BONDSTATSrelay
LOCAL-FIRST / NO ACCOUNT
BONDSTATS / DEVELOPER TOOLS / AUTOMATION CONTROL ROOM

RELAY

Automation with a reason.
Not just a trigger.

A source changes. A rule evaluates it. An action either runs or is deliberately held. Relay makes every step inspectable, lets you rehearse failure cases and exports an automation you can operate yourself. No account, external SDK or hosted dependency.

Open control room Build → test → inspect → export
RELAY / SIGNAL INSTRUMENTAWAITING TEST
LOCAL SIMULATIONS00
LATEST DECISION—
01 SOURCE → 02 COMPARE → 03 DECIDE → 04 DELIVERSIMULATION IS LOCAL · NO WEBHOOKS DISPATCHED
THE MISSION INDEX

Start with a real operation.

Start with a real task, not an empty canvas. Every recipe is editable, testable and exportable as code.

Research intakePreserve new research entries as local Markdown notes; repeated items are held.
CONTROL ROOM / BUILD + REHEARSE

Configure. Challenge. Verify.

Configure the job on the left. Watch its route change on the right, then run the evidence through the decision gate before exporting.

JOB DEFINITIONLOCAL / EDITABLE

Job specification

Use a public HTTPS endpoint. No credentials required.

What counts as an event?

Where should it go?

Optional HTTPS webhook

Only use a receiver you control. Webhook delivery is at-least-once and the receiver should deduplicate event IDs.

Your source and configuration stay in this browser until you export them. A running local Python worker performs the real checks after deployment.
THE EVENT ROUTE

Event decision path

AWAITING EVIDENCE
LIVE CONFIGURATION / SIMULATED DECISIONS
INGEST / 01RSS / Atom

Identify feed entries

COMPARE / 02Deduplicate

Baseline and prior state

DECIDE / 03New item

New IDs, optional keyword

DELIVER / 04Local action

Markdown research note

RELEASE GATEChecking configuration…
Awaiting configuration
DECISION LAB / OFFLINE REHEARSAL

Test before it runs.

Rehearse against an editable source response before deploying. No remote source is fetched and no local action or webhook is executed here.

TRAIN THE TRAINER / THREE-STEP REHEARSAL

Watch a signal become an event.

03 CHECKS

Compare three outcomes against the same rule: first observation, identical replay and a genuine change. The final decision must be justified by the evidence.

Run the guided rehearsal to see the baseline, duplicate and genuine change compared side by side.
PORTABLE EXECUTION / YOUR MACHINE

Export the system.
Keep the control.

Export a complete Python 3.9+ runner, your rules, operating instructions and an MIT licence. The worker checks the source, keeps local state, records its decisions and writes the configured output. No account, external package or BondStats API key is needed. A running computer or server is required for unattended execution.

Configuration ready to export.
EXECUTION / AFTER EXPORTPYTHON 3.9+ · STDLIB
# Rehearse without writing state
python3 relay_runner.py --config relay_config.json --dry-run

# Establish a local baseline
python3 relay_runner.py --config relay_config.json --once

# Watch while this computer stays on
python3 relay_runner.py --config relay_config.json --watch
For execution after terminal closure or restart, schedule --once through Windows Task Scheduler, macOS launchd or Linux cron/systemd. A powered-on device and an accessible source are still necessary.
CONTROL

A baseline is not an alert.

Relay learns the current state before triggering future changes. For numerical rules, it looks for a transition into the matching condition rather than firing at every polling interval.

EVIDENCE

Failures are visible.

The exported runner records checks, held decisions and errors in local JSONL logs. A failed source request is not interpreted as a meaningful new event.

OWNERSHIP

No invisible dependencies.

The runner uses Python's standard library. External sources may change, polling can miss intermediate states and webhook delivery requires receiver-side deduplication.