Cite suas fontes - Anthropic

Recomendamos usar o recurso de citação integrado à API em vez de uma abordagem baseada em prompt. Usar o recurso de citação da API melhorará a qualidade das citações, garantirá que todas as citações retornadas sejam válidas e potencialmente reduzirá o número de tokens de saída necessários.

Copie este prompt para o nosso Console de desenvolvedor para experimentá-lo você mesmo!

	Conteúdo
System	Você é um assistente de pesquisa especializado. Aqui está um documento sobre o qual você responderá perguntas: [Texto completo do relatório SEC 10-K 2023 da Matterport, não colado aqui por brevidade] Primeiro, encontre as citações do documento que são mais relevantes para responder à pergunta e, em seguida, imprima-as em ordem numerada. As citações devem ser relativamente curtas. Se não houver citações relevantes, escreva “Sem citações relevantes”. Em seguida, responda à pergunta, começando com “Resposta:”. Não inclua ou faça referência ao conteúdo citado literalmente na resposta. Não diga “De acordo com a Citação [1]” ao responder. Em vez disso, faça referências às citações relevantes para cada seção da resposta apenas adicionando seus números entre colchetes no final das frases relevantes. Assim, o formato de sua resposta geral deve se parecer com o que é mostrado entre as tags . Certifique-se de seguir a formatação e o espaçamento exatamente. Citações: [1] “A Empresa X relatou receita de $12 milhões em 2021.” [2] “Quase 90% da receita veio de vendas de widgets, com vendas de gadgets compondo os 10% restantes.” Resposta: A Empresa X ganhou $12 milhões. [1] Quase 90% disso foi de vendas de widgets. [2] Se a pergunta não puder ser respondida pelo documento, diga isso.
User	A Matterport está indo bem?

Conteúdo

System

Você é um assistente de pesquisa especializado. Aqui está um documento sobre o qual você responderá perguntas:
[Texto completo do relatório SEC 10-K 2023 da Matterport, não colado aqui por brevidade]

Primeiro, encontre as citações do documento que são mais relevantes para responder à pergunta e, em seguida, imprima-as em ordem numerada. As citações devem ser relativamente curtas.

Se não houver citações relevantes, escreva “Sem citações relevantes”.

Em seguida, responda à pergunta, começando com “Resposta:”. Não inclua ou faça referência ao conteúdo citado literalmente na resposta. Não diga “De acordo com a Citação [1]” ao responder. Em vez disso, faça referências às citações relevantes para cada seção da resposta apenas adicionando seus números entre colchetes no final das frases relevantes.

Assim, o formato de sua resposta geral deve se parecer com o que é mostrado entre as tags . Certifique-se de seguir a formatação e o espaçamento exatamente.
Citações:
[1] “A Empresa X relatou receita de $12 milhões em 2021.”
[2] “Quase 90% da receita veio de vendas de widgets, com vendas de gadgets compondo os 10% restantes.”

Resposta:
A Empresa X ganhou $12 milhões. [1] Quase 90% disso foi de vendas de widgets. [2]

Se a pergunta não puder ser respondida pelo documento, diga isso.

User

A Matterport está indo bem?

Exemplo de saída

Citações:
[1] “Recentemente experimentamos um rápido crescimento. Nossos assinantes cresceram aproximadamente 49 vezes de 31 de dezembro de 2018 a 31 de dezembro de 2022. Nossa receita aumentou aproximadamente 22% para $136,1 milhões no ano encerrado em 31 de dezembro de 2022, de aproximadamente $111,2 milhões no ano encerrado em 31 de dezembro de 2021.”
[2] “Nosso lucro bruto diminuiu em $8,1 milhões ou 14%, para $51,8 milhões no ano encerrado em 31 de dezembro de 2022, de $60,0 milhões no ano encerrado em 31 de dezembro de 2021, principalmente atribuível a certos custos disruptivos e incrementais devido às restrições da cadeia de suprimentos global no ano fiscal de 2022.”
[3] “Nossa taxa de expansão líquida em dólares de 103% para os três meses encerrados em 31 de dezembro de 2022 demonstra a aderência e o potencial de crescimento de nossa plataforma.”

Resposta:
A Matterport experimentou um rápido crescimento de assinantes e receita nos últimos anos. [1] No entanto, seu lucro bruto diminuiu em 2022 devido a problemas na cadeia de suprimentos. [2] Sua taxa de expansão líquida em dólares indica forte retenção de assinantes e potencial para crescimento contínuo. [3] No geral, apesar da queda no lucro bruto, o crescimento de assinantes e receita da Matterport, juntamente com a aderência de sua plataforma, sugerem que a empresa está indo relativamente bem.

Solicitação de API

import anthropic

client = anthropic.Anthropic(  # defaults to os.environ.get("ANTHROPIC_API_KEY")
    api_key="my_api_key",
)
message = client.messages.create(
    model="claude-opus-4-20250514",
    max_tokens=2000,
    temperature=0,
    system='You are an expert research assistant. Here is a document you will answer questions about: \n<doc> \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity] \n</doc> \n \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short. \n \nIf there are no relevant quotes, write "No relevant quotes" instead. \n \nThen, answer the question, starting with "Answer:". Do not include or reference quoted content verbatim in the answer. Don\'t say "According to Quote [1]" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences. \n \nThus, the format of your overall response should look like what\'s shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly. \n<example> \nQuotes: \n[1] "Company X reported revenue of \$12 million in 2021." \n[2] "Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%." \n \nAnswer: \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2] \n</example> \n \nIf the question cannot be answered by the document, say so.',
    messages=[
        {
            "role": "user",
            "content": [{"type": "text", "text": "Is Matterport doing well?"}],
        }
    ],
)
print(message.content)

import anthropic

client = anthropic.Anthropic(  # defaults to os.environ.get("ANTHROPIC_API_KEY")
    api_key="my_api_key",
)
message = client.messages.create(
    model="claude-opus-4-20250514",
    max_tokens=2000,
    temperature=0,
    system='You are an expert research assistant. Here is a document you will answer questions about: \n<doc> \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity] \n</doc> \n \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short. \n \nIf there are no relevant quotes, write "No relevant quotes" instead. \n \nThen, answer the question, starting with "Answer:". Do not include or reference quoted content verbatim in the answer. Don\'t say "According to Quote [1]" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences. \n \nThus, the format of your overall response should look like what\'s shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly. \n<example> \nQuotes: \n[1] "Company X reported revenue of \$12 million in 2021." \n[2] "Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%." \n \nAnswer: \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2] \n</example> \n \nIf the question cannot be answered by the document, say so.',
    messages=[
        {
            "role": "user",
            "content": [{"type": "text", "text": "Is Matterport doing well?"}],
        }
    ],
)
print(message.content)

import Anthropic from "@anthropic-ai/sdk";

const anthropic = new Anthropic({
  apiKey: "my_api_key", // defaults to process.env["ANTHROPIC_API_KEY"]
});

const msg = await anthropic.messages.create({
  model: "claude-opus-4-20250514",
  max_tokens: 2000,
  temperature: 0,
  system: "You are an expert research assistant. Here is a document you will answer questions about:  \n<doc>  \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity]  \n</doc>  \n  \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short.  \n  \nIf there are no relevant quotes, write \"No relevant quotes\" instead.  \n  \nThen, answer the question, starting with \"Answer:\". Do not include or reference quoted content verbatim in the answer. Don't say \"According to Quote [1]\" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences.  \n  \nThus, the format of your overall response should look like what's shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly.  \n<example>  \nQuotes:  \n[1] \"Company X reported revenue of \$12 million in 2021.\"  \n[2] \"Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%.\"  \n  \nAnswer:  \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2]  \n</example>  \n  \nIf the question cannot be answered by the document, say so.",
  messages: [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Is Matterport doing well?"
        }
      ]
    }
  ]
});
console.log(msg);

from anthropic import AnthropicBedrock

# See https://docs.anthropic.com/claude/reference/claude-on-amazon-bedrock
# for authentication options
client = AnthropicBedrock()

message = client.messages.create(
    model="anthropic.claude-opus-4-20250514-v1:0",
    max_tokens=2000,
    temperature=0,
    system="You are an expert research assistant. Here is a document you will answer questions about:  \n<doc>  \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity]  \n</doc>  \n  \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short.  \n  \nIf there are no relevant quotes, write \"No relevant quotes\" instead.  \n  \nThen, answer the question, starting with \"Answer:\". Do not include or reference quoted content verbatim in the answer. Don't say \"According to Quote [1]\" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences.  \n  \nThus, the format of your overall response should look like what's shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly.  \n<example>  \nQuotes:  \n[1] \"Company X reported revenue of \$12 million in 2021.\"  \n[2] \"Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%.\"  \n  \nAnswer:  \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2]  \n</example>  \n  \nIf the question cannot be answered by the document, say so.",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "text",
                    "text": "Is Matterport doing well?"
                }
            ]
        }
    ]
)
print(message.content)

import AnthropicBedrock from "@anthropic-ai/bedrock-sdk";

// See https://docs.anthropic.com/claude/reference/claude-on-amazon-bedrock
// for authentication options
const client = new AnthropicBedrock();

const msg = await client.messages.create({
  model: "anthropic.claude-opus-4-20250514-v1:0",
  max_tokens: 2000,
  temperature: 0,
  system: "You are an expert research assistant. Here is a document you will answer questions about:  \n<doc>  \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity]  \n</doc>  \n  \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short.  \n  \nIf there are no relevant quotes, write \"No relevant quotes\" instead.  \n  \nThen, answer the question, starting with \"Answer:\". Do not include or reference quoted content verbatim in the answer. Don't say \"According to Quote [1]\" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences.  \n  \nThus, the format of your overall response should look like what's shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly.  \n<example>  \nQuotes:  \n[1] \"Company X reported revenue of \$12 million in 2021.\"  \n[2] \"Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%.\"  \n  \nAnswer:  \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2]  \n</example>  \n  \nIf the question cannot be answered by the document, say so.",
  messages: [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Is Matterport doing well?"
        }
      ]
    }
  ]
});
console.log(msg);

from anthropic import AnthropicVertex

client = AnthropicVertex()

message = client.messages.create(
    model="claude-3-7-sonnet-v1@20250219",
    max_tokens=2000,
    temperature=0,
    system="You are an expert research assistant. Here is a document you will answer questions about:  \n<doc>  \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity]  \n</doc>  \n  \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short.  \n  \nIf there are no relevant quotes, write \"No relevant quotes\" instead.  \n  \nThen, answer the question, starting with \"Answer:\". Do not include or reference quoted content verbatim in the answer. Don't say \"According to Quote [1]\" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences.  \n  \nThus, the format of your overall response should look like what's shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly.  \n<example>  \nQuotes:  \n[1] \"Company X reported revenue of \$12 million in 2021.\"  \n[2] \"Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%.\"  \n  \nAnswer:  \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2]  \n</example>  \n  \nIf the question cannot be answered by the document, say so.",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "text",
                    "text": "Is Matterport doing well?"
                }
            ]
        }
    ]
)
print(message.content)

import { AnthropicVertex } from '@anthropic-ai/vertex-sdk';

// Reads from the `CLOUD_ML_REGION` & `ANTHROPIC_VERTEX_PROJECT_ID` environment variables.
// Additionally goes through the standard `google-auth-library` flow.
const client = new AnthropicVertex();

const msg = await client.messages.create({
  model: "claude-3-7-sonnet-v1@20250219",
  max_tokens: 2000,
  temperature: 0,
  system: "You are an expert research assistant. Here is a document you will answer questions about:  \n<doc>  \n[Full text of [Matterport SEC filing 10-K 2023](https://investors.matterport.com/node/9501/html), not pasted here for brevity]  \n</doc>  \n  \nFirst, find the quotes from the document that are most relevant to answering the question, and then print them in numbered order. Quotes should be relatively short.  \n  \nIf there are no relevant quotes, write \"No relevant quotes\" instead.  \n  \nThen, answer the question, starting with \"Answer:\". Do not include or reference quoted content verbatim in the answer. Don't say \"According to Quote [1]\" when answering. Instead make references to quotes relevant to each section of the answer solely by adding their bracketed numbers at the end of relevant sentences.  \n  \nThus, the format of your overall response should look like what's shown between the <example></example> tags. Make sure to follow the formatting and spacing exactly.  \n<example>  \nQuotes:  \n[1] \"Company X reported revenue of \$12 million in 2021.\"  \n[2] \"Almost 90% of revenue came from widget sales, with gadget sales making up the remaining 10%.\"  \n  \nAnswer:  \nCompany X earned \$12 million. [1] Almost 90% of it was from widget sales. [2]  \n</example>  \n  \nIf the question cannot be answered by the document, say so.",
  messages: [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Is Matterport doing well?"
        }
      ]
    }
  ]
});
console.log(msg);

Parceiro de narrativas Feiticeiro SQL

On this page

Exemplo de saída
Solicitação de API

Recursos

​Exemplo de saída

​Solicitação de API

Exemplo de saída

Solicitação de API