Configuration¶

Most Rank Hunter configuration is stored through the typed settings registry and edited under Settings.

The default working database is rank42.db in the checkout unless another database path is supplied.

Registered settings¶

The 0.9.2 core registry includes:

Key	Type/default	Purpose
`ui_theme`	theme id	selected Rank Hunter theme
`ui_appearance`	`light`	light/dark appearance
`science_python`	detected path	Python/Sage executable for scientific jobs
`ratpoints`	detected vendored CPU path	CPU point-search executable
`ratpoints_gpu`	optional detected path	GPU point-search executable
`ratpoints_backend`	`CPU`	default CPU/GPU backend
`pari_rank_timeout`	300 s	default PARI rank budget
`mwrank_rank_timeout`	300 s	default mwrank rank budget
`deep_cert_timeout`	900 s	deep exact-certificate budget
`queue_max_workers`	2	global concurrent queue workers
`queue_resource_limits`	mapping	per-resource-class concurrency ceilings

The registry validates type/range before storing values.

Scientific Python¶

science_python must be an executable that can import Sage.

Changing the string to an arbitrary Python path is not enough; runtime validation must succeed.

Example check:

/path/to/python -c "from sage.all import QQ, EllipticCurve; print('OK')"

Point-search executables¶

ratpoints and ratpoints_gpu point to the CPU and GPU executables.

GPU is optional.

If ratpoints_backend=GPU but GPU validation fails, inspect Diagnostics and either repair CUDA/native build or switch the default back to CPU.

Timeouts¶

Global timeout settings are defaults.

Specific UI/Pipeline/CLI commands may override them.

A timeout is always an operational budget, not a mathematical cutoff.

Increasing pari_rank_timeout does not change what a completed PARI proof means; it only gives the engine more time.

Queue limits¶

queue_max_workers is constrained to a bounded range by core.

queue_resource_limits can prevent too many jobs of one resource class from running simultaneously.

Typical resource classes include:

GPU ratpoints;
Sage-heavy work;
database maintenance.

Set these based on actual RAM/GPU/CPU capacity rather than logical CPU count alone.

Database path¶

Many commands accept:

--db rank42.db

The UI launcher also supplies its configured database path.

Two different database paths are two different Rank Hunter research states.

Always confirm the path before running manual CLI commands.

Database environment variables¶

Compatibility/runtime paths may recognize database environment variables such as:

RANK_HUNTER_DB
RANK42_DB

Prefer explicit --db in scripts when reproducibility matters.

Bootstrap marker¶

The installer can create:

.rank42-ui/science-python

This is bootstrap input for the scientific Python setting. It should not silently override an already-persisted valid user setting.

Secrets¶

The current core settings registry distinguishes secret/non-secret settings. Do not add service credentials as arbitrary plain-text config files if a registered secure path exists.

Configuration vs application state¶

User/runtime configuration belongs in settings.

Current application state (selected page, temporary UI state) and dispatcher heartbeat/service state are separate concerns.

Do not persist transient application state as if it were user configuration.

Reproducible scripts¶

For reproducible command-line research:

pass --db explicitly;
record the command;
record plugin versions;
record important timeout/search bounds;
do not depend on an unrecorded UI selection.