vxcontrol/pentagi · error

knowledge: similarity search (user %d): %w

Error message

knowledge: similarity search (user %d): %w

What it means

The user-scoped branch of doSearch runs SearchUserKnowledgeDocuments, a pgvector cosine-distance query filtered by user namespace. Any SQL execution error is wrapped as 'knowledge: similarity search (user <id>)'.

Source

Thrown at backend/pkg/database/knowledge/knowledge.go:398

	type searchRow struct {
		ID        string
		Document  string
		Cmetadata sql.NullString
		Score     float64
	}

	var rawRows []searchRow

	if userID > 0 {
		rows, err := ks.db.SearchUserKnowledgeDocuments(ctx, database.SearchUserKnowledgeDocumentsParams{
			Embedding:   vecLiteral,
			UserID:      nsOf(strconv.FormatInt(userID, 10)),
			MaxDistance: maxDist,
			Lim:         int32(limit), //nolint:gosec
		})
		if err != nil {
			return nil, fmt.Errorf("knowledge: similarity search (user %d): %w", userID, err)
		}
		rawRows = make([]searchRow, len(rows))
		for i, r := range rows {
			rawRows[i] = searchRow{r.ID, r.Document, r.Cmetadata, r.Score}
		}
	} else {
		rows, err := ks.db.SearchKnowledgeDocuments(ctx, database.SearchKnowledgeDocumentsParams{
			Embedding:   vecLiteral,
			MaxDistance: maxDist,
			Lim:         int32(limit), //nolint:gosec
		})
		if err != nil {
			return nil, fmt.Errorf("knowledge: similarity search (admin): %w", err)
		}
		rawRows = make([]searchRow, len(rows))
		for i, r := range rows {
			rawRows[i] = searchRow{r.ID, r.Document, r.Cmetadata, r.Score}
		}

View on GitHub (pinned to ea665308ba)

Solutions

1. Read the wrapped cause: 'vectors have different dimensions' means re-embed stored data or align the embedding model with the column dimension.
2. Run goose migrations to create/upgrade pgvector tables and extension.
3. Check PostgreSQL health and connection limits if the error is connection-related.
4. Validate the limit parameter if it triggers a query error.

Example fix

// before
Lim: int32(limit), //nolint:gosec
// after
if limit <= 0 || limit > 1000 {
    limit = defaultSearchLimit
}
// then: Lim: int32(limit),

if limit <= 0 {
    limit = defaultSearchLimit
}
if strings.TrimSpace(query) == "" {
    return errors.New("query must not be empty")
}

func isDimensionMismatch(err error) bool {
    return strings.Contains(err.Error(), "different dimensions") || strings.Contains(err.Error(), "does not match")
}

docs, err := store.SearchUserDocuments(ctx, userID, query, filter, limit)
if err != nil {
    if isDimensionMismatch(err) {
        // re-embed corpus or fix model config before retrying
    }
    return err
}

When it happens

Trigger: Calling SearchUserDocuments when the vector column dimension mismatches the embedding model output, the pgvector extension is missing, limit is invalid, or the DB connection fails.

Common situations: Switching embedding models so stored vectors have a different dimension than vecLiteral (pgvector dimension mismatch error); fresh DB without migrations; connection pool exhaustion under load.

Related errors

• knowledge: similarity search (admin): %w
• failed to create flow vector store log: %w
• failed to get flow vector store log: %w
• knowledge: list by flow: %w
• knowledge: list all: %w

AI-assisted analysis of vxcontrol/pentagi@ea665308ba (2026-09-01). Data as JSON: /api/errors/7f091934c2b4eeb9. Report an issue: GitHub.