Search

GET /search searches products, knowledge documents and contacts in one request and returns one ranked list. It is built for a search box in a dashboard.

curl -G https://api.getengageai.app/search \
  -H "Authorization: Bearer $ENGAGE_API_KEY" \
  --data-urlencode "q=ridgelne"

{
  "query": "ridgelne",
  "searched": ["PRODUCT", "DOCUMENT", "CONTACT"],
  "counts": { "PRODUCT": 1 },
  "items": [
    {
      "type": "PRODUCT",
      "id": "…",
      "title": "Ridgeline 2P Tent",
      "subtitle": "TENT-RIDGE-2P",
      "excerpt": "A two-person, three-season backpacking tent.",
      "score": 0.67,
      "updatedAt": "2026-10-12T08:15:00.000Z",
      "groupKey": null
    }
  ]
}

The query has a typo, and the product is still found. Matching uses trigrams, so it tolerates misspellings and partial words.

Parameters

Each item carries its type. Render whatever comes back based on that, because more types will be added over time without a new endpoint. counts gives the number of hits per type before limit was applied, for labels like "3 products, 2 documents". searched lists the types that were searched.

Search is not pagination

Results are a ranking, not a list to page through. limit is a ceiling. If the result you want is not near the top, refine the query.

Search versus retrieval

GET /search matches spelling. It is for people looking for a record.

The reply engine uses a different search that matches meaning as well. To see what the AI would find for a customer's question, use POST /knowledge/preview instead. "Can I send these back?" finds the returns policy through preview, but probably not through /search.

Reliability

If one kind of record cannot be searched for a moment, its results are left out and the rest still come back. A broken index does not break the search box.