This avoids Surreal parse errors and allows arbitrary context key/value benchmarks to run successfully.