exa-js

Our official Javscript SDK. Uses cross-fetch under the hood.

Note: This API is basically the same as metaphor-node but reflects new features associated with Metaphor's rename to Exa. New site is https://exa.ai

https://www.npmjs.com/package/exa-js

Installation

npm install exa-js

Initialization

import Exa from "exa-js";

const exa = new Exa(process.env.EXA_API_KEY);

Common commands

// Basic search
const basicResults = await exa.search("This is a Exa query:");

// Search with date filters
const dateFilteredResults = await exa.search("This is a Exa query:", {
  startPublishedDate: "2019-01-01",
  endPublishedDate: "2019-01-31",
});

// Search with domain filters
const domainFilteredResults = await exa.search("This is a Exa query:", {
  includeDomains: ["www.cnn.com", "www.nytimes.com"],
});

// Search and get text contents
const searchAndTextResults = await exa.searchAndContents(
  "This is a Exa query:",
  { text: true }
);

// Search and get contents with contents options
const searchAndCustomContentsResults = await exa.searchAndContents(
  "This is a Exa query:",
  {
    text: { maxCharacters: 3000 },
  }
);

// Find similar documents
const similarResults = await exa.findSimilar("https://example.com");

// Find similar excluding source domain
const similarExcludingSourceResults = await exa.findSimilar(
  "https://example.com",
  { excludeSourceDomain: true }
);

// Find similar with contents
const similarWithContentsResults = await exa.findSimilarAndContents(
  "https://example.com",
  { text: true }
);

// Get text contents
const textContentsResults = await exa.getContents(["urls"], { text: true });

// Get contents with contents options
const customContentsResults = await exa.getContents(["urls"], {
  text: { includeHtmlTags: true, maxCharacters: 3000 },
});

// Get an answer to a question
const answerResult = await exa.answer(
  "What is the population of New York City?"
);

// Get answer with citation contents and use the exa-pro model, which passes 2 extra queries to exa to increase coverage of the search space.
const answerWithTextResults = await exa.answer(
  "What is the population of New York City?",
  {
    text: true,
    model: "exa-pro",
  }
);

// Get an answer with streaming
for await (const chunk of exa.streamAnswer(
  "What is the population of New York City?"
)) {
  if (chunk.content) {
    process.stdout.write(chunk.content);
  }
  if (chunk.citations) {
    console.log("\nCitations:", chunk.citations);
  }
}

// Get an answer with output schema
const answerResult = await exa.answer(
  "What is the population of New York City?",
  {
    outputSchema: {
      type: "object",
      required: ["answer"],
      additionalProperties: false,
      properties: {
        answer: {
          type: "number",
        },
      },
    },
  }
);

exa.search(query: string, options?: SearchOptions): Promise<SearchResponse>

Performs a search on the Exa system with the given parameters.

const response = await exa.search("funny article about tech culture", {
  numResults: 5,
  includeDomains: ["nytimes.com", "wsj.com"],
  startPublishedDate: "2023-06-12",
});

exa.findSimilar(url: string, options?: FindSimilarOptions): Promise<SearchResponse>

Finds content similar to the specified URL.

const response = await exa.findSimilar(
  "https://waitbutwhy.com/2014/05/fermi-paradox.html",
  {
    numResults: 10,
  }
);

exa.getContents(urls: string[] | Result[]): Promise<GetContentsResponse>

Retrieves the contents of the specified documents.

const response = await exa.getContents([
  "https://blog.samaltman.com/how-to-be-successful",
]);

exa.answer(query: string, options?: AnswerOptions): Promise<AnswerResponse>

Generates an answer to a query using search results as context.

const response = await exa.answer("What is the population of New York City?", {
  text: true,
});

exa.streamAnswer(query: string, options?: { text?: boolean }): AsyncGenerator<AnswerStreamChunk>

Streams an answer as it's being generated, yielding chunks of text and citations. This is useful for providing real-time updates in chat interfaces or displaying partial results as they become available.

// Basic streaming example
for await (const chunk of exa.streamAnswer("What is quantum computing?")) {
  if (chunk.content) {
    process.stdout.write(chunk.content);
  }
  if (chunk.citations) {
    console.log("\nCitations:", chunk.citations);
  }
}

for await (const chunk of exa.streamAnswer("What is quantum computing?", {
  text: true,
})) {
}

Each chunk contains:

• content: A string containing the next piece of generated text
• citations: An array of citation objects containing source information

exa.research.createTask({ instructions, schema }: { instructions: string, output?: { schema?: object }}): Promise<{id: string}>

Exa's research agent can autonomously gather information and return a structured JSON object that conforms to a schema you provide.

import Exa from "exa-js";

const exa = new Exa(process.env.EXA_API_KEY);

const schema = {
  type: "object",
  required: ["answer"],
  properties: {
    answer: { type: "string" },
  },
  additionalProperties: false,
};

const { id: taskId } = await exa.research.createTask({
  instructions: "In ≤3 sentences, explain quantum computing.",
  output: { schema },
});
const result = await exa.research.pollTask(taskId);

Use the status field to poll long-running tasks if needed.

Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change.