Skip to content

memhtml eval discriminate

1. Synopsis

The refusable retrieval gate: every probe must outrank its own wrong-fact twins.

Terminal window
memhtml eval discriminate [options]

2. Arguments

This command takes no positional arguments.

3. Flags

Flag Type Default Values Description
--mode string fake fake, live fake is the deterministic embedder CI measures; live needs AWS_BEARER_TOKEN_BEDROCK and refuses loudly without it.
--seed integer no default no fixed set The fixture corpus seed. A failing run is reproducible from this number.
--now integer no default no fixed set The run instant the fixture corpus anchors its stamps behind, UTC milliseconds since the epoch. The other half of reproducing a failing run: the corpus is a function of (seed, now), and the recency arm ranks on those stamps. Defaults to the clock, and rides back in the report.
--size integer 200 no fixed set Base memories to generate.
--probes integer 36 no fixed set Probes to run. Design §5 wants ≥30.
--mrr-floor string 0.85 no fixed set Mean-reciprocal-rank floor. Lowering it is a deliberate, visible choice.

Every command also accepts the global flags.

4. Response

On success the command writes one JSON envelope to stdout, and its type is eval.discrimination.

On failure it writes the failure envelope, whose code comes from error codes.

The JSON envelope gives the fields both shapes carry. Response types lists every value type takes across the binary.

5. Further reading

No guide block names this command. The guide is the workflow prose the CLI ships beside the command table: which command to reach for, and in what order.

6. Provenance

A loader generates this page from apps/cli/src/commands.ts while the site builds, so no file in the repository holds it: the COMMANDS array there is the one source of argument parsing, of memhtml manifest, and of AGENTS.md. Change the registry and this page changes with it.