Tavily

Tavily

Tavily é uma API de busca potencializada por IA, projetada especificamente para aplicações LLM. Ela oferece capacidades confiáveis de recuperação de informações em tempo real, com recursos otimizados para casos de uso de IA, incluindo busca semântica, extração de conteúdo e recuperação de dados estruturados.

Com o Tavily, você pode:

  • Realizar buscas contextuais: Obtenha resultados relevantes com base em compreensão semântica, e não apenas em correspondência de palavras-chave
  • Extrair conteúdo estruturado: Extraia informações específicas de páginas da web em um formato limpo e utilizável
  • Acessar informações em tempo real: Recupere dados atualizados de toda a web
  • Processar múltiplas URLs simultaneamente: Extraia conteúdo de várias páginas da web em uma única solicitação
  • Receber resultados otimizados para IA: Obtenha resultados de busca formatados especificamente para consumo por sistemas de IA

No Zoen, a integração com o Tavily permite que seus agentes pesquisem a web e extraiam informações como parte dos seus workflows. Isso possibilita cenários sofisticados de automação que exigem informações atualizadas da internet. Seus agentes podem formular consultas de busca, recuperar resultados relevantes e extrair conteúdo de páginas específicas da web para informar seus processos de decisão. Esta integração faz a ponte entre a automação do seu workflow e o vasto conhecimento disponível na web, permitindo que seus agentes acessem informações em tempo real sem intervenção manual. Ao conectar o Zoen ao Tavily, você pode criar agentes que se mantêm atualizados com as informações mais recentes, fornecem respostas mais precisas e entregam mais valor aos usuários.

Usage Instructions

Integrate Tavily into the workflow. Can search the web and extract content from specific URLs. Requires API Key.

Actions

Perform AI-powered web searches using Tavily's search API. Returns structured results with titles, URLs, snippets, and optional raw content, optimized for relevance and accuracy.

Input

ParameterTypeRequiredDescription
querystringYesThe search query to execute (e.g., "latest AI research papers 2024")
max_resultsnumberNoMaximum number of results (1-20, e.g., 5)
topicstringNoCategory type: general, news, or finance (e.g., "news")
search_depthstringNoSearch scope: basic (1 credit) or advanced (2 credits) (e.g., "advanced")
include_answerstringNoLLM-generated response: true/basic for quick answer or advanced for detailed (e.g., "advanced")
include_raw_contentstringNoParsed HTML content: true/markdown or text format (e.g., "markdown")
include_imagesbooleanNoInclude image search results
include_image_descriptionsbooleanNoAdd descriptive text for images
include_faviconbooleanNoInclude favicon URLs
chunks_per_sourcenumberNoMaximum number of relevant chunks per source (1-3, default: 3)
time_rangestringNoFilter by recency: day/d, week/w, month/m, year/y
start_datestringNoEarliest publication date (YYYY-MM-DD format)
end_datestringNoLatest publication date (YYYY-MM-DD format)
include_domainsstringNoComma-separated list of domains to whitelist (e.g., "github.com,stackoverflow.com")
exclude_domainsstringNoComma-separated list of domains to blacklist (e.g., "pinterest.com,reddit.com")
countrystringNoBoost results from specified country (general topic only)
auto_parametersbooleanNoAutomatic parameter configuration based on query intent
apiKeystringYesTavily API Key

Output

ParameterTypeDescription
querystringThe search query that was executed
resultsarrayRanked search results with titles, URLs, content snippets, and optional metadata
titlestringResult title
urlstringResult URL
contentstringBrief description or content snippet
scorenumberRelevance score
raw_contentstringFull parsed HTML content (if requested)
faviconstringFavicon URL for the domain
answerstringLLM-generated answer to the query (if requested)
imagesarrayQuery-related images (if requested)
urlstringImage URL
descriptionstringImage description
auto_parametersobjectAutomatically selected parameters based on query intent (if enabled)
response_timenumberTime taken for the search request in seconds

tavily_extract

Extract raw content from multiple web pages simultaneously using Tavily's extraction API. Supports basic and advanced extraction depths with detailed error reporting for failed URLs.

Input

ParameterTypeRequiredDescription
urlsstringYesURL or array of URLs to extract content from
extract_depthstringNoThe depth of extraction (basic=1 credit/5 URLs, advanced=2 credits/5 URLs)
formatstringNoOutput format: markdown or text (default: markdown)
include_imagesbooleanNoIncorporate images in extraction output
include_faviconbooleanNoAdd favicon URL for each result
apiKeystringYesTavily API Key

Output

ParameterTypeDescription
resultsarraySuccessfully extracted content from URLs
urlstringThe source URL
raw_contentstringFull extracted content from the page
imagesarrayImage URLs (when include_images is true)
faviconstringFavicon URL for the result
failed_resultsarrayURLs that failed to extract content
urlstringThe URL that failed extraction
errorstringError message describing why extraction failed
response_timenumberTime taken for the extraction request in seconds

tavily_crawl

Systematically crawl and extract content from websites using Tavily's crawl API. Supports depth control, path filtering, domain restrictions, and natural language instructions for targeted crawling.

Input

ParameterTypeRequiredDescription
urlstringYesThe root URL to begin the crawl
instructionsstringNoNatural language directions for the crawler (costs 2 credits per 10 pages)
max_depthnumberNoHow far from base URL to explore (1-5, default: 1)
max_breadthnumberNoLinks followed per page level (≥1, default: 20)
limitnumberNoTotal links processed before stopping (≥1, default: 50)
select_pathsstringNoComma-separated regex patterns to include specific URL paths (e.g., /docs/.*)
select_domainsstringNoComma-separated regex patterns to restrict crawling to certain domains
exclude_pathsstringNoComma-separated regex patterns to skip specific URL paths
exclude_domainsstringNoComma-separated regex patterns to block certain domains
allow_externalbooleanNoInclude external domain links in results (default: true)
include_imagesbooleanNoIncorporate images in crawl output
extract_depthstringNoExtraction depth: basic (1 credit/5 pages) or advanced (2 credits/5 pages)
formatstringNoOutput format: markdown or text (default: markdown)
include_faviconbooleanNoAdd favicon URL for each result
apiKeystringYesTavily API Key

Output

ParameterTypeDescription
base_urlstringThe base URL that was crawled
resultsarrayArray of crawled pages with extracted content
urlstringThe crawled page URL
raw_contentstringFull extracted page content
faviconstringFavicon URL for the result
response_timenumberTime taken for the crawl request in seconds
request_idstringUnique identifier for support reference

tavily_map

Discover and visualize website structure using Tavily's map API. Maps out all accessible URLs from a base URL with depth control, path filtering, and domain restrictions.

Input

ParameterTypeRequiredDescription
urlstringYesThe root URL to begin mapping
instructionsstringNoNatural language guidance for mapping behavior (costs 2 credits per 10 pages)
max_depthnumberNoHow far from base URL to explore (1-5, default: 1)
max_breadthnumberNoLinks to follow per level (default: 20)
limitnumberNoTotal links to process (default: 50)
select_pathsstringNoComma-separated regex patterns for URL path filtering (e.g., /docs/.*)
select_domainsstringNoComma-separated regex patterns to restrict mapping to specific domains
exclude_pathsstringNoComma-separated regex patterns to exclude specific URL paths
exclude_domainsstringNoComma-separated regex patterns to exclude domains
allow_externalbooleanNoInclude external domain links in results (default: true)
apiKeystringYesTavily API Key

Output

ParameterTypeDescription
base_urlstringThe base URL that was mapped
resultsarrayArray of discovered URLs during mapping
urlstringDiscovered URL
response_timenumberTime taken for the map request in seconds
request_idstringUnique identifier for support reference

On this page