# `query/engine` — plan + execute orchestration
Glue between planning, materialization, folding, transforms, and spill export. Public planning functions are re-exported at the `query` crate root.
## Files
| `run.rs` | `plan_query_with_tet_mmap_ex`, `plan_read_for_document`, main execute path |
| `operations.rs` | Match `Operation` variant → fold / materialize / transform / covariance |
| `budget.rs` | `ExecutionBudget`, `MemoryStrategy` (in-core vs spill vs streaming) |
| `spill_policy.rs` | Temp spill files, sidecar draft paths, path allowlist (`--spill-allow`) |
## Memory strategies
| `streaming_fold` | Tier-A/B `operation` |
| `in_memory_materialize` / `temp_spill_materialize` | Tier-C ops |
| `mmap_spill` | Top-level `spill` export |
| `transform_ram` / `transform_spill` / `transform_sidecar` | `transform` with `write` routing |
| `capped_in_memory` | Preview-only execute |
See [`docs/query_engine.md`](../../../docs/query_engine.md#memory-budget-and-execution-strategies).
## Key exports
- `plan_query_empty`, `plan_query_with_tet_mmap`, `plan_query_with_tet_mmap_ex`, `plan_read_for_document`
- `materialize_read_plan_*` (re-exported from `materialize`)
- `SpillPathAllowlist`, `ExecutionBudget`, `DEFAULT_MEMORY_BUDGET_BYTES`
## Callers
- `execute.rs` — embedder API
- `bin/tet/query.rs` — CLI `-x`
- `ffi` — `tet_query_json`