Tavily

Tavily enriches LLM prompts with fresh web search context before inference.

Tavily provides a search API purpose-built for AI agents. Portkey’s Tavily guardrail injects relevant web search results into the request before it reaches the LLM, so the model can answer with current online information.

Get Started with Tavily

Requires Backend v1.16.0+.

Using Tavily with Portkey

1. Add Tavily Credentials to Portkey

• Click on the Admin Settings button on Sidebar
• Navigate to Plugins tab under Organisation Settings
• Click on the edit button for the Tavily integration
• Add your Tavily API Key (obtain this from your Tavily account)

2. Add Tavily’s Guardrail Check

• Navigate to the Guardrails page and click the Create button
• Search for “Tavily Online Search” and click Add
• Configure the search behaviour (all parameters are optional):

Parameter	Type	Description	Default
prefix	string	Text inserted before the Tavily search context in the prompt	\n<web_search_context>
suffix	string	Text inserted after the Tavily search context in the prompt	\n</web_search_context>
maxResults	number	Maximum number of Tavily search results to return	5
searchDepth	enum	Latency vs. relevance tradeoff: advanced, basic, fast, ultra-fast	—
chunksPerSource	number	Maximum number of chunks returned per source (advanced/fast depths)	3
topic	enum	Category of results: general, news, finance	—
timeRange	enum	Filter by recency: day / week / month / year (or d / w / m / y)	—
startDate	string	Return results after this date (YYYY-MM-DD)	—
endDate	string	Return results before this date (YYYY-MM-DD)	—
includeAnswer	boolean | basic | advanced	Include a Tavily-generated answer in the response	false
includeRawContent	boolean | markdown | text	Include cleaned, parsed HTML content for each result	false
includeImages	boolean	Include query-related images and per-result images	false
includeImageDescriptions	boolean	Include descriptive text for each image	false
includeFavicon	boolean	Include the favicon URL for each result	false
includeDomains	string[]	Restrict results to the listed domains	—
excludeDomains	string[]	Exclude results from the listed domains	—
country	string	Boost results from a specific country (only when topic = general)	—
autoParameters	boolean	Let Tavily automatically configure search parameters from the query	false
exactMatch	boolean	Return only results matching exact quoted phrases	false
includeUsage	boolean	Include Tavily credit usage information in the response metadata	false
safeSearch	boolean	Enterprise-only. Filters adult or unsafe content. Not supported for fast/ultra-fast depths.	false
timeout	number	Maximum time to wait for the Tavily request (ms)	30000

• Set any actions you want on your check, and create the Guardrail!

Guardrail Actions allow you to orchestrate your guardrails logic. You can learn more about them here

Check Name	Description	Supported Hooks
Tavily Online Search	Adds relevant Tavily web results to the prompt so the model can answer with current online information	beforeRequestHook

3. Add Guardrail ID to a Config and Make Your Request

• When you save a Guardrail, you’ll get an associated Guardrail ID — add this ID to the input_guardrails param in your Portkey Config
• Create these Configs in Portkey UI, save them, and get an associated Config ID to attach to your requests. More here.

{
  "input_guardrails": ["guardrails-id-xxx"]
}

NodeJS

const portkey = new Portkey({
    apiKey: "PORTKEY_API_KEY",
    config: "pc-***"
});

Python

portkey = Portkey(
    api_key="PORTKEY_API_KEY",
    config="pc-***"
)

cURL

curl https://api.portkey.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "x-portkey-api-key: $PORTKEY_API_KEY" \
  -H "x-portkey-config: $CONFIG_ID" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "What are the latest changes to the EU AI Act?"}]
  }'

For more, refer to the Config documentation.

Get Support

If you face any issues with the Tavily integration, join the Portkey community forum for assistance.