Flux Store
$15
// operating system · $15

Preflight and health-gate your optimization pipeline.

Inspect registered module scripts, ledgers, and locks; supervise your configured tasks with output checks; and record changes you have already made. Includes two synthetic local examples.

Get Instant Access · $15Buy · $15
flux@store
# Setup outline · see included README
› Inspect registered paths, ledgers, and locks
› Review the preflight report
› Run registered tasks through health checks
› Record results and externally completed changes
cortex · portable workflow
1Inspect registered paths, ledgers, and locksscoped
2Review the preflight reportscoped
3Run registered tasks through health checksscoped
4Record results and externally completed changesscoped
Workflow overviewRead · configure · verify
$ run --pipeline
1
Inspect registered paths, ledgers, and locks
2
Review the preflight report
3
Run registered tasks through health checks
4
Record results and externally completed changes
$ explain --how-it-works

How it actually works

Exactly what it does — and when.

1Preflight checks registered paths, ledgers, and locks and writes a report without editing target source
2The supervisor checks each registered task against its declared output rule
3The adoption tracker records externally completed changes; it does not implement them
In practice · A scheduled task exits zero but writes nothing

CORTEX detects that the registered output check failed, records the failed run, and follows its retry or alarm path. The operator investigates the task, makes any required changes outside the adoption tracker, and records the supporting evidence.

$ cat spec.txt
IncludedPreflight, supervisor, adoption tracker, PID lock, two examples
SetupNode.js 18+ · run the documented local examples
ProvidersBring your own models or modules
BundleStandalone or in the $199 bundle
$ benchmark --vs-alternatives
This product
A cron job with an exit-code check
Preflight
Read-only target inspection
Job starts immediately
Success
Declared output must change
Exit zero is enough
Lock
Holder PID liveness is checked
Age-only locks can wedge
Adoption
Records changes already made and caller evidence
No separate adoption record
$ diff --before-after
Without it
  • A scheduled job exits zero after doing nothing
  • A dead PID leaves a stale lock
  • Changes are recorded without their supporting evidence
With CORTEX
  • Run a read-only preflight before changing a target
  • Check whether each registered step produced its declared output
  • Use PID-liveness locking and a health ledger around scheduled work

Use it to…

Real ways people run CORTEX.

Preflight an unfamiliar pipeline

Inspect detected entry points and proposed opportunities before approving any write.

Supervise scheduled steps

Wrap registered self-play, distillation, or other modules in declared output checks.

Recover a dead lock holder

Use PID liveness rather than lock age alone when deciding whether a run is still active.

What's inside

Preflight checks
Module supervision
Execution ledgers
PID-safe orchestration
MIT-licensed source

Who it's for

Maintainers of local optimization jobs

need silent no-ops to surface

Teams with scheduled learning scripts

need output checks and single-flight locking

Operators approving pipeline changes

need a read-only plan before mutation

Questions, answered

✓ Instant delivery✓ Yours to own✓ Use on every project✓ Built to ship results

Ready to put CORTEX to work?

Get Instant Access · $15

Instant download · one-time price · own forever · all sales final

← Back to all products
CORTEX
$15Get Instant Access