Expand description
The named parse limit sets of spec 002, 3.2.5. Each document type uses
exactly one of them through crate::Document::LIMITS.
Constants§
- BACKEND_
OUTPUT_ V1 - Backend outputs; masses, logits and scores are JSON numbers.
- DEFINITION_
V1 - Decision definitions.
- DESCRIPTOR_
V1 - Backend descriptors and calibration artifacts.
- DYNAMIC_
QUESTION_ V1 - Dynamic questions,
rustev.dynamic-question/1(spec 019, 3.2): the request’s own bounds (128 KiB canonical, 64 KiB of context values) are checked after this scan; the scan admits at most twice the canonical bound in raw bytes, so whitespace and escapes cannot force an unbounded allocation. - EXCHANGE_
V1 - Exchange records,
rustev.remote-exchange/1(spec 009, 3.9.1): a fixed maximum size, with free text cut to 256 bytes before escaping. - PLAN_V1
- Compiled plans.
- RECORD_
V1 - Judgments, evidence records and evaluation reports.
- REMOTE_
V1 - The
rustev.remote/1documents (spec 009, 3.10): descriptors inside a describe answer, projections as strings, outputs as JSON numbers. A transport also bounds what it reads by the negotiated maximum. - REPLAY_
V1 - Replay bundles and eval-owned documents (spec 004, 3.2 and 3.5).
- REQUEST_
V1 - Request identity documents:
REPLAY_V1tightened to 4 MiB (spec 004, 3.4.2). The projection inside still obeys the plan’s own bound. - SNAPSHOT_
V1 - Context snapshots.
- TRACE_
V1 - Span lines,
rustev.trace/1(spec 028, 3.10.1): at most 64 KiB per line. An attribute value is at most 4 096 bytes before escaping, so a string’s raw length is bounded at six times that.