Astrolabe icon

Astrolabe

Sovereign EU AI for the social economy — chat, embeddings and knowledge bases (OpenAI-compatible API)

Actions16

Knowledge Base → Search

AI-generated

Summary

Searches a selected knowledge base semantically when a query is supplied, or lists and filters its contents without embedding cost when the query is blank. Supports date and facet filters, sorting, grouping, pagination, graph expansion, and numeric aggregation.

Inputs

  • Knowledge Base Name or ID (required) — Select the base to search, or provide its slug or ID using an expression.
  • Query — Semantic search text. Leave blank for listing and filtering without scores or embedding cost; graph expansion requires a query.
  • Aggregate (Numeric Facet, JSON) — Numeric facet name or JSON array of names to aggregate server-side. Available in analytic mode with Query left blank.
  • Date Range From / To — Include only chunks whose business date is on or after the start date and/or on or before the end date (YYYY-MM-DD).
  • Facets Equality (JSON) — JSON object of exact-match metadata filters, such as {"theme":"billing"}.
  • Filters (JSON) — JSON array of typed facet comparisons using eq, neq, gt, gte, lt, lte, in, or contains.
  • Graph Expansion (JSON) — With a query, include chunks linked to semantic results by facets. Configure seed count, facet keys, match mode, depth, and maximum chunks.
  • Max Results (Top K) — Maximum results, default 8 and capped at 50.
  • Offset (Pagination) — Number of results to skip, useful for paging through listings.
  • Return (Group By) — Return one result per chunk, group excerpts by document, or return one result per business date. Defaults to chunks.
  • Sort By — Sort by relevance, business date, creation date, or document name. Relevance requires a query; without one it falls back to business date descending.
  • Sort Direction — Sort ascending or descending; defaults to descending.

Output shape

A single n8n item containing search metadata and a results array.

The item includes group_by, total, truncated, distinct_documents, aggregates, and results; unavailable metadata is returned as null and results defaults to an empty array. Aggregates contain field, count, sum, avg, min, and max values when requested.

Links

Discussion