Semantic search over a space's indexed files

GET

/spaces/:spaceId/search

GET

/spaces/:spaceId/search

$ curl -G https://api.cloudraker.com/spaces/spaceId/search \
>      -H "Authorization: Bearer <token>" \
>      -H "Content-Type: application/json" \
>      --data-urlencode "q=Find the latest quarterly financial report" \
>      -d limit=10

200Retrieved

1 {
2   "matches": [
3     {
4       "id": "chunk_9f8e7d6c5b4a3210",
5       "namespace": "space/a1b2c3d4e5f6g7h8i9j0/file_01J9KLMNOPQ",
6       "tags": [
7         "finance",
8         "quarterly-report",
9         "2024-Q1"
10       ],
11       "score": 0.987,
12       "metadata": {
13         "fileId": "file_01J9KLMNOPQ",
14         "page": 12,
15         "bbox": [
16           100,
17           150,
18           400,
19           300
20         ]
21       }
22     },
23     {
24       "id": "chunk_4a3b2c1d0e9f8g7h",
25       "namespace": "space/a1b2c3d4e5f6g7h8i9j0/file_02XYZABCDE",
26       "tags": [
27         "finance",
28         "summary"
29       ],
30       "score": 0.876,
31       "metadata": {
32         "fileId": "file_02XYZABCDE",
33         "page": 3
34       }
35     }
36   ]
37 }

Runs semantic search over every indexed file in the space :spaceId. Authorization: the caller needs space:read on the space; organization administrators bypass this check. Lacking access returns 404 (existence is hidden), never 403. Query params: q (required) is the natural-language query text; callers never send vectors. An empty or whitespace-only q returns 400. limit (optional, default 20) is a positive integer, capped at 100; larger values are clamped to 100, and a non-integer or < 1 returns 400. The search is always scoped to the space in the path. Returns { matches }, an array ranked by relevance; each match has id, namespace, tags[], score (higher = closer), and a metadata object carrying file grounding (file id, page, bbox, and/or audio timecode) when the source was indexed with it. There is no cursor pagination — bound the result count with limit.

Runs semantic search over every indexed file in the space `:spaceId`. Authorization: the caller needs `space:read` on the space; organization administrators bypass this check. Lacking access returns **404** (existence is hidden), never 403. Query params: `q` (required) is the natural-language query text; callers never send vectors. An empty or whitespace-only `q` returns 400. `limit` (optional, default 20) is a positive integer, capped at 100; larger values are clamped to 100, and a non-integer or `< 1` returns 400. The search is always scoped to the space in the path. Returns `{ matches }`, an array ranked by relevance; each match has `id`, `namespace`, `tags[]`, `score` (higher = closer), and a `metadata` object carrying file grounding (file id, page, bbox, and/or audio timecode) when the source was indexed with it. There is no cursor pagination — bound the result count with `limit`.

Authentication

AuthorizationBearer

Bearer authentication of the form Bearer <token>, where token is your auth token.

Path parameters

spaceIdstringRequired

Query parameters

qstringRequired>=1 character

Natural-language query text. Empty or whitespace-only returns 400. Callers never send vectors.

limitintegerOptional1-100Defaults to 20

Maximum matches. Defaults to 20, clamped to 100. A non-integer or value below 1 returns 400.

Response

Ranked semantic matches within the space.

matcheslist of objects

Ranked by relevance. No cursor.