Repository navigation
sim-capacity: release every order at exactly submit+latency by default; receipts re-run (old kept as superseded) - #254
Merged
seathatflowsinourveins merged 2 commits intoSep 25, 2026
Conversation
…atency by default
On the pinned NautilusTrader 2.0.0rc5 engine, a deferred order was released
only when its own instrument's next quote, or any due clock timer, was
processed at or after submit + latency -- not exactly at that instant
(measured in sim-engine-crosscheck: extra delay of median 28ms SPY / 53ms
NVDA, p90 135ms/263ms). CapacityExerciserParams.release_alert_latency_ns
registers one otherwise-inert per-order set_time_alert_ns at exactly
submit+latency, fixing this; runner.run_one wires it from a new
exact_release parameter (default True), with --release-timing legacy on the
CLI to reproduce the old behavior on demand.
- exerciser.py: release_alert_latency_ns field + the per-order alert.
- runner.py: exact_release/--release-timing wiring, check_release_timing
(asserted by cmd_run: no fill before submit+latency, none inexact when
exact release is active), release_timing in the receipt.
- tests/test_sim_capacity.py: ExactReleaseTimingTests on a sparse synthetic
quote set, exact mode exactness and legacy mode slack; verified failing
with the alert disabled and passing restored. All 65 sim-capacity tests
pass on system Python (28 skip cleanly) and the pinned runtime (0 skipped).
- Re-ran the paper-parity and elite-tier receipts (with latency sweep) from
retained --replay data: receipts/20260925-*.json. submit-to-fill is
exactly 70.0ms (min=p50=p90=max) at the 70ms primary latency; every
latency-sweep point has zero early/inexact release-timing violations;
fills/min, NBBO-touch (0 violations), recount-agrees and fee totals all
still hold. Prior receipts kept unmodified as
receipts/20260924-*-superseded.json.
- experiment.json: dated 2026-09-25 observations for the new runs and
tests, timing limitation marked resolved (old text kept, correction
appended), qualification_run_ids repointed at the new elite-tier run,
every {path, sha256} pin refreshed. validate_convergence.py
--all-recorded passes for all 22 recorded records.
- manifests/evidence.json: refreshed/added entries, sorted,
json.dumps(indent=2) formatting confirmed byte-identical on reserialize.
validate.py and evidence_manifest.py --check both pass.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
With #248 on main, the dashboard's simulation lane references the cross-check receipt rather than the renamed 20260924 receipt. The grand-dashboard tests that failed on this branch's older base (missing evidence reference) now pass: 190 dashboard and sim tests OK. validate.py, evidence_manifest.py --check and validate_convergence.py --all-recorded pass. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
seathatflowsinourveins
added
the
lane:trading
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
seathatflowsinourveins
deleted the
claude/sim-capacity-exact-latency-20260925
branch
September 25, 2026 14:55
seathatflowsinourveins
pushed a commit
that referenced
this pull request
Sep 25, 2026
…0260925 Two use-stage receipts, native_proven, recorded with scripts/host_receipts.py at origin/main 720e294 on the WSL2 Threadripper/RTX 4090 host from #231, each bound to its layer winner's current pin: - socraticode 1.14.0 (foundation/semantic-rag): through mcporter, the health check reaches Qdrant and the vLLM-served Nemotron embedder, the index of a separate clone of this repository is green, a gate-ladder question returns scripts/trading_gates.py and catalogs/us-equities/gates-20260922.json in the top eight, and an unrelated control query returns neither. The -2 generation supersedes the first recording: it drops an unestablished "read-only clone" wording and asserts the launched package version against the winner pin. - ai-memory 2.3.2 (foundation/durable-memory): read-only status of the host service (counts and flags only), then write, search, read-back and delete of a scoped probe page on a disposable scratch server with its own data directory; after deletion the same search finds nothing and read-page fails. The scratch server and directory are removed. Rebased onto d2daa16 (#254) with the docs/lanes.md hot-file protocol: main's manifests/evidence.json plus the three receipts re-registered; the component evidence matrix is regenerated with main's component_matrix.py, and the new-host grand list is unchanged by these receipts. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This was referenced Sep 25, 2026
seathatflowsinourveins
added a commit
that referenced
this pull request
Sep 25, 2026
… incentive board and data lake, and the recounted WSL host (#288) - simulation: 866 to 884 fills per min at 70 ms exact release, 30 of 30 minutes, and 672 at 1 s (#254, sim-capacity/receipts/20260925-elite-tier.json). - paper: adds the margin financing model (#243); still blocked on the paper keys (HTTP 401). - mover-early-entry-studies: v1 to v3, with the v3 core merged (#190) and not frozen. - New gates: - incentive-board-forward (#187, #220): 0 orders; - data-lake-source (#208): minute-bar pilot not run. - wsl-workstation-host: the foundation lane's wording. Counts were recounted on main 0074a0c by the last independent_session verdict per receipt, superseded excluded: 69 receipts, 56 agree, 1 needs changes, 0 awaiting review, 12 superseded (#285). The foundation lane acked the strings and the recount rule. Co-authored-by: Scout <scout@local> Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
The sim-capacity NautilusTrader runs now release every latency-delayed order at exactly submit + configured latency, and the committed receipts are re-run with that as the default. The old behaviour stays available behind an explicit switch (
--release-timing), and both earlier receipts are kept byte-identical as*-superseded.json.Why
The engine cross-check (#245) measured a NautilusTrader 2.0.0rc5 timing property. A delayed order is released only when the next quote on its own instrument, or some other due timer, is processed at or after submit + latency. That added a median of 28 ms (SPY) and 53 ms (NVDA) beyond the configured latency, and a p90 of 135 ms and 263 ms.
The tested fix is one otherwise-inert
set_time_alert_nsper order at submit + latency. It makes every order resolve at exactly the configured latency, matching hftbacktest's timing in #245.Results
Elite latency sweep (new). Every point clears 180 in all 30 minutes, and every exact-release check shows zero early and zero inexact releases.
Paper parity is rate-limited to 180 per minute by its own
RiskEngineConfig. Its predeclared rule is that it must never exceed 180 per minute (max 180), not that it must sustain it. This rule is unchanged, andexperiment.jsonrecords it.Checks
tests/test_sim_capacity.pyhas 65 tests, 3 of them new (ExactReleaseTimingTests).experiment.jsonhas 3 new dated observations. The timing limitation is marked resolved, with the old text kept and a correction appended. Its pins cover the new and superseded receipts.validate.py,evidence_manifest.py --checkandvalidate_convergence.py --all-recordedpass on the tree with main merged in.test_grand_dashboard,test_native_dashboard_data,test_new_host_grand_list,test_sim_capacity,test_sim_engine_crosscheck): 190 tests OK with main merged in.Evidence class
Infrastructure evidence only: a simulation capacity replay on retained SIP quotes. It is not broker throughput and not strategy performance (NS-06).
Review and process
/usr/bin/git. The commits are secret-scanned clean, and such agent types are no longer used here.🤖 Generated with Claude Code