Semantic search over feedback records

client.feedbackRecords.search.performSemanticSearch(, ?): SearchPerformSemanticSearchResponse { data, limit, next_cursor }

POST/v1/feedback-records/search/semantic

Embeds the search query and returns feedback record IDs with similarity scores (cosine, 0..1). Only available when embeddings are configured (EMBEDDING_PROVIDER and EMBEDDING_MODEL set). Supported providers: openai, google (Gemini Developer API / Google AI Studio), google-gemini (Gemini Enterprise Agent Platform API). When embeddings are disabled, this endpoint returns 503 Service Unavailable. Request body must include query and tenant_id (required for tenant isolation).

ParametersExpand Collapse

params: SearchPerformSemanticSearchParams { query, tenant_id, cursor, 2 more }

query: string

Body param: Search query text (embedded and compared via cosine similarity)

minLength1

tenant_id: string

Body param: Tenant ID (required for isolation; must match feedback record tenant_id)

minLength1

cursor?: string

Query param: Omit for the first page. For the next page, use the exact value from the previous response's next_cursor. Opaque (base64-encoded); keyset pagination.

limit?: number

Query param: Number of results to return (default 10, max 100). Consistent with list endpoints.

minimum1

maximum100

min_score?: number

Query param: Minimum similarity score (0..1); only results with score >= min_score are returned. Default 0.7 to reduce noise.

formatfloat

minimum0

maximum1

ReturnsExpand Collapse

SearchPerformSemanticSearchResponse { data, limit, next_cursor }

data: Array<Data>

List of feedback record IDs with similarity scores (0 = unrelated, 1 = identical). Consistent with list endpoints.

feedback_record_id: string

Feedback record UUID

formatuuid

field_label: string

Label of the feedback field (included in embedding for context). May be empty if the source record had no label.

score: number

Similarity score (0..1)

formatdouble

value_text: string

value_text of the feedback record (the text that was embedded). May be empty if the source had no text; embeddings are only created for records with non-empty value_text, but the field can be cleared after embedding creation.

limit: number

Limit used in query (echoed for consistency with list endpoints)

next_cursor?: string

Opaque cursor for the next page (keyset paging). Present only when there may be more results (full page returned). Omit when no next page. Use this exact value as the cursor query param for the next page.

Semantic search over feedback records

import FormbricksHub from '@formbricks/hub';

const client = new FormbricksHub({
  apiKey: process.env['HUB_API_KEY'], // This is the default and can be omitted
});

const response = await client.feedbackRecords.search.performSemanticSearch({
  query: 'What do users think about login speed?',
  tenant_id: 'org-123',
});

console.log(response.data);

{
  "data": [
    {
      "feedback_record_id": "182bd5e5-6e1a-4fe4-a799-aa6d9a6ab26e",
      "field_label": "field_label",
      "score": 0,
      "value_text": "value_text"
    }
  ],
  "limit": 0,
  "next_cursor": "eyJkIjowLjEsImkiOiIwMThlMTIzNC01Njc4LTlhYmMtZGVmMC0xMTExMTExMTExMTEifQ=="
}

Returns Examples

{
  "data": [
    {
      "feedback_record_id": "182bd5e5-6e1a-4fe4-a799-aa6d9a6ab26e",
      "field_label": "field_label",
      "score": 0,
      "value_text": "value_text"
    }
  ],
  "limit": 0,
  "next_cursor": "eyJkIjowLjEsImkiOiIwMThlMTIzNC01Njc4LTlhYmMtZGVmMC0xMTExMTExMTExMTEifQ=="
}