awrise

Docs · Source · pip install awrise · The Aither World

The Aither World is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. awnix is the Linux underneath it; awrise is one of its 66 bricks — each installs on its own, runs offline, and needs no account.

Start here: Wake something on a schedule, let it do one thing, and put it back to sleep.

Wake something on a schedule, let it do one thing, and put it back to sleep.

A scheduled job fails as a silence. A wake that never fired because the host was off, a wake that overlapped the run before it, and a wake that hung forever holding its slot are indistinguishable from outside: in every case nothing happened and nothing said so. Cron has no memory -- a missed minute is simply gone -- so "not due yet" and "due, and never ran" read identically.

awrise's product is not execution, it is the record. The host scheduler (cron, a systemd timer, launchd, Task Scheduler) is the clock; awrise run-due is one pass that exits; the wake ledger is the memory. Every wake ends in exactly one terminal row carrying a reason, and awrise history answers what happened to all of them.

No daemon. No background process. Standard library only -- no runtime dependencies.

Install

pip install awrise

Python 3.10 or newer. Linux, macOS and Windows.

Quick start

awrise add --name backup --every 1h --run "rsync -a src/ dest/"
awrise install-clock        # register THIS machine's scheduler, idempotently
awrise install --cron       # or --systemd-user / --systemd-system / --launchd / --schtasks
awrise install --check      # 0 installed and ticking, 1 a measured no, 2 unjudged
awrise history              # what happened to every wake and why
awrise status               # per-job verdict; exit 1 on a failing job

install writes nothing until you ask it to: --print renders the entry to stdout, --dry-run names every file and command it would touch, and an install is reported as successful only after it has been read back from the scheduler.

install-clock is install for an operator who does not want to know which scheduler the host has: it picks the native one, refuses to run twice (a second call that finds the same entry, the same interval, the same interpreter and the same payload bytes says so and changes nothing), and when the scheduler refuses the create it prints the exact command to run from an elevated shell rather than a diagnosis.

On Windows the two entries it registers are not equal: the repeating one registers for an ordinary user, and the at-startup one needs elevation. A refused at-startup entry therefore does NOT cost the host its clock -- the repeating entry is installed, and the gap (no wake after an unattended reboot, until somebody logs on) is printed by install-clock, install --check and doctor on every run, with its one-line fix.

A pass you ran by hand is not the clock ticking. Every adapter launches run-due --invoker <kind>, and freshness is judged on those rows only: a ledger full of hand-run passes reads as a clock that has never fired, which is what it is.

Commands

command	what it does
add --name N --every 15m --run "..."	register a job; refuses an empty command, a duplicate name, an unparseable interval, and a timeout at or above the interval (--allow-overrun to insist)
set --name N key=value ...	change one spec key; an unknown key is refused, never stored and ignored
enable --name N / disable --name N	a disabled job that comes due writes skipped_disabled, never silence
remove --name N	forget the job (refused while a wake of it is in flight)
list [--json]	the jobs, with the interval exactly as entered, the next due time, and the windows missed
run-due [--quiet] [--dry-run] [--prune 30d]	one pass: reconcile, tick, run every due job, exit
run --name N [--force]	run one job now, due or not
history [--job N] [--since 7d] [--event E] [--json] [--judge]	the ledger; --judge turns it into a verdict (drift or absence) and exits 2 on a window too young to judge
status	per-job verdict and the host-clock line; exit 1 on a failing job
explain --name N	why that job did or did not fire on the last tick
checks [--since 7d] [--json]	WL001-WL004 against the record: an unclosed wake, a reason outside the closed set, a job that should have woken and did not, a stale lock
prewarm [--apply] [--json] [--ledger-dir D] [--exclude GLOB] [--park-after] [--allow-derived-units]	read the usage ledger and propose one daily wake per unit something asked for yesterday; prints, and schedules nothing without --apply
prune [--keep 30d] [--dry-run] [--force]	drop ledger day files older than a window
install [--cron|--systemd-user|--systemd-system|--launchd|--schtasks] [--every 1m] [--print] [--dry-run] [--check] [--uninstall]	register (or judge, or remove) the host clock that runs run-due
install-clock [--every 5m] [--force]	register THIS machine's host clock, idempotently; on a refusal it prints the exact command to run elevated
reconcile [--restore] [--reset]	close orphaned wakes and break dead locks; --restore brings back the backup store
doctor	what of the aw* family is installed, and the one thing to fix

Every command exits 0 clean, 1 a measured no, 2 could not judge -- never 0 on silence.

The job spec

add and set accept exactly these keys, and refuse every other one:

key	default	meaning
every	required	Ns / Nm / Nh / Nd / Nw and compounds (1h30m); stored as a number, displayed exactly as entered
run	required	the command; empty is refused at add time, not skipped at run time
timeout_s	300	the whole process tree is killed at the bound, on POSIX and on Windows
cwd	none	the only environment knob; a missing directory is an error row, not a crash
at	none	HH:MM UTC daily anchor: due is the next anchor after the last start — or, until it has ever run, after it was added — so a job added at 14:00 waits for 03:00 and a laptop asleep at 07:00 catches up once and re-anchors instead of walking later every day
missed	catch_up_once	windows lost while nothing ran: catch up exactly once, or skip them
detach	false	spawn and close the wake at once; the child outlives the pass, and keeps the job's lock until it exits — so the overlap rule below holds for it too
enabled	true	a disabled job still leaves a row when it comes due
executor	shell	python, http, awrun, agent and session are optional, and import nothing until a job names them
wake	none	a unit to wake before the job runs and wait for; the wake is measured and recorded (--wake my-model.service)
park_after	false	with wake: ask for the unit to be put back to sleep afterwards; refused without wake, refused with detach, and skipped whenever the wake closed with the work still outstanding (a queued run-queue item)
wake_required	true	a failed wake means the command does NOT run; false runs it anyway and still records the failure
report	off	--report-relay '#channel' and --card-after N: a relay line, and one decision card after a failure streak
report.memory	false (set NAME report.memory=true)	recall this job's own past wakes into its env before it runs, and remember the one that just finished -- see Memory
predict	off (set NAME predict=warn|skip)	consult this job's own history before it fires -- see Predict

Overlap is not one of them: it is fixed behaviour. A job whose previous run still holds its lock is recorded skipped_overlap and is never started a second time.

Waking what the job needs

A job that talks to something which is asleep can name it:

awrise add --name nightly-render --every 1d --at 03:00 \
           --wake my-model.service --park-after \
           --run "render --all"
awrise prewarm                     # what yesterday's usage says should be pre-warmed
awrise prewarm --apply             # schedule exactly that

The unit is started by a host agent that owns the service manager; awrise never does. They meet on a directory both can write -- one file per request, one per answer -- so there is no port, no token and no daemon of awrise's own:

<plane>/requests/<id>.json   {id, action, unit, wait_s, created_at}
<plane>/results/<id>.json    {id, ok, state, reason, wake_s}
<plane>/agent.json           {alive_at}

Three rules make the failures loud rather than slow:

• no fresh heartbeat, no request. Nothing is written where nothing reads it, so a machine with no agent fails the wake in milliseconds instead of spending its whole budget.
• an unanswered request is failed, never done, and it is withdrawn -- an agent that wakes up late cannot start a unit for a job that already gave up on it.
• a failed wake does not run the command (unless wake_required: false). A command run against a unit that is not up fails in a way that reads like the command's own bug.

The wake time is the agent's own measurement when it made one, and it lands in the finished row as wake_s beside wake_unit, wake_state and park_state -- so "the command took 4s" and "waking what it talks to took eleven minutes" are two numbers in the record instead of one duration nobody can explain.

prewarm reads the usage ledger -- one <service>.json per service, written by the services themselves, holding when each was last asked for something. For every service used on the chosen day it proposes one daily job at that hour, and prints the EVIDENCE it used (basis=first_request_at when the record knows its first use of the day, last_request_at when that is all it has). A service with no entry for that day gets no proposal: silence is not a schedule. Nothing is written without --apply, and an absent ledger directory is exit 2 -- could not judge -- never an empty plan.

Every proposal also says where its UNIT NAME came from. unit=declared means the ledger record named the unit; unit=derived means awrise assembled one from the service name, which is a guess -- service names and unit names are written by different hands and they do not always agree. --apply schedules the declared ones and REFUSES the derived ones (exit 1, with the list), because a nightly job that names a unit nobody runs fails into a log and reads exactly like a quiet machine. --allow-derived-units says you checked them yourself.

The ledger

$AWRISE_HOME/ledger/YYYY-MM-DD.jsonl, one JSON object per line, flushed and fsynced per row. The writer refuses an event or a state outside the closed set, and refuses an empty reason, so a row can only come from something that was measured. Output tails are redacted before they are written.

awrise history --since 7d --json

Each pass also writes a tick and a tick_end row carrying the gap since the previous one, so a clock that stopped is a visible absence rather than a silence.

Memory

report.memory: true gives a job a small, ancestor-safe memory of its own past wakes, backed by awm -- off by default, and inert until set:

awrise set nightly-render report.memory=true

Before the job runs, awrise recalls its own last 20 finished wakes and exports them as JSON in AWRISE_MEMORY_JSON (an empty array [] on the first run, never absent). After it finishes, the outcome (state, reason, duration_s, exit_code, ts) is remembered under a key unique to that wake, so the next wake's recall includes it. awrise explain --name NAME shows the same payload live, read-only, for a job with report.memory: true.

The memory is scoped awrise:<host>:<job> -- one job, one host -- and the store lives at $AWRISE_HOME/awm/memory.db, isolated per AWRISE_HOME. AWRISE_MEMORY_JSON is inert data only: awrise never evaluates, sources or shell-interpolates it, and a shell_command job that does (eval $AWRISE_MEMORY_JSON) is that job's own choice and risk, not something this sink does or endorses.

Both the recall and the remember are bounded by a tight timeout and fail OPEN: a missing or broken awm, or a store that cannot answer in time, never blocks or fails the wake -- it costs one report_error ledger row and nothing else.

Predict

predict: warn or predict: skip gives a job a gate over its OWN history, backed by awpredict -- off by default, and inert until set:

awrise set nightly-render predict=warn    # log the verdict, always fire
awrise set nightly-render predict=skip    # refuse to fire on a bad-outcome verdict
awrise predict --name nightly-render      # the live verdict, read-only, right now

Before the job runs, awrise counts its own judged finished wakes (a success, or one of the failure / timeout / error / orphaned bad states -- a policy skip or a cancellation says nothing about the command's own behaviour, so it is not counted). Below five of them the verdict is UNJUDGED, fewer than 5 historical rows and the job fires exactly as if predict were off -- there is no awpredict sentinel for "not enough data yet"; awrise decides that itself, before ever calling the engine. At or above five, awrise asks a cached awpredict MLPWorldModel what its own history says the next wake's outcome looks like, and reads good or bad off the predicted reward.

predict: warn only LOGS the verdict -- on the job's started ledger row, as prediction: {verdict, reason, confidence, mode, rows} -- and always fires anyway; a bad-outcome verdict is information, not a veto. predict: skip is the only policy that can refuse to fire: on a bad-outcome verdict it writes a skipped_predicted row instead of running the job -- UNLESS the job's own PREVIOUS row was already skipped_predicted, in which case a real attempt is forced regardless of what the new prediction says, so a job can never be skipped twice running on a prediction alone. awrise --dry-run agrees with the real pass: a job that would be skipped_predicted next pass shows as hold, never would_fire.

The call is bounded by a tight timeout and fails OPEN: a missing or broken awpredict, a cold engine with nothing to say yet, or a call that does not return in time all degrade to UNJUDGED -- the job fires exactly as if predict were off. A genuine failure (a timeout or an exception, never "too few rows") also costs one report_error ledger row, same as the memory sink above -- never a failed wake.

Configuration

export AWRISE_HOME=/custom/path      # default ~/.aither/awrise, created 0700

# only for `wake` / `park_after` / `prewarm` -- unset on a machine that has no unit agent
export AITHER_GPU_UNIT_PLANE_DIR=/srv/plane      # where the unit agent answers
export AITHER_LIBRARY=/srv/library               # ...or <AITHER_LIBRARY>/Data/compute/gpu_unit_plane
export AITHER_USAGE_LEDGER_DIR=/srv/usage        # default: the plane's sibling `usage/`
export AWRISE_WAKE_BUDGET_S=600                  # how long a unit may take to come up healthy

With none of these set there is no plane: a job that names wake is an error row saying so, which is the point -- the alternative is a request written into a guessed directory that nobody ever reads.

awrise refuses to run when the home or the store is writable by anyone but its owner. Keep it on a native path: one home shared by a Windows and a WSL view of the same directory is protected by neither one's permissions.

Self-test

awrise --self-test           # run every case: 0 all passed, 1 a case failed, 2 a case broke
awrise --self-test --list    # the case names, one per line

Every case has a negative twin, so a case that can no longer fail is itself a failure. The list below is the output of awrise --self-test --list and must stay equal to it (AWR001).

Self-test verifies

• interval_parse_accepts_s_m_h_d_w_and_compounds
• interval_parse_rejects_empty_zero_negative_garbage
• store_crash_mid_save_keeps_jobs
• store_corrupt_exits_2_and_restore_returns_bak
• store_v1_migrates_with_tz_aware_stamps_and_keeps_v1_bak
• ledger_refuses_unknown_event_state_and_empty_reason
• ledger_accepts_the_closed_vocabulary
• ledger_redacts_output_tails
• run_due_failure_exit_code_propagates
• every_terminal_state_recorded_exactly_once
• reconcile_closes_orphan_and_does_not_refire
• spec_keys_accepted_equal_keys_read
• add_refuses_empty_command_duplicate_and_bad_timeout
• run_due_is_idempotent_inside_the_window
• python_m_awrise_entry_exists
• started_row_precedes_exec_and_stamp_follows_finished
• finished_wake_whose_stamp_was_lost_is_not_refired
• ledger_concurrent_appends_lose_nothing
• status_is_unjudged_for_a_job_that_never_woke
• store_refuses_malformed_records_with_exit_2
• add_refuses_names_outside_the_grammar
• reconcile_never_copies_an_unreadable_ts_into_the_store
• add_refuses_timeout_ge_interval
• overlap_is_skipped_while_the_lock_is_held_and_a_stale_lock_is_broken
• timeout_kills_the_whole_process_tree
• detach_closes_at_spawn_and_leaves_the_child_alive
• clock_skew_is_an_error_row_then_a_fire
• healthy_store_is_byte_stable_across_passes
• readded_job_never_inherits_a_removed_jobs_wake
• a_stale_lock_is_broken_by_exactly_one_pass
• release_never_removes_another_wakes_lock
• a_foreign_ledger_row_does_not_wedge_the_next_pass
• clock_skew_on_a_disabled_job_heals_in_one_pass
• timeout_past_the_platform_wait_is_refused
• hostclock_renders_every_adapter_the_same_way_twice
• schtasks_payload_is_crlf_hidden_and_inside_the_length_cap
• install_print_and_dry_run_touch_nothing
• install_check_is_unjudged_without_a_record_and_red_without_the_payload
• install_refuses_to_claim_an_entry_it_cannot_read_back
• doctor_never_exits_zero_while_printing_a_measured_no
• every_pass_records_a_tick_with_the_gap_since_the_last_one
• explain_reports_overdue_windows
• missed_windows_are_one_row_and_one_catch_up_fire
• missed_policy_skip_drops_the_windows_without_running
• at_anchor_fires_once_past_the_anchor_and_reanchors
• dry_run_says_what_would_fire_and_moves_nothing
• a_missed_row_needs_a_measured_clock_gap
• a_wake_in_flight_is_not_recorded_as_lost_windows
• dry_run_agrees_with_the_pass_on_a_skewed_disabled_job
• one_unreadable_line_cannot_buy_a_verdict
• the_window_guard_is_per_job_not_the_longest_cadence
• prune_is_unjudged_when_the_clock_cannot_be_read
• history_judge_flags_drift
• history_judge_is_unjudged_on_a_young_ledger
• checks_rules_each_fire_and_each_stay_quiet
• prune_removes_only_files_outside_the_window
• list_json_carries_the_exact_interval_and_the_next_due
• an_unreadable_clock_override_is_a_verdict_not_a_traceback
• optional_executors_never_import_their_packages_at_module_level
• an_unknown_executor_never_falls_through_to_the_shell
• a_bearer_path_outside_the_confinement_is_refused
• a_permission_mode_outside_the_allowlist_is_refused_at_add
• a_sink_that_cannot_run_is_a_row_not_a_lost_pass
• a_job_with_no_sinks_configured_writes_no_report_rows
• http_gaierror_is_skipped_unresolvable
• http_5xx_is_failure_with_body_tail
• card_raised_after_n_failures
• status_judges_a_detached_wake_by_its_receipt
• reconcile_stamps_a_detached_wake_from_its_receipt
• report_block_validated_at_add
• cwd_missing_is_error
• memory_off_by_default_touches_neither_env_nor_store
• memory_on_exports_recalled_wakes_and_excludes_ancestor_scope
• memory_failure_is_a_report_error_row_not_a_failed_wake
• explain_shows_the_same_recalled_memory_payload
• predict_off_writes_no_prediction_and_builds_no_engine
• predict_under_threshold_is_unjudged_and_still_fires
• predict_engine_is_built_exactly_once_across_due_checks
• predict_timeout_fails_open_and_writes_a_report_error_row
• predict_skip_holds_on_a_bad_verdict_and_never_skips_twice_running
• dry_run_predict_skip_agrees_with_the_real_pass
• import_routine_translates_a_shaped_yaml_and_writes_nothing_without_apply
• import_routine_apply_with_i_am_the_runner_writes_the_job_exactly_once
• import_routine_apply_without_the_runner_flag_is_refused
• import_routine_refuses_cron_and_non_shell_actions_by_name
• import_routine_refuses_a_block_pattern_and_a_metacharacter_under_apply
• import_routine_action_args_is_refused_by_name_and_never_concatenated
• history_import_fleet_is_read_only_and_flags_unparseable_records

Design

• The record is the product. Nothing is templated: every row comes from a measured outcome, and a pass that cannot judge something says so rather than reporting success.
• No daemon. A daemon that dies is the failure it was meant to prevent. A host clock that stops is a gap between ticks, and the ledger can see it.
• One pass, then exit. run-due reconciles, fires what is due, and returns an exit code a cron wrapper can act on.
• Nothing optional is imported until it is used. The http, awrun, agent and session executors, the relay and card sinks, and the awm-backed memory sink are guarded, off by default, and carry no default URL.
• Atomic state. jobs.json is written through a temporary file, fsynced and replaced, after every outcome; the previous copy is kept as jobs.json.bak. A corrupt store exits 2 and is restorable -- it is never silently replaced by an empty one.

Licence

MIT. See LICENSE.

The aw family

Standalone tools that share one idea: replace something you would otherwise have to trust with something you can check.

Each installs on its own, works offline, and needs no account.

	instead of trusting	you check
awdk	a framework's idea of how your agents should run	one loop you can read, pointed at a backend you already pay for
awskills	that an agent knows your procedure	the procedure written down, versioned, and loadable by any agent
awpack	that the pack you want shipped inside somebody's SDK, under whatever licence that SDK happens to carry	the pack as its own versioned artifact, with its own licence, that any agent runtime can install
awm	that memory stayed in its lane	tenant:user:project scopes, so a write cannot cross a boundary
awdesk	that the agent is somewhere behind a browser tab	a tray icon, a face on your desktop, and the decision card that pops when it needs you
awnode	a vendor's cloud with every prompt	a local gateway routing to backends you chose
awgraph	that grep found everything	an AST + tree-sitter call graph an agent can traverse
awgit	that no one else is editing this file	a lease, refused at commit time if you do not hold it
awdelphi	one agent's confident take on a decision	the round trace, the anonymity, and who dissents
awtoll	that your tooling is saving you context	the measured token cost of each tool call, and what the alternative cost
awseal	that the artifact came from who you think	an Ed25519 seal — the key that verifies is not the key that forges
awshare	that the download is intact	content-addressed bundles, verified on fetch
awnest	that there is a person on the other end	a verdict with evidence, where "we could not tell" is not "yes"
awrena	a leaderboard someone can edit, and votes nobody counted	a scored duel with both answers kept, and a result bound to them
awnboard	a share link anyone who sees it can use	an invitation addressed to one person, for one gate, revocable
awnix	that the box is what you left it as	an immutable image you built, with atomic rollback
awrecover	that the restore worked	a restore that fully lands or does not land at all
awstorage	a du you ran last month, and a peers file that says 3 TB free	an inventory snapshot per node with a diff since the last one, and each tree classified re-fetchable or not
awrelay	a SaaS in the middle of your agents	findings, alerts and coordination over your own transport
awask	that anyone read the paragraph where you asked	the ask itself, with a button that steers the session that raised it
awmail	a mailbox somebody else can read	mail your agents send and receive over your own server
awfind	one vendor's idea of the web	results from whichever providers you configured
awbrowse	that the page said what you were told	the render, the DOM and the requests it made
awvoice	that a cloud vendor may hold your audio	a transcript and a wav from a service you host
awvision	a filename and a caption somebody wrote	what a model actually reports about the pixels
awscreen	a selector that was true when the page was written	the elements actually rendered, by what they look like
gawbbonet	the model to keep a 300-message campaign coherent by itself	campaign facts recalled from scoped memory you can list and edit
aitherkvcache	a vendor's quantisation defaults	sub-byte KV cache kernels you can benchmark yourself
awrtifact	a hand-rolled split script and a hand-edited worker manifest	byte-verified parts in a release, served with Range + CORS, sizes asserted by a live gate
AitherZero	a pile of scripts nobody has numbered	numbered, discoverable automation with declarative playbooks
AitherConnect	what a page tells your browser to do	a federated search and desktop bridge you host
awreason	a confident paragraph	the phases it went through, and every tool call it made to get there
awrecurse	that everything you pasted in was actually read	which slices it opened, and what it concluded from each
awprism	the first explanation that fits	the ranked alternatives, and the observation that separates them
awrepl	what the agent believes the value is	the value, printed from the live session
awresearch	a summary of pages nobody opened	every claim against the source it came from
awfocus	twelve terminal tabs and a bad memory	one command that names every session, finds any transcript, and opens or steers the one you want
awgym	that a world model learned anything from the games it saw	transitions captured from real play, fed back, and the retrodiction score falling on grids it never saw
awpredict	a model because it trained without erroring	its prediction against a self-updating lookup, on the rows that are actually novel
awsh	that you already know the name of the command	what it decided your line meant, before it acts on it
awrise (you are here)	that a scheduled agent ran at all, and ran exactly once	a durable record of every wake -- fired, skipped, overlapped or timed out -- each with its reason
awkno	that the docs site is up, or that you remember the family	the whole ecosystem in your terminal, with no network at all
awwall	that a service only talks to the hosts you think it talks to	an explicit egress allowlist, where a denial names the rule that denied it
awembed	a general-purpose embedder that has never seen your code	a held-out split of whole directories, scored teacher vs student vs int8
awtax	a closed tax app's sealed file you can never read again	a plain, provider-neutral schema of every figure, with the page it came from
awsettings	that you will remember to re-approve the same thing on every box you work from	one profile, unioned rather than overwritten, with the credentials left behind
awavatar	a cloud 3D vendor's opaque task id	a manifest with a sha256, a licence and a rig-audit verdict per file

awnix is the ground floor — A Linux you can hand to an agent — immutable base, capabilities included.

The aw family

Standalone tools that share one idea: replace something you would otherwise have to trust with something you can check.

Each installs on its own, works offline, and needs no account.

	instead of trusting	you check
awdk	a framework's idea of how your agents should run	one loop you can read, pointed at a backend you already pay for
awskills	that an agent knows your procedure	the procedure written down, versioned, and loadable by any agent
awpack	that the pack you want shipped inside somebody's SDK, under whatever licence that SDK happens to carry	the pack as its own versioned artifact, with its own licence, that any agent runtime can install
awm	that memory stayed in its lane	tenant:user:project scopes, so a write cannot cross a boundary
awdesk	that the agent is somewhere behind a browser tab	a tray icon, a face on your desktop, and the decision card that pops when it needs you
awnode	a vendor's cloud with every prompt	a local gateway routing to backends you chose
awgraph	that grep found everything	an AST + tree-sitter call graph an agent can traverse
awgit	that no one else is editing this file	a lease, refused at commit time if you do not hold it
awdelphi	one agent's confident take on a decision	the round trace, the anonymity, and who dissents
awclassify	a filename, a folder, or whoever last touched it	doc_type, visibility, audience and topics, with the evidence lines that decided each
awdecide	a hosted classifier's probability that never learns whether it was right	the decision, its probability, and the calibration curve from your own resolved outcomes
awtoll	that your tooling is saving you context	the measured token cost of each tool call, and what the alternative cost
awseal	that the artifact came from who you think	an Ed25519 seal — the key that verifies is not the key that forges
awshare	that the download is intact	content-addressed bundles, verified on fetch
awnest	that there is a person on the other end	a verdict with evidence, where "we could not tell" is not "yes"
awrena	a leaderboard someone can edit, and votes nobody counted	a scored duel with both answers kept, and a result bound to them
awnboard	a share link anyone who sees it can use	an invitation addressed to one person, for one gate, revocable
awnix	that the box is what you left it as	an immutable image you built, with atomic rollback
awrecover	that the restore worked	a restore that fully lands or does not land at all
awstorage	a du you ran last month, and a peers file that says 3 TB free	an inventory snapshot per node with a diff since the last one, and each tree classified re-fetchable or not
awrelay	a SaaS in the middle of your agents	findings, alerts and coordination over your own transport
awask	that anyone read the paragraph where you asked	the ask itself, with a button that steers the session that raised it
awmail	a mailbox somebody else can read	mail your agents send and receive over your own server
awswarm	that a model either fits your GPU or it doesn't run at all	a placement plan and an acquisition-probability estimate before you spend on a run
awfind	one vendor's idea of the web	results from whichever providers you configured
awbrowse	that the page said what you were told	the render, the DOM and the requests it made
awvoice	that a cloud vendor may hold your audio	a transcript and a wav from a service you host
awvision	a filename and a caption somebody wrote	what a model actually reports about the pixels
awscreen	a selector that was true when the page was written	the elements actually rendered, by what they look like
awbeads	that a layout your users built survives the next deploy	the arrangement as data you can read back, diff, and hand to another surface
awbonsai	that inference always means a request left the machine	a WebGPU model answering on the tab's own GPU, with a consent record logged before it ever loaded
gawbbonet	the model to keep a 300-message campaign coherent by itself	campaign facts recalled from scoped memory you can list and edit
aitherkvcache	a vendor's quantisation defaults	sub-byte KV cache kernels you can benchmark yourself
awrtifact	a hand-rolled split script and a hand-edited worker manifest	byte-verified parts in a release, served with Range + CORS, sizes asserted by a live gate
AitherZero	a pile of scripts nobody has numbered	numbered, discoverable automation with declarative playbooks
AitherConnect	what a page tells your browser to do	a federated search and desktop bridge you host
awreason	a confident paragraph	the phases it went through, and every tool call it made to get there
awrecurse	that everything you pasted in was actually read	which slices it opened, and what it concluded from each
awprism	the first explanation that fits	the ranked alternatives, and the observation that separates them
awrepl	what the agent believes the value is	the value, printed from the live session
awreport	that the report you pasted carried no token in it	a redacted report, and the duplicate it merged into instead of filing twice
awresearch	a summary of pages nobody opened	every claim against the source it came from
awfocus	twelve terminal tabs and a bad memory	one command that names every session, finds any transcript, and opens or steers the one you want
awgym	that a world model learned anything from the games it saw	transitions captured from real play, fed back, and the retrodiction score falling on grids it never saw
awpredict	a model because it trained without erroring	its prediction against a self-updating lookup, on the rows that are actually novel
awevolve	that your optimisation loop is finding anything	every version it kept, the score that version earned, and the edit that produced it
awsh	that you already know the name of the command	what it decided your line meant, before it acts on it
awmine	that a session's lesson survived the session	a row per outcome, a candidate per lesson, and the transcript line each one came from
awrise (you are here)	that a scheduled agent ran at all, and ran exactly once	a durable record of every wake -- fired, skipped, overlapped or timed out -- each with its reason
awkno	that the docs site is up, or that you remember the family	the whole ecosystem in your terminal, with no network at all
awwall	that a service only talks to the hosts you think it talks to	an explicit egress allowlist, where a denial names the rule that denied it
awembed	a general-purpose embedder that has never seen your code	a held-out split of whole directories, scored teacher vs student vs int8
awtax	a closed tax app's sealed file you can never read again	a plain, provider-neutral schema of every figure, with the page it came from
awsettings	that you will remember to re-approve the same thing on every box you work from	one profile, unioned rather than overwritten, with the credentials left behind
awavatar	a cloud 3D vendor's opaque task id	a manifest with a sha256, a licence and a rig-audit verdict per file

awnix is the ground floor — A Linux you can hand to an agent — immutable base, capabilities included.

The Aitherium ecosystem

Every repository here is public. Each publishes an aither-manifest.json beside its page, so any surface can read every sibling's — the network is browsable from any node in it.

repo	what it is	pages
awdk	Build AI agent fleets — 3 lines, any backend, local or cloud	docs
awskills	Portable agent skills — self-contained procedures an agent loads on demand	docs
awpack	First-party agent packs — the ones we build, versioned and installable on their own	docs
awm	A portable, scoped agent memory	docs
awdesk	Aither World Desk -- the desktop body of AitherOS Online: tray, avatars, decision cards, the Living Desktop as an overlay	docs
awnode	A lightweight local gateway — bridges your apps to the AI backends you chose	docs
awrun	A priority-aware queue and dispatcher for agentic runs and ad-hoc CI builds. It also judges whether the runner pool is big enough for the queue it is draining, and can ask a host to grow it -- reserving capacity is zero-sum, so a saturated pool needs more of it, not a different share of it	docs
awgraph	A semantic code graph for agents — AST + tree-sitter, call graphs	docs
awgit	Semantic version control on top of git — edit-ops and leases	docs
awdelphi	Anonymous multi-round expert panels — a converged answer with a trace	docs
awclassify	Classify any document -- what it is, who may read it, who it is for, what it is about	—
awdecide	One typed-decision contract -- choice / score / bool with a probability -- over a ladder of backends you already run (rules, tiny local models, an LLM's logprobs), fail-closed, with a Brier ledger that resolves every decision against its outcome	—
awtoll	What every tool call costs you in context, measured from your own transcripts	docs
awseal	Sign an artifact so a stranger can verify it	docs
awshare	Publish an artifact and fetch it back verified	docs
awdit	An append-only audit trail whose gaps are DETECTABLE	docs
awbac	Role-based access control that fails closed and explains itself	docs
awiam	Who is this caller? A directory and session store that fails honestly	docs
awtunnel	Reach a service that has no public address	docs
awnest	Prove there is a human before you let them into the nest	docs
awrena	Put two agents head to head and get a verdict you can check	docs
awnboard	A front gate you can put in front of anything, and hand someone the key to	docs
awnix	A Linux you can hand to an agent — immutable base, capabilities included	docs
awrecover	Labelled snapshots with an all-or-nothing restore	docs
awstorage	Every drive on every node, indexed, classified and diffed -- so you can see what you own before you delete it	docs
awrelay	Portable agent messaging — findings, alerts, coordination	docs
awask	Your agent asks you a question — and acts on your answer	docs
awmail	Give an agent an email address — send, and actually receive	docs
awnet	The agentic web — agents host a mesh, and agents join one	docs
awswarm	Run one model too big for any single GPU across a pool of small ones	—
awfind	A portable search client — query, results, ranking	docs
awbrowse	A portable browser client — navigate, console, network, DOM, screenshot	docs
awvoice	Hear and speak — transcribe audio, synthesize a voice	docs
awvision	See an image — describe it, ask it a question, compare two	docs
awscreen	See this machine — what is on screen, and where to click it	docs
awkit	Render an agent panel from a tool result — one component, any React app	—
awbeads	A spatial canvas for a page — arrange things, connect them, and keep the arrangement	—
awbonsai	Run a real model in the visitor's own browser — no server round trip, no upload	—
awknowledge	How to run a coding agent so the result survives — the laws, with evidence	docs
awbrain	Your history as a wiki of linked markdown — claims pinned to the evidence	—
gawbbonet	GobboNet campaigns with a real agent brain — scoped memory, graph recall	docs
aitherkvcache	Near-optimal KV cache quantization for LLM inference — sub-byte compression	docs
awrtifact	Deliberately chunk artifacts into GitHub release assets — the productized aitherkvcache mirror lane	docs
AitherZero	PowerShell 7+ automation framework — numbered, self-describing scripts	docs
AitherConnect	Browser extension — federated AI search, page context, and the Living OS overlay	docs
awreason	A portable reasoning client — sessions, phases, thoughts, and the chain that produced the answer	docs
awrecurse	Answer a question over a context far larger than the window — recursively, with the trace kept	docs
awprism	Turn a failure into ranked hypotheses — and say what would confirm each one	docs
awrepl	A REPL an agent can actually use — state that survives between turns	docs
awreport	File a bug report that has already scrubbed your secrets and collapsed the duplicate	—
awresearch	Ask a research question, get a cited report you can check	docs
awfocus	See, search and steer every Claude session from one command	docs
awgym	An ARC training gym — a game a world model can watch, and six roles that play through it	docs
awpredict	Predict what your environment does next, and how surprised you were	docs
awevolve	Point an agent at a file and a command that scores it, and let it improve	—
awsh	Your terminal answers you -- type a question where a command would go	docs
awmine	Mine what your agents did -- outcomes, lessons and procedures out of the transcripts they left behind	—
awrise (you are here)	Wake an agent on a schedule, let it do one thing, and put it back to sleep	docs
awkno	The man page for the Aither World — every brick, stack and law, offline	docs
awwall	Say what a workload may reach, and watch everything else fail closed	docs
awrouter	OpenRouter for your own fleet: pick a model backend by cost/latency/ capability, fail over, fit the context window, stream. Standalone, OpenAI-compatible, no Aither-specifics required to be valuable	—
awembed	Train an embedding model that knows your corpus, and prove it beats the big one	docs
awtax	Turn any tax PDF -- returns, W-2, 1099, statements, even scans -- into structured data you can check	docs
awflow	A deterministic workflow runtime — chain agent calls with journal replay and budget control	docs
awsettings	Your agent's permissions and config, following you to the next machine	docs
awavatar	One character spec in, a rigged, animated, multi-style avatar pack out	docs

<script src="aither-constellation.js"></script>