Skip to main content
Agentic Search returns Knowledge, Episodes, Topics, Entities, and relations in one mixed result list.

​
Overview

const client = new NdxClient({
  apiKey: process.env.NEURADEX_API_KEY,
  projectId: 'your-project-id',
});

const result = await client.search.agentic('authentication discussion', {
  limit: 20,
  includeTypes: ['knowledge', 'episode', 'topic', 'entity', 'relation'],
});

​
Methods

MethodDescription
agentic(query, options?)Unified cross-layer search (single array)

​
agentic()

Searches across multiple knowledge layers and returns a single mixed array. Filter by type as needed for your UI or workflow.

​
Parameters

​
query
string
required
Natural language search query
​
options.limit
number
Max total results
​
options.perTypeLimit
number
Max results per type
​
options.includeTypes
string[]
Types to include (“knowledge” | “episode” | “topic” | “entity” | “relation”)
​
options.expandGraph
boolean
Include broader related context
​
options.maxHops
number
Depth of related context
​
options.expandLimit
number
Seed count used to broaden results
​
options.relationLimit
number
Max number of relations

​
Returns

interface AgenticSearchResponse {
  results: AgenticSearchResult[];
  meta: {
    query: string;
    projectId: string;
    usedTypes: AgenticSearchType[];
    expandGraph: boolean;
    counts: Record<AgenticSearchType, number>;
  };
}

​
Example

const result = await client.search.agentic('deploy failures', {
  limit: 15,
  includeTypes: ['episode', 'knowledge', 'relation'],
});

const episodes = result.results.filter((item) => item.type === 'episode');
const knowledges = result.results.filter((item) => item.type === 'knowledge');

​
Next Steps

Knowledge API

Fetch details and updates for knowledge

MCP Tools

Use Agentic Search via MCP