Search & Memory Recall

Skopik features built-in search and knowledge retrieval across all platform resources. Rather than managing third-party vector databases or custom chunking pipelines, Skopik automatically indexes Agent brain files, durable memories, Session transcripts, and final deliverables into a unified search index.

Automatic zero-config indexing

Indexing is fully implicit:

• Writing or updating a file via the Files API automatically queues background chunking, embedding generation, and lexical tokenization.
• Deleting a file automatically removes its associated chunks from search.
• Closing or updating Session transcripts automatically indexes new turn dialogue.

No manual indexing endpoints or ETL pipelines are needed.

Search modes

Target-scoped search

Query specific resource domains using targets:

const response = await skopik.search.search({
	query: 'rate limiting architecture in enterprise API gateway',
	targets: ['documents', 'memories'],
	mode: 'hybrid',
	limit: 10,
	highlight: true,
})

// Access results grouped by target
const docHits = response.results.documents?.hits ?? []
const memoryHits = response.results.memories?.hits ?? []

for (const hit of docHits) {
	console.log(`Document: ${hit.path} (Score: ${hit.score})`)
	console.log(`  Snippet: ${hit.text}`)
	console.log(`  Highlights:`, hit.highlights)
}

curl "https://api.skopik.com/api/v1/search" \
  -H "Authorization: Bearer $SKOPIK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "rate limiting architecture in enterprise API gateway",
    "targets": ["documents", "memories"],
    "mode": "hybrid",
    "limit": 10,
    "highlight": true
  }'

Search target descriptions

Memory recall with context positioning

When searching Agent memories, pass a context object containing agentId or sessionId. Skopik automatically traverses the full organizational hierarchy (session → agent → user → org) to rank memories most relevant to that specific execution context:

const response = await skopik.search.search({
	query: 'user preferred coding style and framework conventions',
	targets: ['memories'],
	context: {
		agentId: agent.agentId,
		sessionId: session.sessionId,
	},
	limit: 5,
})

for (const memory of response.results.memories?.hits ?? []) {
	console.log(`[Memory] ${memory.path}: ${memory.text}`)
}

Reference search for precise code blocks

Use skopik.search.references() to find exact code symbols, configuration keys, or policy references across files with character-level offsets and line ranges:

const response = await skopik.search.references({
	query: 'export function createClientTool',
	mode: 'indexed',
	caseSensitive: true,
	wholeWord: false,
	limit: 20,
})

for (const hit of response.hits) {
	console.log(`File: ${hit.path}`)
	console.log(`  Lines: ${hit.headingPath?.join(' > ') ?? 'root'}`)
	console.log(`  Character Range: ${hit.matchStart} - ${hit.matchEnd}`)
	console.log(`  Snippet:\n${hit.snippet}`)
}

curl "https://api.skopik.com/api/v1/search/references" \
  -H "Authorization: Bearer $SKOPIK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "export function createClientTool",
    "mode": "indexed",
    "caseSensitive": true,
    "limit": 20
  }'

Filter scopes

Filter searches to specific agents, session workspaces, or folder paths:

const response = await skopik.search.search({
	query: 'database migration policy',
	filters: {
		scope: 'agent',
		agentId: agent.agentId,
		pathPrefix: 'references/database/',
	},
	limit: 10,
})

API reference