Copy

Para casos de uso e orientações de primeiros passos, consulte Extração de texto (Qwen-OCR) .

API compatível com OpenAI

Endpoints

Região	base_url do SDK	Endpoint HTTP
Singapura	`https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1`	`POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions`
EUA (Virgínia)	`https://dashscope-us.aliyuncs.com/compatible-mode/v1`	`POST https://dashscope-us.aliyuncs.com/compatible-mode/v1/chat/completions`
China (Pequim)	`https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1`	`POST https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1/chat/completions`

O Model Studio lançou domínios específicos por workspace para as regiões China (Pequim) e Singapura. Os novos domínios dedicados oferecem desempenho superior e maior estabilidade para solicitações de inferência. Recomendamos a migração para os novos domínios:

China (Pequim): de https://dashscope.aliyuncs.com para https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com
Singapura: de https://dashscope-intl.aliyuncs.com para https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com

{WorkspaceId} é o ID do seu workspace, disponível na página Workspace Details no console do Model Studio. O domínio existente permanece totalmente funcional.

Pré-requisitos

Obtenha uma chave de API e defina-a como variável de ambiente. Caso utilize o SDK da OpenAI, instale o SDK.

Início rápido

Utilize o endpoint de conclusões de chat compatível com OpenAI. Envie uma mensagem de user contendo a URL da imagem e o prompt de texto. O modelo extrai o texto e o retorna em choices[0].message.content.

Sem streaming

Python

from openai import OpenAI
import os

PROMPT_TICKET_EXTRACTION = """
Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image.
You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?).
Return the data in JSON format as follows: {'invoice_number': 'xxx', 'departure_station': 'xxx', 'arrival_station': 'xxx', 'departure_date_and_time':'xxx', 'seat_number': 'xxx','ticket_price':'xxx', 'id_card_number': 'xxx', 'passenger_name': 'xxx'},
"""

try:
    client = OpenAI(
        # If the environment variable is not configured, replace with: api_key="sk-xxx"
        api_key=os.getenv("DASHSCOPE_API_KEY"),
        # Singapore region. For US (Virginia), use https://dashscope-us.aliyuncs.com/compatible-mode/v1
        # For China (Beijing), use https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1
        base_url="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
    )
    completion = client.chat.completions.create(
        model="qwen-vl-ocr-2025-11-20",
        messages=[
            {
                "role": "user",
                "content": [
                    {
                        "type": "image_url",
                        "image_url": {"url":"https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg"},
                        # Minimum pixel count. Images below this are upscaled.
                        "min_pixels": 32 * 32 * 3,
                        # Maximum pixel count. Images above this are downscaled.
                        "max_pixels": 32 * 32 * 8192
                    },
                    # Custom prompt. Without this, the model uses: "Please output only the text content from the image without any additional descriptions or formatting."
                    {"type": "text",
                     "text": PROMPT_TICKET_EXTRACTION}
                ]
            }
        ])
    print(completion.choices[0].message.content)
except Exception as e:
    print(f"Error message: {e}")

Node.js

import OpenAI from 'openai';

const PROMPT_TICKET_EXTRACTION = `
Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image.
You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?).
Return the data in JSON format as follows: {'invoice_number': 'xxx', 'departure_station': 'xxx', 'arrival_station': 'xxx', 'departure_date_and_time':'xxx', 'seat_number': 'xxx','ticket_price':'xxx', 'id_card_number': 'xxx', 'passenger_name': 'xxx'}
`;

const client = new OpenAI({
  // If the environment variable is not configured, replace with: apiKey: "sk-xxx"
  apiKey: process.env.DASHSCOPE_API_KEY,
  // For China (Beijing), use https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1
  baseURL: 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1',
});

async function main() {
  const response = await client.chat.completions.create({
    model: 'qwen-vl-ocr-2025-11-20',
    messages: [
      {
        role: 'user',
        content: [
          { type: 'text', text: PROMPT_TICKET_EXTRACTION},
          {
            type: 'image_url',
            image_url: {
              url: 'https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg',
            },
              // Minimum pixel count. Images below this are upscaled.
              "min_pixels": 32 * 32 * 3,
              // Maximum pixel count. Images above this are downscaled.
              "max_pixels": 32 * 32 * 8192
          }
        ]
      }
    ],
  });
  console.log(response.choices[0].message.content)
}

main();

curl

curl -X POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "model": "qwen-vl-ocr-2025-11-20",
  "messages": [
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url":"https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg"},
                    "min_pixels": 3072,
                    "max_pixels": 8388608
                },
                {"type": "text", "text": "Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image. You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?). Return the data in JSON format as follows: {\'invoice_number\': \'xxx\', \'departure_station\': \'xxx\', \'arrival_station\': \'xxx\', \'departure_date_and_time\':\'xxx\', \'seat_number\': \'xxx\',\'ticket_price\':\'xxx\', \'id_card_number\': \'xxx\', \'passenger_name\': \'xxx\'}"}
            ]
        }
    ]
}'

Streaming

Defina stream como true para receber os resultados incrementalmente à medida que o modelo os gera. Python

import os
from openai import OpenAI

PROMPT_TICKET_EXTRACTION = """
Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image.
You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?).
Return the data in JSON format as follows: {'invoice_number': 'xxx','departure_station': 'xxx', 'arrival_station': 'xxx', 'departure_date_and_time':'xxx', 'seat_number': 'xxx','ticket_price':'xxx', 'id_card_number': 'xxx', 'passenger_name': 'xxx'},
"""

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    base_url="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
)
completion = client.chat.completions.create(
    model="qwen-vl-ocr-2025-11-20",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url":"https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg"},
                    "min_pixels": 32 * 32 * 3,
                    "max_pixels": 32 * 32 * 8192
                },
                {"type": "text","text": PROMPT_TICKET_EXTRACTION}
            ]
        }
    ],
    stream=True,
    stream_options={"include_usage": True}
)

for chunk in completion:
    print(chunk.model_dump_json())

Node.js

import OpenAI from 'openai';

const PROMPT_TICKET_EXTRACTION = `
Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image.
You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?).
Return the data in JSON format as follows: {'invoice_number': 'xxx', 'departure_station': 'xxx', 'arrival_station': 'xxx', 'departure_date_and_time':'xxx', 'seat_number': 'xxx','ticket_price':'xxx', 'id_card_number': 'xxx', 'passenger_name': 'xxx'}
`;

const openai = new OpenAI({
  apiKey: process.env.DASHSCOPE_API_KEY,
  baseURL: 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1',
});

async function main() {
  const response = await openai.chat.completions.create({
    model: 'qwen-vl-ocr-2025-11-20',
    messages: [
      {
        role: 'user',
        content: [
          { type: 'text', text: PROMPT_TICKET_EXTRACTION},
          {
            type: 'image_url',
            image_url: {
              url: 'https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg',
            },
              "min_pixels": 32 * 32 * 3,
              "max_pixels": 32 * 32 * 8192
          }
        ]
      }
    ],
    stream: true,
    stream_options:{"include_usage": true}
  });
  let fullContent = ""
  console.log("Streaming output content:")
  for await (const chunk of response) {
    if (chunk.choices[0] && chunk.choices[0].delta.content != null) {
      fullContent += chunk.choices[0].delta.content;
      console.log(chunk.choices[0].delta.content);
    }
  }
  console.log(`Full output content: ${fullContent}`)
}

main();

curl

curl -X POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "model": "qwen-vl-ocr-2025-11-20",
  "messages": [
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url":"https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg"},
                    "min_pixels": 3072,
                    "max_pixels": 8388608
                },
                {"type": "text", "text": "Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image. You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?). Return the data in JSON format as follows: {\'invoice_number\': \'xxx\', \'departure_station\': \'xxx\', \'arrival_station\': \'xxx\', \'departure_date_and_time\':\'xxx\', \'seat_number\': \'xxx\',\'ticket_price\':\'xxx\', \'id_card_number\': \'xxx\', \'passenger_name\': \'xxx\'}"}
            ]
        }
    ],
    "stream": true,
    "stream_options": {"include_usage": true}
}'

Parâmetros da solicitação

Parâmetro	Tipo	Obrigatório	Descrição
`model`	string	Sim	Nome do modelo. Consulte Modelos recomendados para ver os modelos suportados.
`messages`	array	Sim	Array de objetos de mensagem que fornece contexto ao modelo.

Objeto de mensagem Cada mensagem exige um role (deve ser user) e um array content com os seguintes tipos de elementos:

Parâmetro	Tipo	Obrigatório	Descrição
`type`	string	Sim	`text` para entrada de texto, `image_url` para entrada de imagem.
`text`	string	Não	Prompt de texto. Padrão: `"Please output only the text content from the image without any additional descriptions or formatting"`.
`image_url.url`	string	Sim (quando `type` é `image_url`)	URL ou Data URL codificada em Base64 da imagem. Para arquivos locais, consulte Extração de texto.
`min_pixels`	integer	Não	Limiar mínimo de pixels. Imagens abaixo deste valor são ampliadas. Consulte Controle de resolução de imagem.
`max_pixels`	integer	Não	Limiar máximo de pixels. Imagens acima deste valor são reduzidas. Consulte Controle de resolução de imagem.

Parâmetros de geração

Parâmetro	Tipo	Padrão	Descrição
`stream`	boolean	`false`	Defina como `true` para receber respostas incrementais conforme o modelo gera a saída.
`stream_options.include_usage`	boolean	`false`	Quando `stream` for `true`, defina isto como `true` para incluir o uso de tokens no último chunk.
`max_tokens`	integer	Variável	Máximo de tokens na saída. Exceder este valor trunca a resposta. Consulte Limites de tokens de saída.
`temperature`	float	`0.01`	Controla a diversidade da saída. Valores mais altos produzem texto mais variado. Intervalo: [0, 2).
`top_p`	float	`0.001`	Limiar de amostragem por núcleo. Valores mais altos aumentam a diversidade. Intervalo: (0, 1.0]. Defina `temperature` ou `top_p`, mas não ambos.
`top_k`	integer	`1`	Limita o conjunto de tokens candidatos durante a amostragem. Se o valor for None ou maior que 100, a política top_k não é ativada e apenas a política top_p tem efeito. Deve ser >= 0. Não é um parâmetro padrão da OpenAI — passe via `extra_body` no SDK Python: `extra_body={"top_k": xxx}`. No SDK Node.js ou HTTP, passe no nível superior.
`repetition_penalty`	float	`1.0`	Penalidade para sequências repetidas. Valores acima de 1.0 reduzem a repetição. Não é um parâmetro padrão da OpenAI — passe via `extra_body` no SDK Python.
`presence_penalty`	float	`0.0`	Controla a repetição de conteúdo. Intervalo: [-2.0, 2.0]. Valores positivos reduzem a repetição.
`seed`	integer	--	Garante resultados reproduzíveis quando o mesmo valor é usado com parâmetros idênticos. Intervalo: [0, 2^31 - 1].
`logprobs`	boolean	`false`	Defina como `true` para retornar as probabilidades logarítmicas dos tokens de saída.
`top_logprobs`	integer	`0`	Número de tokens mais prováveis a serem retornados por etapa. Intervalo: [0, 5]. Efetivo apenas quando `logprobs` for `true`.
`stop`	string ou array	--	Palavras de parada ou IDs de token. A geração para quando uma string especificada ou `token_id` aparece. Não misture strings e `token_id` s no mesmo array.

Resposta

Resposta sem streaming (`chat.completion`)

{
  "id": "chatcmpl-ba21fa91-dcd6-4dad-90cc-6d49c3c39094",
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "logprobs": null,
      "message": {
        "content": "```json\n{\n    \"seller_name\": \"null\",\n    \"buyer_name\": \"Cai Yingshi\",\n    \"price_excluding_tax\": \"230769.23\",\n    \"organization_code\": \"null\",\n    \"invoice_code\": \"142011726001\"\n}\n```",
        "refusal": null,
        "role": "assistant",
        "annotations": null,
        "audio": null,
        "function_call": null,
        "tool_calls": null
      }
    }
  ],
  "created": 1763283287,
  "model": "qwen-vl-ocr-latest",
  "object": "chat.completion",
  "service_tier": null,
  "system_fingerprint": null,
  "usage": {
    "completion_tokens": 72,
    "prompt_tokens": 1185,
    "total_tokens": 1257,
    "completion_tokens_details": {
      "accepted_prediction_tokens": null,
      "audio_tokens": null,
      "reasoning_tokens": null,
      "rejected_prediction_tokens": null,
      "text_tokens": 72
    },
    "prompt_tokens_details": {
      "audio_tokens": null,
      "cached_tokens": null,
      "image_tokens": 1001,
      "text_tokens": 184
    }
  }
}

Campo	Tipo	Descrição
`id`	string	Identificador único da solicitação.
`choices`	array	Conteúdo gerado pelo modelo.
`choices[].finish_reason`	string	`stop` quando a geração for concluída normalmente, `length` quando truncada devido ao limite de tokens.
`choices[].index`	integer	Posição no array `choices`.
`choices[].message.content`	string	Texto extraído ou saída estruturada do modelo.
`choices[].message.role`	string	Sempre `assistant`.
`choices[].message.refusal`	string	Sempre `null`.
`choices[].message.audio`	object	Sempre `null`.
`choices[].message.function_call`	object	Sempre `null`.
`choices[].message.tool_calls`	array	Sempre `null`.
`created`	integer	Timestamp UNIX da solicitação.
`model`	string	Modelo utilizado.
`object`	string	Sempre `chat.completion`.
`service_tier`	string	Sempre `null`.
`system_fingerprint`	string	Sempre `null`.
`usage.completion_tokens`	integer	Contagem de tokens de saída.
`usage.prompt_tokens`	integer	Contagem de tokens de entrada.
`usage.total_tokens`	integer	Soma de `prompt_tokens` e `completion_tokens`.
`usage.completion_tokens_details.text_tokens`	integer	Tokens de saída de texto. Outros campos em `completion_tokens_details` são sempre `null`.
`usage.prompt_tokens_details.image_tokens`	integer	Tokens de entrada de imagem.
`usage.prompt_tokens_details.text_tokens`	integer	Tokens de entrada de texto. Outros campos em `prompt_tokens_details` são sempre `null`.

Resposta com streaming (`chat.completion.chunk`)

Quando stream é true, a resposta é entregue como uma série de chunks de Server-Sent Event (SSE). Cada chunk segue a mesma estrutura da resposta sem streaming, com estas diferenças:

object é sempre chat.completion.chunk.
choices[].delta substitui choices[].message. O objeto delta possui os mesmos campos que message.
choices[].delta.role é retornado apenas no primeiro chunk.
finish_reason é null durante a geração, stop na conclusão, ou length se truncado.
Quando include_usage é true, o último chunk tem um array choices vazio e inclui o objeto usage.

{"id":"chatcmpl-f6fbdc0d-78d6-418f-856f-f099c2e4859b","choices":[{"delta":{"content":"","function_call":null,"refusal":null,"role":"assistant","tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}],"created":1764139204,"model":"qwen-vl-ocr-latest","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-f6fbdc0d-78d6-418f-856f-f099c2e4859b","choices":[{"delta":{"content":"```","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}],"created":1764139204,"model":"qwen-vl-ocr-latest","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-f6fbdc0d-78d6-418f-856f-f099c2e4859b","choices":[{"delta":{"content":"json","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}],"created":1764139204,"model":"qwen-vl-ocr-latest","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
......
{"id":"chatcmpl-f6fbdc0d-78d6-418f-856f-f099c2e4859b","choices":[{"delta":{"content":"","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":"stop","index":0,"logprobs":null}],"created":1764139204,"model":"qwen-vl-ocr-latest","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-f6fbdc0d-78d6-418f-856f-f099c2e4859b","choices":[],"created":1764139204,"model":"qwen-vl-ocr-latest","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":{"completion_tokens":141,"prompt_tokens":513,"total_tokens":654,"completion_tokens_details":{"accepted_prediction_tokens":null,"audio_tokens":null,"reasoning_tokens":null,"rejected_prediction_tokens":null,"text_tokens":141},"prompt_tokens_details":{"audio_tokens":null,"cached_tokens":null,"image_tokens":332,"text_tokens":181}}}

Controle de resolução de imagem

min_pixels e max_pixels controlam o redimensionamento da imagem antes do processamento. A proporção de tokens por pixel depende da versão do modelo:

Modelo	Pixels por token	Padrão de min_pixels(mínimo)	Padrão de max_pixels****	Máximo de max_pixels****
`qwen3.5-ocr`, `qwen-vl-ocr-latest`, `qwen-vl-ocr-2025-11-20`	32 x 32 = 1.024	3.072 (3 tokens)	8.388.608 (8.192 tokens)	30.720.000 (30.000 tokens)
`qwen-vl-ocr`, `qwen-vl-ocr-2025-08-28` e anteriores	28 x 28 = 784	3.136 (4 tokens)	6.422.528 (8.192 tokens)	23.520.000 (30.000 tokens)

Comportamento de redimensionamento:

Se a contagem de pixels da imagem estiver abaixo de min_pixels, a imagem será ampliada até exceder min_pixels.
Caso a contagem de pixels esteja dentro de [min_pixels, max_pixels], a imagem original será usada sem redimensionamento.
Quando a contagem de pixels exceder max_pixels, a imagem será reduzida para ficar abaixo de max_pixels.

Limites de tokens de saída

Modelo	max_tokenspadrão e máximo
`qwen3.5-ocr`, `qwen-vl-ocr-latest`, `qwen-vl-ocr-2025-11-20`, `qwen-vl-ocr-2024-10-28`	Igual ao comprimento máximo de saída do modelo. Consulte Seleção de modelo.
`qwen-vl-ocr`, `qwen-vl-ocr-2025-04-13`, `qwen-vl-ocr-2025-08-28`	4.096

Para qwen-vl-ocr, qwen-vl-ocr-2025-04-13 e qwen-vl-ocr-2025-08-28 , max_tokens tem como padrão 4096. Para aumentar esse valor (4097–8192), entre em contato com seu gerente comercial informando: ID da sua conta Alibaba Cloud, tipo de imagem (ex.: documentos, e-commerce, contratos), nome do modelo, QPS estimado e volume diário de solicitações, além da porcentagem de solicitações que excedem 4096 tokens de saída.

API DashScope

Endpoints

Região	Endpoint HTTP
Singapura	`POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation`
EUA (Virgínia)	`POST https://dashscope-us.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation`
China (Pequim)	`POST https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation`

Configuração da URL base do SDK: Python:

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

Java (Método 1 — construtor):

import com.alibaba.dashscope.protocol.Protocol;
MultiModalConversation conv = new MultiModalConversation(Protocol.HTTP.getValue(), "https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1");

Java (Método 2 — bloco estático):

import com.alibaba.dashscope.utils.Constants;
Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";

Substitua o domínio por dashscope-us.aliyuncs.com para a região EUA (Virgínia) ou {WorkspaceId}.cn-beijing.maas.aliyuncs.com para a região China (Pequim). Para a região China (Pequim), não é necessário definir base_url nas chamadas do SDK.

Obtenha uma chave de API e defina-a como variável de ambiente. Se você utilizar o SDK DashScope, também deve instalar o SDK DashScope.

Tarefas integradas

A API DashScope fornece tarefas de OCR integradas por meio do parâmetro ocr_options. Cada tarefa usa um prompt padrão otimizado, eliminando a necessidade de uma mensagem text.

Tarefa	Valor de ocr_options.task****	Formato de saída
Reconhecimento geral de texto	`text_recognition`	Texto simples
Reconhecimento de alta precisão	`advanced_recognition`	Texto simples com caixas delimitadoras
Extração de informações	`key_information_extraction`	Pares chave-valor estruturados
Análise de tabelas	`table_parsing`	Estrutura da tabela
Análise de documentos	`document_parsing`	Estrutura do documento
Reconhecimento de fórmulas	`formula_recognition`	Fórmulas LaTeX
Reconhecimento multilíngue	`multi_lan`	Texto multilíngue

Reconhecimento de alta precisão

Retorna texto com dados posicionais para cada linha reconhecida. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "advanced_recognition"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

// dashscope SDK version >= 2.21.8
import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.ADVANCED_RECOGNITION)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "advanced_recognition"
    }
  }
}
'

Extração de informações

Extrai dados estruturados de chave-valor de imagens. Especifique os campos a serem extraídos em task_config.result_schema. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [
      {
        "role":"user",
        "content":[
          {
              "image":"http://duguang-labelling.oss-cn-shanghai.aliyuncs.com/demo_ocr/receipt_zh_demo.jpg",
              "min_pixels": 3072,
              "max_pixels": 8388608,
              "enable_rotate": False
          }
        ]
      }
    ]

params = {
  "ocr_options":{
    "task": "key_information_extraction",
    "task_config": {
      "result_schema": {
          "Ride Date": "Corresponds to the ride date and time in the image, in the format YYYY-MM-DD, for example, 2025-03-05",
          "Invoice Code": "Extract the invoice code from the image, usually a combination of numbers or letters",
          "Invoice Number": "Extract the number from the invoice, usually composed of only digits."
      }
    }
  }
}

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    **params)

print(response.output.choices[0].message.content[0]["ocr_result"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;
import com.google.gson.JsonObject;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "http://duguang-labelling.oss-cn-shanghai.aliyuncs.com/demo_ocr/receipt_zh_demo.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        JsonObject resultSchema = new JsonObject();
        resultSchema.addProperty("Ride Date", "Corresponds to the ride date and time in the image, in the format YYYY-MM-DD, for example, 2025-03-05");
        resultSchema.addProperty("Invoice Code", "Extract the invoice code from the image, usually a combination of numbers or letters");
        resultSchema.addProperty("Invoice Number", "Extract the number from the invoice, usually composed of only digits.");

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.KEY_INFORMATION_EXTRACTION)
                .taskConfig(OcrOptions.TaskConfig.builder().resultSchema(resultSchema).build())
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("ocr_result"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "http://duguang-labelling.oss-cn-shanghai.aliyuncs.com/demo_ocr/receipt_zh_demo.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "key_information_extraction",
      "task_config": {
        "result_schema": {
          "Ride Date": "Corresponds to the ride date and time in the image, in the format YYYY-MM-DD, for example, 2025-03-05",
          "Invoice Code": "Extract the invoice code from the image, usually a combination of numbers or letters",
          "Invoice Number": "Extract the number from the invoice, usually composed of only digits."
        }
      }
    }
  }
}
'

Análise de tabelas

Extrai a estrutura de tabelas a partir de imagens. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "http://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/doc_parsing/tables/photo/eng/17.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "table_parsing"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/doc_parsing/tables/photo/eng/17.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.TABLE_PARSING)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "http://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/doc_parsing/tables/photo/eng/17.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "table_parsing"
    }
  }
}
'

Análise de documentos

Extrai o layout estrutural e o texto de documentos. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "https://img.alicdn.com/imgextra/i1/O1CN01ukECva1cisjyK6ZDK_!!6000000003635-0-tps-1500-1734.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "document_parsing"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://img.alicdn.com/imgextra/i1/O1CN01ukECva1cisjyK6ZDK_!!6000000003635-0-tps-1500-1734.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.DOCUMENT_PARSING)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "https://img.alicdn.com/imgextra/i1/O1CN01ukECva1cisjyK6ZDK_!!6000000003635-0-tps-1500-1734.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "document_parsing"
    }
  }
}
'

Reconhecimento de fórmulas

Extrai fórmulas matemáticas de imagens e as retorna no formato LaTeX. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "http://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/formula_handwriting/test/inline_5_4.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "formula_recognition"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "http://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/formula_handwriting/test/inline_5_4.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.FORMULA_RECOGNITION)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "http://duguang-llm.oss-cn-hangzhou.aliyuncs.com/llm_data_keeper/data/formula_handwriting/test/inline_5_4.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "formula_recognition"
    }
  }
}
'

Reconhecimento geral de texto

Extrai texto simples de imagens sem formatação estrutural. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "text_recognition"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.TEXT_RECOGNITION)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/ctdzex/biaozhun.jpg",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "text_recognition"
    }
  }
}
'

Reconhecimento multilíngue

Reconhece texto em vários idiomas a partir de imagens. Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

messages = [{
            "role": "user",
            "content": [{
                "image": "https://img.alicdn.com/imgextra/i2/O1CN01VvUMNP1yq8YvkSDFY_!!6000000006629-2-tps-6000-3000.png",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192,
                "enable_rotate": False}]
            }]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model='qwen-vl-ocr-2025-11-20',
    messages=messages,
    ocr_options={"task": "multi_lan"}
)
print(response["output"]["choices"][0]["message"].content[0]["text"])

Java

import java.util.Arrays;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.aigc.multimodalconversation.OcrOptions;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://img.alicdn.com/imgextra/i2/O1CN01VvUMNP1yq8YvkSDFY_!!6000000006629-2-tps-6000-3000.png");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        map.put("enable_rotate", false);

        OcrOptions ocrOptions = OcrOptions.builder()
                .task(OcrOptions.Task.MULTI_LAN)
                .build();
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map
                        )).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .ocrOptions(ocrOptions)
                .build();
        MultiModalConversationResult result = conv.call(param);
        System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent().get(0).get("text"));
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '
{
  "model": "qwen-vl-ocr-2025-11-20",
  "input": {
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "image": "https://img.alicdn.com/imgextra/i2/O1CN01VvUMNP1yq8YvkSDFY_!!6000000006629-2-tps-6000-3000.png",
            "min_pixels": 3072,
            "max_pixels": 8388608,
            "enable_rotate": false
          }
        ]
      }
    ]
  },
  "parameters": {
    "ocr_options": {
      "task": "multi_lan"
    }
  }
}
'

Streaming (DashScope)

Ative a saída em streaming para receber resultados incrementalmente. O método varia conforme o SDK:

SDK Python: Defina stream=True e incremental_output=True.
SDK Java: Utilize a interface streamCall.
HTTP: Configure o cabeçalho X-DashScope-SSE: enable.

Python

import os
import dashscope

dashscope.base_http_api_url = 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1'

PROMPT_TICKET_EXTRACTION = """
Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image.
You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?).
Return the data in JSON format as follows: {'invoice_number': 'xxx','departure_station': 'xxx', 'arrival_station': 'xxx', 'departure_date_and_time':'xxx', 'seat_number': 'xxx','ticket_price':'xxx', 'id_card_number': 'xxx', 'passenger_name': 'xxx'},
"""

messages = [
    {
        "role": "user",
        "content": [
            {
                "image": "https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg",
                "min_pixels": 32 * 32 * 3,
                "max_pixels": 32 * 32 * 8192},
            {
                "type": "text",
                "text": PROMPT_TICKET_EXTRACTION
            }
        ]
    }
]

response = dashscope.MultiModalConversation.call(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="qwen-vl-ocr-2025-11-20",
    messages=messages,
    stream=True,
    incremental_output=True,
)
full_content = ""
print("Streaming output content:")
for response in response:
    try:
        print(response["output"]["choices"][0]["message"].content[0]["text"])
        full_content += response["output"]["choices"][0]["message"].content[0]["text"]
    except:
        pass
print(f"Full content: {full_content}")

Java

import java.util.*;

import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversation;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationParam;
import com.alibaba.dashscope.aigc.multimodalconversation.MultiModalConversationResult;
import com.alibaba.dashscope.common.MultiModalMessage;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.exception.UploadFileException;
import io.reactivex.Flowable;
import com.alibaba.dashscope.utils.Constants;

public class Main {

    static {
        Constants.baseHttpApiUrl="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1";
    }

    public static void simpleMultiModalConversationCall()
            throws ApiException, NoApiKeyException, UploadFileException {
        MultiModalConversation conv = new MultiModalConversation();
        Map<String, Object> map = new HashMap<>();
        map.put("image", "https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg");
        map.put("max_pixels", 8388608);
        map.put("min_pixels", 3072);
        MultiModalMessage userMessage = MultiModalMessage.builder().role(Role.USER.getValue())
                .content(Arrays.asList(
                        map,
                        Collections.singletonMap("text", "Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image. You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?). Return the data in JSON format as follows: {\'invoice_number\': \'xxx\', \'departure_station\': \'xxx\', \'arrival_station\': \'xxx\', \'departure_date_and_time\':\'xxx\', \'seat_number\': \'xxx\',\'ticket_price\':\'xxx\', \'id_card_number\': \'xxx\', \'passenger_name\': \'xxx\'"))).build();
        MultiModalConversationParam param = MultiModalConversationParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-vl-ocr-2025-11-20")
                .message(userMessage)
                .incrementalOutput(true)
                .build();
        Flowable<MultiModalConversationResult> result = conv.streamCall(param);
        result.blockingForEach(item -> {
            try {
                List<Map<String, Object>> contentList = item.getOutput().getChoices().get(0).getMessage().getContent();
                if (!contentList.isEmpty()){
                    System.out.println(contentList.get(0).get("text"));
                }//
            } catch (Exception e){
                System.exit(0);
            }
        });
    }

    public static void main(String[] args) {
        try {
            simpleMultiModalConversationCall();
        } catch (ApiException | NoApiKeyException | UploadFileException e) {
            System.out.println(e.getMessage());
        }
        System.exit(0);
    }
}

curl

curl --location 'https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--header 'X-DashScope-SSE: enable' \
--data '
{
    "model": "qwen-vl-ocr-2025-11-20",
    "input": {
        "messages": [
            {
              "role": "user",
              "content": [
                  {
                      "image": "https://img.alicdn.com/imgextra/i2/O1CN01ktT8451iQutqReELT_!!6000000004408-0-tps-689-487.jpg",
                      "min_pixels": 3072,
                      "max_pixels": 8388608
                  },
                  {"type": "text", "text": "Please extract the invoice number, train number, departure station, arrival station, departure date and time, seat number, seat class, ticket price, ID card number, and passenger name from the train ticket image. You must accurately extract the key information. Do not omit or fabricate information. Replace any single character that is blurry or obscured by strong light with an English question mark (?). Return the data in JSON format as follows: {\'invoice_number\': \'xxx\', \'departure_station\': \'xxx\', \'arrival_station\': \'xxx\', \'departure_date_and_time\':\'xxx\', \'seat_number\': \'xxx\',\'ticket_price\':\'xxx\', \'id_card_number\': \'xxx\', \'passenger_name\': \'xxx\'}"}
              ]
            }
        ]
    },
    "parameters": {
        "incremental_output": true
    }
}'

Parâmetros da solicitação

Parâmetro	Tipo	Obrigatório	Descrição
`model`	string	Sim	Nome do modelo. Consulte Modelos recomendados para ver os modelos suportados.
`input.messages`	array	Sim	Um array de objetos de mensagem.

Objeto de mensagem Cada mensagem requer um role (deve ser user) e um campo content (string ou array). Use uma string para entrada apenas de texto. Utilize um array se a entrada incluir dados de imagem, com estes campos:

Parâmetro	Tipo	Obrigatório	Descrição
`image`	string	Não	URL, Data URL Base64 ou caminho local da imagem. Consulte Passagem de arquivos locais.
`text`	string	Não	Prompt de texto. Padrão: `"Please output only the text content from the image without any additional descriptions or formatting"`. Não é obrigatório ao usar uma tarefa integrada.
`enable_rotate`	boolean	Não	Defina como `true` para corrigir imagens inclinadas. Padrão: `false`.
`min_pixels`	integer	Não	Limiar mínimo de pixels. Consulte Controle de resolução de imagem.
`max_pixels`	integer	Não	Limiar máximo de pixels. Consulte Controle de resolução de imagem.

Parâmetros de geração Configure-os no objeto parameters para chamadas HTTP.

Parâmetro	Tipo	Padrão	Descrição
`max_tokens`	integer	Variável	Máximo de tokens na saída. Consulte Limites de tokens de saída. No SDK Java, use `maxTokens`.
`stream`	boolean	`false`	Ativa a saída em streaming. Apenas SDK Python. Para Java, use `streamCall`. Para HTTP, defina `X-DashScope-SSE: enable`.
`incremental_output`	boolean	`false`	Quando `true` (recomendado), cada chunk contém apenas conteúdo novo. Quando `false`, cada chunk contém a sequência completa até o momento. No SDK Java, use `incrementalOutput`.
`temperature`	float	`0.01`	Controla a diversidade da saída. Intervalo: [0, 2).
`top_p`	float	`0.001`	Limiar de amostragem por núcleo. Intervalo: (0, 1.0]. Defina `temperature` ou `top_p`, mas não ambos.
`top_k`	integer	`1`	Limita o conjunto de tokens candidatos durante a amostragem. Se o valor for None ou maior que 100, a política top_k não é ativada e apenas a política top_p tem efeito. Deve ser >= 0.
`repetition_penalty`	float	`1.0`	Penalidade para sequências repetidas. Valores acima de 1.0 reduzem a repetição.
`presence_penalty`	float	`0.0`	Controla a repetição de conteúdo. Intervalo: [-2.0, 2.0].
`seed`	integer	--	Garante resultados reproduzíveis. Intervalo: [0, 2^31 - 1].
`logprobs`	boolean	`false`	Defina como `true` para retornar probabilidades logarítmicas. Modelos suportados: `qwen-vl-ocr-2025-04-13` e posteriores. No SDK Java, use o mesmo nome. Para HTTP, coloque em `parameters`.
`top_logprobs`	integer	`0`	Número de tokens mais prováveis por etapa. Intervalo: [0, 5]. Efetivo apenas quando `logprobs` for `true`. No SDK Java, use `topLogprobs`. Para HTTP, coloque em `parameters`.
`stop`	string ou array	--	Palavras de parada ou IDs de token. A geração para quando uma string especificada ou `token_id` aparece. Não misture strings e `token_id` s no mesmo array.

Parâmetros de tarefas integradas (ocr_options) Ao usar uma tarefa integrada, passe ocr_options em parameters (HTTP), como argumento nomeado (SDK Python) ou via builder OcrOptions (SDK Java).

Parâmetro	Tipo	Obrigatório	Descrição
`ocr_options.task`	string	Sim	Nome da tarefa integrada. Valores válidos: `text_recognition`, `key_information_extraction`, `document_parsing`, `table_parsing`, `formula_recognition`, `multi_lan`, `advanced_recognition`.
`ocr_options.task_config`	object	Não	Configuração para `key_information_extraction`.
`ocr_options.task_config.result_schema`	object	Não	Objeto JSON especificando campos a extrair. As chaves são nomes de campos, os valores são descrições opcionais para melhorar a precisão. Suporta até três níveis de aninhamento.

Exemplo de result_schema****:

"result_schema": {
     "invoice_number": "The unique identification number of the invoice, usually a combination of numbers and letters.",
     "issue_date": "The date the invoice was issued. Extract it in YYYY-MM-DD format, for example, 2023-10-26.",
     "seller_name": "The full company name of the seller shown on the invoice.",
     "total_amount": "The total amount on the invoice, including tax. Extract the numerical value and keep two decimal places, for example, 123.45."
}

No SDK Java, este parâmetro é OcrOptions . A versão mínima do SDK Python DashScope é 1.22.2. A versão mínima do SDK Java é 2.18.4. Para advanced_recognition , é necessário SDK Java >= 2.21.8.

Resposta

A API DashScope utiliza formato de resposta idêntico para saída com e sem streaming.

{
  "status_code": 200,
  "request_id": "8f8c0f6e-6805-4056-bb65-d26d66080a41",
  "code": "",
  "message": "",
  "output": {
    "text": null,
    "finish_reason": null,
    "choices": [
      {
        "finish_reason": "stop",
        "message": {
          "role": "assistant",
          "content": [
            {
              "ocr_result": {
                "kv_result": {
                  "price_excluding_tax": "230769.23",
                  "invoice_code": "142011726001",
                  "organization_code": "null",
                  "buyer_name": "Cai Yingshi",
                  "seller_name": "null"
                }
              },
              "text": "```json\n{\n    \"price_excluding_tax\": \"230769.23\",\n    \"invoice_code\": \"142011726001\",\n    \"organization_code\": \"null\",\n    \"buyer_name\": \"Cai Yingshi\",\n    \"seller_name\": \"null\"\n}\n```"
            }
          ]
        }
      }
    ],
    "audio": null
  },
  "usage": {
    "input_tokens": 926,
    "output_tokens": 72,
    "characters": 0,
    "image_tokens": 754,
    "input_tokens_details": {
      "image_tokens": 754,
      "text_tokens": 172
    },
    "output_tokens_details": {
      "text_tokens": 72
    },
    "total_tokens": 998
  }
}

Campo	Tipo	Descrição
`status_code`	string	`200` indica sucesso. O SDK Java lança uma exceção em vez de retornar este campo.
`request_id`	string	Identificador único da solicitação. No SDK Java, este é `requestId`.
`code`	string	Código de erro. Vazio em caso de sucesso. Apenas o SDK Python retorna este campo.
`output.text`	string	Sempre `null`.
`output.finish_reason`	string	`null` durante a geração, `stop` quando concluída, `length` quando truncada.
`output.choices[].finish_reason`	string	Mesmos valores que `output.finish_reason`.
`output.choices[].message.role`	string	Sempre `assistant`.
`output.choices[].message.content[].text`	string	Texto extraído ou saída formatada do modelo.
`output.choices[].message.content[].ocr_result`	object	Retornado para tarefas integradas (`key_information_extraction`, `advanced_recognition`).
`output.choices[].message.content[].ocr_result.kv_result`	object	Resultados da extração de chave-valor (para `key_information_extraction`).
`output.choices[].message.content[].ocr_result.words_info`	array	Resultados de linhas de texto com dados posicionais (para `advanced_recognition`).
`output.choices[].message.content[].ocr_result.words_info[].rotate_rect`	array	`[center_x, center_y, width, height, angle]` — retângulo delimitador rotacionado.
`output.choices[].message.content[].ocr_result.words_info[].location`	array	`[x1, y1, x2, y2, x3, y3, x4, y4]` — quatro vértices no sentido horário a partir do canto superior esquerdo.
`output.choices[].message.content[].ocr_result.words_info[].text`	string	Conteúdo da linha de texto.
`output.choices[].message.logprobs`	object	Informações de probabilidade logarítmica, retornadas quando `logprobs` é `true`.
`usage.input_tokens`	integer	Contagem de tokens de entrada.
`usage.output_tokens`	integer	Contagem de tokens de saída.
`usage.characters`	integer	Fixado em 0.
`usage.total_tokens`	integer	Soma de `input_tokens` e `output_tokens`.
`usage.image_tokens`	integer	Tokens correspondentes à entrada de imagem.
`usage.input_tokens_details.image_tokens`	integer	Tokens de entrada de imagem.
`usage.input_tokens_details.text_tokens`	integer	Tokens de entrada de texto.
`usage.output_tokens_details.text_tokens`	integer	Tokens de saída de texto.

Modelos suportados

Modelo	Descrição
`qwen3.5-ocr`	Baseado na arquitetura Qwen3.5. Mais rápido e preciso. Grandes melhorias na extração de informações, posicionamento de texto e suporte a conversas multi-turno. Comprimento de contexto estendido para 128K.
`qwen-vl-ocr-latest`	Aponta sempre para a versão mais recente.
`qwen-vl-ocr-2025-11-20`	Snapshot datado mais recente.
`qwen-vl-ocr-2025-08-28`	Versão anterior.
`qwen-vl-ocr-2025-04-13`	Versão anterior.
`qwen-vl-ocr-2024-10-28`	Versão anterior.
`qwen-vl-ocr`	Modelo base.

Códigos de erro

Se uma chamada de modelo retornar um erro, consulte Mensagens de erro para resolver o problema.

​API compatível com OpenAI

​Endpoints

​Pré-requisitos

​Início rápido

​Sem streaming

​Streaming

​Parâmetros da solicitação

​Resposta

​Resposta sem streaming (chat.completion)

​Resposta com streaming (chat.completion.chunk)

​Controle de resolução de imagem

​Limites de tokens de saída

​API DashScope

​Endpoints

​Tarefas integradas

​Reconhecimento de alta precisão

​Extração de informações

​Análise de tabelas

​Análise de documentos

​Reconhecimento de fórmulas

​Reconhecimento geral de texto

​Reconhecimento multilíngue

​Streaming (DashScope)

​Parâmetros da solicitação

​Resposta

​Modelos suportados

​Códigos de erro

API compatível com OpenAI

Endpoints

Pré-requisitos

Início rápido

Sem streaming

Streaming

Parâmetros da solicitação

Resposta

Resposta sem streaming (`chat.completion`)

Resposta com streaming (`chat.completion.chunk`)

Controle de resolução de imagem

Limites de tokens de saída

API DashScope

Endpoints

Tarefas integradas

Reconhecimento de alta precisão

Extração de informações

Análise de tabelas

Análise de documentos

Reconhecimento de fórmulas

Reconhecimento geral de texto

Reconhecimento multilíngue

Streaming (DashScope)

Parâmetros da solicitação

Resposta

Modelos suportados

Códigos de erro