.

response = client.chat.completions.create(
    model="paddleocr/pp-ocrv6",
    messages=[{"role": "user", "content": [{"type": "image_url", "image_url": {"url": "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg"}}]}],
    extra_body={
        "method": "ocr",
        "method_params": {"lang": "en", "score_threshold": 0.5},
    },
)

const response = await client.chat.completions.create({
  model: "paddleocr/pp-ocrv6",
  messages: [{ role: "user", content: [{ type: "image_url", image_url: { url: "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg" } }] }],
  method: "ocr",
  method_params: { lang: "en", score_threshold: 0.5 },
});

vlmrun gw chat https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg \
  -m paddleocr/pp-ocrv6 --method ocr \
  --method-params '{"lang": "en", "score_threshold": 0.5}'

curl https://gateway.vlm.run/v1/openai/chat/completions \
  -X POST \
  -H "Authorization: Bearer $VLMRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "paddleocr/pp-ocrv6",
    "method": "ocr",
    "method_params": {
      "lang": "en",
      "score_threshold": 0.5
    },
    "messages": [{"role": "user", "content": [{"type": "image_url", "image_url": {"url": "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.receipt/playground/2.jpg"}}]}]
  }'

video_fps is the cadence for a native video_url. facebook/sam3.1 track reads sampling from method_params.

Response Formats

JSON mode for programmatic use, text mode for agents and MCP.

Models

Models grouped by task, with their methods.

Visual Intelligence

Detection, segmentation, pose, and OCR samples.

Document OCR

End-to-end recipe from model selection to response parsing.

Overview

Platform

OpenAI Compatibility

TypeSafe Compatibility

MCP

Misc

The `method` field

Method Parameters

Response Formats

Models

Visual Intelligence

Document OCR

​The method field

​Method Parameters

​Related

Response Formats

Models

Visual Intelligence

Document OCR

The `method` field

Method Parameters

Related