Python SDK Specification

For ChatGPT-based Python SDK assistance, go here.

Getting started

Install the exa-py SDK

Bash
pip install exa_py

and then instantiate an Exa client

Python
from exa_py import Exa

import os

exa = Exa(os.getenv('EXA_API_KEY'))

Get API Key

Follow this link to get your API key

`search` Method

Perform an Exa search given an input query and retrieve a list of relevant results as links.

Input Example:

Python
`result = exa.search(
  "hottest AI startups",
  use_autoprompt=True,
  num_results=2
)

Input Parameters:

Parameter	Type	Description	Default
query	str	The input query string.	Required
num_results	Optional[int]	Number of search results to return.	10
include_domains	Optional[List[str]]	List of domains to include in the search.	None
exclude_domains	Optional[List[str]]	List of domains to exclude in the search.	None
start_crawl_date	Optional[str]	Results will only include links crawled after this date.	None
end_crawl_date	Optional[str]	Results will only include links crawled before this date.	None
start_published_date	Optional[str]	Results will only include links with a published date after this date.	None
end_published_date	Optional[str]	Results will only include links with a published date before this date.	None
use_autoprompt	Optional[bool]	If true, convert query to a query best suited for Exa.	False
type	Optional[str]	The type of search, keyword or neural.	”auto”
category	Optional[str]	A data category to focus on when searching, with higher comprehensivity and data cleanliness. Currently, the available categories are: company, research paper, news, linkedin profile, github, tweet, movie, song, personal site, pdf and financial report.	None

Returns Example:

JSON
{
  "autopromptString": "Here is a link to one of the hottest AI startups:",
  "results": [
    {
      "score": 0.17025552690029144,
      "title": "Adept: Useful General Intelligence",
      "id": "https://www.adept.ai/",
      "url": "https://www.adept.ai/",
      "publishedDate": "2000-01-01",
      "author": null
    },
    {
      "score": 0.1700288951396942,
      "title": "Home | Tenyx, Inc.",
      "id": "https://www.tenyx.com/",
      "url": "https://www.tenyx.com/",
      "publishedDate": "2019-09-10",
      "author": null
    }
  ],
  "requestId": "a78ebce717f4d712b6f8fe0d5d7753f8"
}

Return Parameters:

SearchResponse[Result]

Field	Type	Description
results	List[Result]	List of Result objects
autoprompt_string	Optional[str]	Exa query created by autoprompt functionality

Result Object:

Field	Type	Description
url	str	URL of the search result
id	str	Temporary ID for the document
title	Optional[str]	Title of the search result
score	Optional[float]	Similarity score between query/url and result
published_date	Optional[str]	Estimated creation date
author	Optional[str]	Author of the content, if available

`search_and_contents` Method

Perform an Exa search given an input query and retrieve a list of relevant results as links, optionally including the full text and/or highlights of the content.

Input Example:

Python
`# Search with full text content
result_with_text = exa.search_and_contents(
    "AI in healthcare",
    text=True,
    num_results=2
)

# Search with highlights
result_with_highlights = exa.search_and_contents(
    "AI in healthcare",
    highlights=True,
    num_results=2
)

# Search with both text and highlights
result_with_text_and_highlights = exa.search_and_contents(
    "AI in healthcare",
    text=True,
    highlights=True,
    num_results=2
)

Input Parameters:

Parameter	Type	Description	Default
query	str	The input query string.	Required
text	Union[TextContentsOptions, Literal[True]]	If provided, includes the full text of the content in the results.	None
highlights	Union[HighlightsContentsOptions, Literal[True]]	If provided, includes highlights of the content in the results.	None
num_results	Optional[int]	Number of search results to return.	10
include_domains	Optional[List[str]]	List of domains to include in the search.	None
exclude_domains	Optional[List[str]]	List of domains to exclude in the search.	None
start_crawl_date	Optional[str]	Results will only include links crawled after this date.	None
end_crawl_date	Optional[str]	Results will only include links crawled before this date.	None
start_published_date	Optional[str]	Results will only include links with a published date after this date.	None
end_published_date	Optional[str]	Results will only include links with a published date before this date.	None
use_autoprompt	Optional[bool]	If true, convert query to a query best suited for Exa.	False
type	Optional[str]	The type of search, keyword or neural.	”auto”
category	Optional[str]	A data category to focus on when searching, with higher comprehensivity and data cleanliness. Currently, the available categories are: company, research paper, news, linkedin profile, github, tweet, movie, song, personal site, pdf and financial report.	None

Returns Example:

JSON
`{
  "results": [
    {
      "score": 0.20826785266399384,
      "title": "2023 AI Trends in Health Care",
      "id": "https://aibusiness.com/verticals/2023-ai-trends-in-health-care-",
      "url": "https://aibusiness.com/verticals/2023-ai-trends-in-health-care-",
      "publishedDate": "2022-12-29",
      "author": "Wylie Wong",
      "text": "While the health care industry was initially slow to [... TRUNCATED IN THESE DOCS FOR BREVITY ...]",
      "highlights": [
        "But to do so, many health care institutions would like to share data, so they can build a more comprehensive dataset to use to train an AI model. Traditionally, they would have to move the data to one central repository. However, with federated or swarm learning, the data does not have to move. Instead, the AI model goes to each individual health care facility and trains on the data, he said. This way, health care providers can maintain security and governance over their data."
      ],
      "highlightScores": [
        0.5566554069519043
      ]
    },
    {
      "score": 0.20796334743499756,
      "title": "AI in healthcare: Innovative use cases and applications",
      "id": "https://www.leewayhertz.com/ai-use-cases-in-healthcare",
      "url": "https://www.leewayhertz.com/ai-use-cases-in-healthcare",
      "publishedDate": "2023-02-13",
      "author": "Akash Takyar",
      "text": "The integration of AI in healthcare is not [... TRUNCATED IN THESE DOCS FOR BREVITY ...]",
      "highlights": [
        "The ability of AI to analyze large amounts of medical data and identify patterns has led to more accurate and timely diagnoses. This has been especially helpful in identifying complex medical conditions, which may be difficult to detect using traditional methods. Here are some examples of successful implementation of AI in healthcare. IBM Watson Health: IBM Watson Health is an AI-powered system used in healthcare to improve patient care and outcomes. The system uses natural language processing and machine learning to analyze large amounts of data and provide personalized treatment plans for patients."
      ],
      "highlightScores": [
        0.6563674807548523
      ]
    }
  ],
  "requestId": "d8fd59c78d34afc9da173f1fe5aa8965"
}

Return Parameters:

The return type depends on the combination of text and highlights parameters:

SearchResponse[ResultWithText]: When only text is provided.
SearchResponse[ResultWithHighlights]: When only highlights is provided.
SearchResponse[ResultWithTextAndHighlights]: When both text and highlights are provided.

`SearchResponse\[ResultWithTextAndHighlights\]`

Field	Type	Description
results	List[ResultWithTextAndHighlights]	List of ResultWithTextAndHighlights objects
autoprompt_string	Optional[str]	Exa query created by autoprompt functionality

`ResultWithTextAndHighlights` Object

Field	Type	Description
url	str	URL of the search result
id	str	Temporary ID for the document
title	Optional[str]	Title of the search result
score	Optional[float]	Similarity score between query/url and result
published_date	Optional[str]	Estimated creation date
author	Optional[str]	Author of the content, if available
text	str	Text of the search result page (always present)
highlights	List[str]	Highlights of the search result (always present)
highlight_scores	List[float]	Scores of the highlights (always present)

Note: If neither text nor highlights is specified, the method defaults to including the full text content.

`find_similar` Method

Find a list of similar results based on a webpage’s URL.

Input Example:

Python
similar_results = exa.find_similar(
    "miniclip.com",
    num_results=2,
    exclude_source_domain=True
)

Input Parameters:

Parameter	Type	Description	Default
url	str	The URL of the webpage to find similar results for.	Required
num_results	Optional[int]	Number of similar results to return.	None
include_domains	Optional[List[str]]	List of domains to include in the search.	None
exclude_domains	Optional[List[str]]	List of domains to exclude from the search.	None
start_crawl_date	Optional[str]	Results will only include links crawled after this date.	None
end_crawl_date	Optional[str]	Results will only include links crawled before this date.	None
start_published_date	Optional[str]	Results will only include links with a published date after this date.	None
end_published_date	Optional[str]	Results will only include links with a published date before this date.	None
exclude_source_domain	Optional[bool]	If true, excludes results from the same domain as the input URL.	None
category	Optional[str]	A data category to focus on when searching, with higher comprehensivity and data cleanliness.	None

Returns Example:

JSON
{
  "results": [
    {
      "score": 0.8777582049369812,
      "title": "Play New Free Online Games Every Day",
      "id": "https://www.minigames.com/new-games",
      "url": "https://www.minigames.com/new-games",
      "publishedDate": "2000-01-01",
      "author": null
    },
    {
      "score": 0.87653648853302,
      "title": "Play The best Online Games",
      "id": "https://www.minigames.com/",
      "url": "https://www.minigames.com/",
      "publishedDate": "2000-01-01",
      "author": null
    }
  ],
  "requestId": "08fdc6f20e9f3ea87f860af3f6ccc30f"
}

Return Parameters:

SearchResponse[_Result]: The response containing similar results and optional autoprompt string.

`SearchResponse\[Results\]`

Field	Type	Description
results	List[ResultWithTextAndHighlights]	List of ResultWithTextAndHighlights objects
autoprompt_string	Optional[str]	Exa query created by autoprompt functionality

`Results` Object

Field	Type	Description
url	str	URL of the search result
id	str	Temporary ID for the document
title	Optional[str]	Title of the search result
score	Optional[float]	Similarity score between query/url and result
published_date	Optional[str]	Estimated creation date
author	Optional[str]	Author of the content, if available

`find_similar_and_contents` Method

Find a list of similar results based on a webpage’s URL, optionally including the text content or highlights of each result.

Input Example:

Python
# Find similar with full text content
similar_with_text = exa.find_similar_and_contents(
    "https://example.com/article",
    text=True,
    num_results=2
)

# Find similar with highlights
similar_with_highlights = exa.find_similar_and_contents(
    "https://example.com/article",
    highlights=True,
    num_results=2
)

# Find similar with both text and highlights
similar_with_text_and_highlights = exa.find_similar_and_contents(
    "https://example.com/article",
    text=True,
    highlights=True,
    num_results=2
)

Input Parameters:

Parameter	Type	Description	Default
url	str	The URL of the webpage to find similar results for.	Required
text	Union[TextContentsOptions, Literal[True]]	If provided, includes the full text of the content in the results.	None
highlights	Union[HighlightsContentsOptions, Literal[True]]	If provided, includes highlights of the content in the results.	None
num_results	Optional[int]	Number of similar results to return.	None
include_domains	Optional[List[str]]	List of domains to include in the search.	None
exclude_domains	Optional[List[str]]	List of domains to exclude from the search.	None
start_crawl_date	Optional[str]	Results will only include links crawled after this date.	None
end_crawl_date	Optional[str]	Results will only include links crawled before this date.	None
start_published_date	Optional[str]	Results will only include links with a published date after this date.	None
end_published_date	Optional[str]	Results will only include links with a published date before this date.	None
exclude_source_domain	Optional[bool]	If true, excludes results from the same domain as the input URL.	None
category	Optional[str]	A data category to focus on when searching, with higher comprehensivity and data cleanliness.	None

Returns:

The return type depends on the combination of text and highlights parameters:

SearchResponse[ResultWithText]: When only text is provided or when neither text nor highlights is provided (defaults to including text).
SearchResponse[ResultWithHighlights]: When only highlights is provided.
SearchResponse[ResultWithTextAndHighlights]: When both text and highlights are provided.

The response contains similar results and an optional autoprompt string.

Note: If neither text nor highlights is specified, the method defaults to including the full text content.

`answer` Method

Generate an answer to a query using Exa’s search and LLM capabilities. This method returns an AnswerResponse with the answer and a list of citations. You can optionally retrieve the full text of each citation by setting text=True.

Input Example:

Python
response = exa.answer("What is the capital of France?")

print(response.answer)       # e.g. "Paris"
print(response.citations)    # list of citations used

# If you want the full text of the citations in the response:
response_with_text = exa.answer(
    "What is the capital of France?",
    text=True
)
print(response_with_text.citations[0].text)  # Full page text

Input Parameters:

Parameter	Type	Description	Default
query	str	The question to answer.	Required
text	Optional[bool]	If true, the full text of each citation is included in the result.	False
stream	Optional[bool]	Note: If true, an error is thrown. Use stream_answer() instead for streaming responses.	None

Returns Example:

JSON
{
  "answer": "The capital of France is Paris.",
  "citations": [
    {
      "id": "https://www.example.com/france",
      "url": "https://www.example.com/france",
      "title": "France - Wikipedia",
      "publishedDate": "2023-01-01",
      "author": null,
      "text": "France, officially the French Republic, is a country in... [truncated for brevity]"
    }
  ]
}

Return Parameters:

Returns an AnswerResponse object:

Field	Type	Description
answer	str	The generated answer text
citations	List[AnswerResult]	List of citations used to generate the answer

`AnswerResult` object

Field	Type	Description
id	str	Temporary ID for the document
url	str	URL of the citation
title	Optional[str]	Title of the content, if available
published_date	Optional[str]	Estimated creation date
author	Optional[str]	The author of the content, if available
text	Optional[str]	The full text of the content (if text=True)

`stream_answer` Method

Generate a streaming answer to a query with Exa’s LLM capabilities. Instead of returning a single response, this method yields chunks of text and/or citations as they become available.

Input Example:

Python
stream = exa.stream_answer("What is the capital of France?", text=True)

for chunk in stream:
    if chunk.content:
        print("Partial answer:", chunk.content)
    if chunk.citations:
        for citation in chunk.citations:
            print("Citation found:", citation.url)

Input Parameters:

Parameter	Type	Description	Default
query	str	The question to answer.	Required
text	Optional[bool]	If true, includes full text of each citation in the streamed response.	False

Return Type:

A StreamAnswerResponse object, which is iterable. Iterating over it yields StreamChunk objects:

`StreamChunk`

Field	Type	Description
content	Optional[str]	Partial text content of the answer so far.
citations	Optional[List[AnswerResult]]	Citations discovered in this chunk, if any.

Use stream.close() to end the streaming session if needed.

SDKs

​Getting started

Get API Key

​search Method

​Input Example:

​Input Parameters:

​Returns Example:

​Return Parameters:

​Result Object:

​search_and_contents Method

​Input Example:

​Input Parameters:

​Returns Example:

​Return Parameters:

​SearchResponse\[ResultWithTextAndHighlights\]

​ResultWithTextAndHighlights Object

​find_similar Method

​Input Example:

​Input Parameters:

​Returns Example:

​Return Parameters:

​SearchResponse\[Results\]

​Results Object

​find_similar_and_contents Method

​Input Example:

​Input Parameters:

​Returns:

​answer Method

​Input Example:

​Input Parameters:

​Returns Example:

​Return Parameters:

​AnswerResult object

​stream_answer Method

​Input Example:

​Input Parameters:

​Return Type:

​StreamChunk

Getting started

`search` Method

Input Example:

Input Parameters:

Returns Example:

Return Parameters:

Result Object:

`search_and_contents` Method

Input Example:

Input Parameters:

Returns Example:

Return Parameters:

`SearchResponse\[ResultWithTextAndHighlights\]`

`ResultWithTextAndHighlights` Object

`find_similar` Method

Input Example:

Input Parameters:

Returns Example:

Return Parameters:

`SearchResponse\[Results\]`

`Results` Object

`find_similar_and_contents` Method

Input Example:

Input Parameters:

Returns:

`answer` Method

Input Example:

Input Parameters:

Returns Example:

Return Parameters:

`AnswerResult` object

`stream_answer` Method

Input Example:

Input Parameters:

Return Type:

`StreamChunk`