Skip to main content

​
Search

Search is the product area for fast, single-call web search. You send a query and get back a ranked list of results immediately — no session to create, no polling, no synthesis step. It’s the lightweight counterpart to AI Research.

​
What Search is for

Use Search when you need:
  • ranked web results for a query, not an analyzed answer
  • the fastest possible turnaround for a single lookup
  • a building block for your own research or ranking logic
  • a cheap way to check what’s publicly indexed about a topic
Common examples:
  • pull search results to display alongside other data
  • feed raw results into your own downstream processing
  • quick fact-checking or discovery before a deeper lookup
  • populate a “related links” or “sources” panel

​
How it works

Search is synchronous: you call the endpoint and get the result in the same response.
  1. You send a query and optional settings
  2. NeuralVerge runs the search
  3. You get ranked results back immediately
There is no session_id and nothing to poll.

​
Inputs

The Search endpoint accepts:
  • query — the search query, in plain text
  • settings.country — bias results toward a country
  • settings.language — bias results toward a language
  • settings.max_results — cap the number of results returned

​
Outputs

Search results include:
  • a ranked list of results, each with title, url, and snippet
  • a session_id for record-keeping, even though there is nothing to poll
  • usage metadata
Choose Search when:
  • you already know what to look for and just need the results
  • you don’t need cross-source synthesis or a written summary
  • latency and cost matter more than depth
Choose AI Research instead when the task requires reasoning across multiple sources and a structured final answer, not just a list of links. In the API Reference, Search includes the endpoint for running a search.

​
Typical product pattern

A common integration pattern is:
  • call Search directly from your backend or edge function
  • render title / url / snippet for each result
  • optionally feed the results into your own ranking or extraction step