Derive heartbeat cadence from the shared coordination frame

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a13eab-1e6a-7b33-9c37-372e90ada3a2
This commit is contained in:
tegwick 2026-10-15 20:41:49 +02:00
parent db850aded5
commit a786b061da
9 changed files with 98 additions and 27 deletions

View file

@ -9,6 +9,7 @@ from typing import Any, Callable
from urllib.error import HTTPError, URLError
from urllib.request import Request, urlopen
from .coordination import FRAME_SECONDS
from .store import Engine
@ -91,7 +92,7 @@ class OutboxWorker:
engine: Engine,
sink: Callable[[dict[str, Any]], None],
*,
heartbeat_interval_seconds: int = 300,
heartbeat_interval_seconds: int = FRAME_SECONDS,
) -> None:
self.engine = engine
self.sink = sink

View file

@ -8,6 +8,7 @@ from wsgiref.simple_server import make_server
from .api import App
from .audit import AuditCoreSink, OutboxWorker
from .coordination import FRAME_SECONDS
from .auth import Identity, JWTAuthenticator, StaticTokenAuthenticator
from .store import Engine
@ -45,7 +46,7 @@ def _parser() -> argparse.ArgumentParser:
serve.add_argument("--audit-url")
serve.add_argument("--audit-token-file")
serve.add_argument("--outbox-poll-seconds", type=float, default=5)
serve.add_argument("--heartbeat-seconds", type=int, default=86400)
serve.add_argument("--heartbeat-seconds", type=int, default=FRAME_SECONDS)
migrate = sub.add_parser("migrate", help="apply repeatable schema migrations")
_common_db(migrate)
verify = sub.add_parser("verify", help="verify schema and SQLite integrity")

View file

@ -0,0 +1,6 @@
"""Generated cadence timings; regenerate with NetKingdom coordination-frame tool."""
# Source: coordination-engine/config/coordination-frame.json
FRAME_CONFIG_SHA256 = "d15e092387f128ffdfd00b9815707a05ff77c5bb0ba4d9a428fbc37b8c440b67"
FRAME_SECONDS = 3600
RECEIVER_WAIT_SECONDS = 7200
EFFECTIVE_DEADLINE_SECONDS = 10800

View file

@ -25,14 +25,15 @@
},
"heartbeat": {
"event_class": "audit-core.heartbeat",
"interval_seconds": 86400,
"interval_seconds": 3600,
"assertion": "nothing-to-report",
"missing": "finding",
"extensions": {
"approval-engine": {
"asserted_class": "issuance",
"requested_observer_max_gap_seconds": 172800,
"observer_registration_status": "pending"
"requested_observer_max_gap_seconds": 7200,
"observer_registration_status": "pending",
"effective_deadline_seconds": 10800
}
}
}
@ -59,14 +60,15 @@
},
"heartbeat": {
"event_class": "audit-core.heartbeat",
"interval_seconds": 86400,
"interval_seconds": 3600,
"assertion": "nothing-to-report",
"missing": "finding",
"extensions": {
"approval-engine": {
"asserted_class": "use",
"requested_observer_max_gap_seconds": 172800,
"observer_registration_status": "pending"
"requested_observer_max_gap_seconds": 7200,
"observer_registration_status": "pending",
"effective_deadline_seconds": 10800
}
}
}
@ -93,14 +95,15 @@
},
"heartbeat": {
"event_class": "audit-core.heartbeat",
"interval_seconds": 86400,
"interval_seconds": 3600,
"assertion": "nothing-to-report",
"missing": "finding",
"extensions": {
"approval-engine": {
"asserted_class": "supersession",
"requested_observer_max_gap_seconds": 172800,
"observer_registration_status": "pending"
"requested_observer_max_gap_seconds": 7200,
"observer_registration_status": "pending",
"effective_deadline_seconds": 10800
}
}
}
@ -127,14 +130,15 @@
},
"heartbeat": {
"event_class": "audit-core.heartbeat",
"interval_seconds": 86400,
"interval_seconds": 3600,
"assertion": "nothing-to-report",
"missing": "finding",
"extensions": {
"approval-engine": {
"asserted_class": "revocation",
"requested_observer_max_gap_seconds": 172800,
"observer_registration_status": "pending"
"requested_observer_max_gap_seconds": 7200,
"observer_registration_status": "pending",
"effective_deadline_seconds": 10800
}
}
}
@ -191,6 +195,16 @@
"status": "candidate",
"schema_sha256": "6d52692eb1e4d1325e0d6062d95acded98beddaae5bcfe24c6f0e91f1b6be6ae"
}
},
"netkingdom": {
"coordination_frame": {
"ref": "coordination-engine/config/coordination-frame.json",
"sha256": "d15e092387f128ffdfd00b9815707a05ff77c5bb0ba4d9a428fbc37b8c440b67",
"frame_seconds": 3600,
"heartbeat_frames": 1,
"registered_wait_frames": 2,
"effective_deadline_frames": 3
}
}
}
}

View file

@ -32,6 +32,8 @@ spec:
app.kubernetes.io/name: approval-engine
template:
metadata:
annotations:
coordination.coulomb.social/frame-sha256: "d15e092387f128ffdfd00b9815707a05ff77c5bb0ba4d9a428fbc37b8c440b67"
labels:
app.kubernetes.io/name: approval-engine
spec:
@ -72,6 +74,8 @@ spec:
- approval-engine
- --jwks-url
- http://keycape.sso.svc.cluster.local:8080/jwks
- --heartbeat-seconds
- "3600"
- --audit-url
- http://audit-core.audit-core.svc.cluster.local:8080
- --audit-token-file

View file

@ -35,15 +35,18 @@ whichever class happens to be busy and says nothing about the quiet one — here
that is `revocation`, whose silence is the only silence that matters. Requested
`heartbeat_classes`, all four at the same bound:
| Class | Longest gap that is not yet a finding |
| Class | Registered wait before receiver grace |
| --- | --- |
| `issuance` | 172800s (48h) |
| `use` | 172800s (48h) |
| `supersession` | 172800s (48h) |
| `revocation` | 172800s (48h) |
| `issuance` | 2f = 7200s (2h) |
| `use` | 2f = 7200s (2h) |
| `supersession` | 2f = 7200s (2h) |
| `revocation` | 2f = 7200s (2h) |
Twice the deployed `--heartbeat-seconds` default of 86400, so one missed cycle
is not a finding and two consecutive ones are. The engine emits all four in a
These are registered base waits, not the effective alert threshold. Founder
coordination-frame decision 2026-10-15 sets f=3600s. The prepared emitter
interval is f; the receiver wait is 2f=7200s, and Audit Core's existing 1.5
grace gives an effective deadline of 3f=10800s. This source request is not a
live registry receipt or proof that deployment has changed pace. The engine emits all four in a
single transaction: a partial emission would report some classes healthy and
others stalled, which is a worse signal than none.
@ -83,3 +86,12 @@ owners and bounds; this registration does not claim their completion.
First provisioning still requires the platform-owned custody record linking
the receiver's protected sender registry and this mounted credential. No
credential value belongs in this document, Git, Hub or a rollout receipt.
Current timing values derive from coordination-engine/config/coordination-frame.json
through net-kingdom/tools/coordination-frame/render.py. The generated declaration,
consumer constants and deployment pod annotation carry its SHA-256. Existing
credential scope and the native September 15 bearer-revocation proof are
retained; do not repeat that unchanged proof as if it were missing. Actual
lifecycle reconciliation, class registration, lag findings and operational
observer readback remain APPROVAL-WP-0002-T04.

View file

@ -21,10 +21,10 @@ becomes a missing positive rather than a quiet month.
| Field | Value |
| --- | --- |
| Class | `heartbeat` |
| Interval | 24 hours (SHOULD also be emittable on demand) |
| Interval | f = 1 hour; SHOULD also be emittable on demand |
| Assertion | `nothing-to-report` |
| Counts | `{issuance, use, supersession, revocation}` of committed outbox rows |
| Missing heartbeat | **finding** after the requested 172800s receiver bound; registration pending |
| Missing heartbeat | **finding** after effective 3f = 3h (registration 2f = 2h plus receiver 1.5 grace); admission pending |
### Reconciliation
@ -48,7 +48,7 @@ case §9.6 names.
## What is a finding
- No per-class heartbeat arrives beyond the requested 48h receiver bound (24h emission interval).
- No per-class heartbeat arrives beyond the effective 3f deadline (f emission interval).
- `audit-core` count for a class is less than this engine's drained count
for that class.
- Outbox depth or age exceeds the lag bound in `cadence.yaml` (drain stuck).
@ -143,3 +143,19 @@ snapshot proves completeness or resistance to source compromise.
This implements T04's missing bounded source computation. Receiver registration,
observed lag/findings and native reconciliation remain blocked owner acceptance;
no existing live image is promoted by this change.
## Configurable coordination frame — 2026-10-15
The timing paragraphs in the dated October 3 migration section describe its
historical daily defaults. Current source defaults and the prepared manifest
are derived from coordination-engine/config/coordination-frame.json, initially
f=3600s by founder decision. The generated approval_engine/coordination.py and
cadence.yaml retain the config digest; CLI and direct OutboxWorker defaults
use FRAME_SECONDS. The deployment argument explicitly supplies the same value
and its pod annotation retains the frame digest. Runtime applies remain under
the owner change gate; no hourly live cadence is claimed by this source edit.
Class mappings, ownership, atomic outbox behavior and count-window accounting
are unchanged. A different frame is regenerated for emitter and observer
settings together using net-kingdom/tools/coordination-frame/render.py.

View file

@ -6,7 +6,7 @@ the workplan remains `blocked` after completing reachable source work.
| Task | Owner return required | Evidence that clears the wait |
| --- | --- | --- |
| T01 — authenticated approver evidence | key-cape and informed-decision; founder for release rollout | Exact deployed issuer revision and client/profile receipt for requester and human paths; human sub/type, tenant/source, resource audience, scopes and assurance captured as non-secret facts and correlated with a stored human-control entry. Promote scanned schema-v6 image under per-change approval, back up/migrate once, then verify native entry and issuance/use provenance. No tokens or secrets in receipts. |
| T04 — delivery and reconciliation | audit-core, with railiance-platform for deployed sender/custody acceptance | Accept current cadence.yaml inventory for source approval-engine and tenant tenant:platform, explicit redact policy and four transition heartbeat classes at the requested 172800s bound. Return actual lifecycle event ids/drain acknowledgment and independent accepted counts for an agreed bounded comparison, accounting for occurrence versus receiver acceptance time, delayed delivery, duplicate acknowledgments and boundary crossings; source pending rows remain lag. Prove startup beats, missing-class and lag findings (depth 100 / age 3600s), and scoped sender bearer revocation. |
| T04 — delivery and reconciliation | audit-core, with railiance-platform for deployed sender/custody acceptance | Accept current cadence.yaml inventory for source approval-engine and tenant tenant:platform, explicit redact policy and four transition heartbeat classes at the frame-derived 2f=7200s registered wait (effective 3f=10800s after receiver grace). Return actual lifecycle event ids/drain acknowledgment and independent accepted counts for an agreed bounded comparison, accounting for occurrence versus receiver acceptance time, delayed delivery, duplicate acknowledgments and boundary crossings; source pending rows remain lag. Prove startup beats, missing-class and lag findings (depth 100 / age 3600s), and retain the delivered September 15 scoped sender bearer-revocation receipt. |
| T05 — live PEP consumption | secrets-engine, with informed-decision for the act-specific human review and access-engine for decision binding | Name the served PEP route/workload, immutable component revision, engine image/endpoint, authenticated caller identities and exact binding. Prove claim-before-decision and CAS consume after ALLOW before protected side effect; same-digest retry, different-digest conflict, spent-on-failure, revoked/expired/superseded/wrong-binding/unavailable denial with no protected callback. Preserve the native approval and decision digests, declared human_control where required, and consumer-owned side-effect/refusal evidence. |
For live whitehat T-06 only, T05's named surface must additionally be covered by
@ -76,3 +76,11 @@ population and reconcile lag/duplicate/boundary cases; matching since/until
strings against occurrence counts is insufficient. Existing owner task bindings
remain key-cape (T01), audit-core (T04) and secrets-engine (T05). No new owner
receipt or named live T-06 surface arrived since the preceding review.
2026-10-15 timing follow-up: founder selected shared coordination f=1h,
replacing a fixed day for this adoption. Generated source/config bindings are
prepared, not deployed. Canonical base is coordination-engine/config/coordination-frame.json;
NetKingdom retains regeneration and drift checks. Lifecycle drain/count-window,
registry admission and independent observation remain open. No credential,
JWT/approval validity, issuer scope or production image pin is changed.

View file

@ -2,6 +2,7 @@ import json
from pathlib import Path
from approval_engine.audit import audit_envelope
from approval_engine.coordination import FRAME_SECONDS, RECEIVER_WAIT_SECONDS
from approval_engine.store import EVENT_CLASSES, HEARTBEAT_CLASSES
ROOT = Path(__file__).resolve().parents[1]
@ -30,8 +31,16 @@ def test_cadence_inventory_matches_durable_emitter(engine):
assert beat["event_class"] == observed[cls]["type"] == "audit-core.heartbeat"
assert beat["assertion"] == observed[cls]["data"]["assertion"]
assert beat["extensions"]["approval-engine"]["asserted_class"] == cls
assert beat["interval_seconds"] == 86400
assert beat["extensions"]["approval-engine"]["requested_observer_max_gap_seconds"] == 172800
assert beat["interval_seconds"] == FRAME_SECONDS
assert beat["extensions"]["approval-engine"]["requested_observer_max_gap_seconds"] == RECEIVER_WAIT_SECONDS
assert entries[cls]["extensions"]["netkingdom"]["rate_monitoring"] == "forbidden"
assert entries[cls]["reconciliation"]["undrained_local"] == "lag-not-divergence"
assert "heartbeat" not in entries["heartbeat"] # transport control has no recursive heartbeat
def test_default_runtime_heartbeat_matches_declared_coordination_frame():
from approval_engine.cli import _parser
args = _parser().parse_args(["serve"])
assert args.heartbeat_seconds == FRAME_SECONDS
assert _parser().parse_args(["serve", "--heartbeat-seconds", "600"]).heartbeat_seconds == 600