Key concepts to understand how Linkup works
/search
endpoint allows you to discover and access relevant web content based on a natural language query. Once the content is retrieved, it can serve as factual grounding for Large Language Models (LLMs), helping them produce more accurate and informed responses.
Initial Query | Improved Query | Explanation |
---|---|---|
What is the website of the company named Total? | What is the website of the French company named Total? | Adding the country (“French”) helps narrow down results, making it easier to identify the correct company. |
standard
: Returns results more quickly, suitable for low-latency scenarios. Costs €0.005 per call.deep
: Continues to search iteratively if it doesn’t find sufficient information on the first attempt. This may take longer, but often yields more comprehensive results. Costs €0.05 per call.deep
, the system repeatedly evaluates whether it has found enough information. If not, it continues searching—potentially for tens of seconds—until it locates what it needs.
sourcedAnswer
: Returns a natural language answer with source attributions.searchResults
: Provides chunks of contextual data suitable for grounding in LLM prompts.structured
: Produces a response following a specified JSON schema, ideal for structured data extraction.includeImages
: Boolean parameter to include relevant images in search resultsfromDate
: Filter results from a specific date (format: YYYY-MM-DD)toDate
: Filter results until a specific date (format: YYYY-MM-DD)includeDomains
: Filter results to only include specific domainsexcludeDomains
: Filter results to exclude specific domainsstandard
for quick, general queriesdeep
for complex research or when accuracy is criticalsourcedAnswer
for direct answers with citationssearchResults
for LLM groundingstructured
when you need specific data points returned in a json format. Read our structured output guide to know more.