Synapse REST API

Perform a synchronous autocomplete search query against a SearchIndex.

Purpose-built for type-ahead input. The request shape is deliberately narrow: there is no size (the server caps every response at 8 hits), no from / search_after (a dropdown does not paginate), no sort (results are ordered by relevance), and no aggregations / responseParts (autocomplete returns matching hits only). For anything beyond a type-ahead lookup — scored relevance, faceting, totals, deep pagination — use the async POST /search/query/async/start endpoint with SearchIndexQuery instead.

Allowlisted top-level keys

Only query and _source are accepted on the searchQuery; any other top-level key returns HTTP 400 naming the offender.

query — required. The top-level clause must be one of prefix, match_phrase_prefix, or match_bool_prefix; any other clause type (including compound clauses such as bool) is rejected with HTTP 400. The same per-clause guarantees as the async search endpoint apply: scripts, cross-index references, and the wrapper form are rejected; depth, total-clause, and inline terms array length are capped; prefix values starting with * or ? are rejected (a leading wildcard forces a full inverted-index scan).

_source — optional. Source filter; same shape as on SearchQuery._source. Narrowing this to the columns the dropdown actually displays reduces response size, especially for wide indexes.

Field references

On text-typed columns (STRING, STRING_LIST, MEDIUMTEXT, LARGETEXT, LINK), use {columnName}.keyword for prefix and match_phrase_prefix (these are exact-match operations against the raw / non-tokenized sub-field). match_bool_prefix uses the bare column name (it analyzes the input). Numeric / boolean / keyword columns always use the bare name.

Example — prefix match on the title.keyword sub-field:

"searchQuery": { "query": { "prefix": { "title.keyword": "can" } } }

Resource URL

https://repo-prod.prod.sagebase.org/repo/v1/search/autocomplete