VIOLET_PART_SPEC_OA_TODO.md carves the self-contained pending items from the VIOLET plan so a
separate agent can work without colliding with in-flight files:
1. Parity-pin tests for sizing.py's hand-replicated arithmetic (regime/ob/strength/dc) vs the
real NDAlphaEngine methods — test-only.
2. Multi-exchange OB provider seam (VioletVenueOBProvider scaffold + interface + design note)
— the BingX/3rd-party swap point, NOT wired live.
3. Base-fraction sizing study (analysis script + report) — pure read-only analysis, no
behavior change.
Each task: exact affected (new-only) files, V-TYPES + DARK + never-touch-shared rules, required
tests + pass criteria. Includes an explicit COMMIT/BRANCH policy (own git worktree, or partial
commits with `git commit -- <paths>`) prompted by today's three shared-index collisions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Launcher shadow path now keeps the persistent state BLUE keeps across scans (faithful, not
per-scan single-shot): ONE OBFeatureEngine wired lazily on first assets and kept for the
service lifetime, reading BLUE's EXTANT published OBF feed via HZOBProvider (read-only HZ
entry-listener cache — NO new OB storage; provider is the swap seam for a future direct
BingX/3rd-party OB stream); per-scan ob_bar_idx into step_live/get_market; and prior_boost_beta
carried across scans (stale exf keeps the prior). shadow_decision_step threads ob_engine +
bar_idx + prior into source_live_blue_sizing_factors; build_shadow_live_source seeds the state
+ a default ob_engine_factory (injectable for tests).
TODO_HZBRIDGE markers added at all 3 VIOLET->Hazelcast touch points: per operator the upcoming
dolphinng5_predict/hzbridge must become the sanctioned HZ connection (silent client-death /
lockup / dropout mitigation) — refactor ASAP once it ships.
Doc VIOLET_OB_FEED_AND_AGENT_COORDINATION.md: OB-feed sourcing doctrine, HZ-bridge TODO, and
the multi-agent worktree + doctrine/release/status scheme (prompted by the shared-index
incidents where agents' staged files cross-contaminate commits).
6 launcher tests (+wired-once/bar_idx-increments, +prior-carries). violet-only; partial commit
(only my 4 paths) to avoid sweeping a concurrent agent's staged files.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Operator directive: VIOLET must do IDENTICALLY what BLUE does for the three parity
flags — approximation cannot guarantee bit-for-bit functioning. Reworked
live_blue_source.py to call BLUE's OWN code paths, not reconstruct/substitute them.
boost/beta — was reading the published DOLPHIN_FEATURES.acb_boost scalar
(acb_processor_service's daily value). BLUE's trader does NOT use that for sizing; it
recomputes live via acb.get_dynamic_boost_from_hz(exf_latest, w750_velocity, direction)
with a bare AdaptiveCircuitBreaker() and NO ob_engine (nautilus_event_trader.py
on_exf_update:4769 / rollover prewarm:2710). New _source_boost_beta replicates that call
exactly (reads exf_latest + latest_eigen_scan.w750_velocity; 0.0→None like BLUE; on stale
exf ValueError → neutral, mirroring BLUE's "ACB Stale Data Fallback"). The published
acb_boost is never read. Test pins bit-identity against the real ACB.
signal-gen (dc_status) — was AlphaSignalGenerator() bare defaults; coincidentally equal to
BLUE today, but BLUE builds it from ENGINE_KWARGS (trader:128-133, threaded at
esf_alpha_orchestrator.py:180-191), so a champion retune would silently diverge. Now
constructed with BLUE_SIGNAL_GEN_KWARGS (vel_div_* imported from the kernel constants).
Test parses ENGINE_KWARGS from the trader source and asserts each param matches — drift
becomes a red test, not a silent miss.
OB — was a reinvented HazelcastOBProvider reading asset_*_ob with custom parsing. Now uses
BLUE's OWN HZOBProvider + OBFeatureEngine, wired exactly as _wire_obf
(nautilus_event_trader.py:4967-4980): step_live(assets, bar_idx) then get_market. Engine
is injectable + persistent so OB accumulation matches BLUE across scans (caller owns it).
Deleted: HazelcastOBProvider, _extract_acb, the status-label mc path. mc_scale fix
(begin_day cat/env thresholds) retained. Module docstring + structural-divergence doc
updated: all three flags FIXED; only _derive_mc_scale remains hand-replicated (pinned by
formula test). OPEN follow-up: launcher shadow_decision_step should pass a persistent
ob_engine + bar_idx for cross-scan OB history.
34 tests (33 + live-HZ smoke deselected): boost/beta-vs-ACB bit-identity (incl. w750=0→None,
no-exf, stale ValueError, ignores acb_boost), signal-gen ENGINE_KWARGS pin (parametrized),
OB wiring (step_live call, HZ coords, neutral paths), mc_scale formula, sequence dc/selector
parity, anomaly handling. violet-only; no shared-file edits.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
REVIEW of the V3.4c/d/e work (other agents) for BLUE-algo compliance found one
real bug and one architectural concern.
BUG FIXED — mc_scale derivation (live_blue_source.py):
The adapter mapped `mc_forewarner_latest.status == "ORANGE"` → 0.5. But BLUE's sizing
mc_scale is NOT the MC service's `status` label (that label, from mc_forewarner_flow.py,
uses GREEN<0.10/ORANGE<0.30/RED — observability only). The live trader re-derives the
haircut in begin_day (esf_alpha_orchestrator.py:956-962) from the SAME published fields
`catastrophic_prob` + `envelope_score` with DIFFERENT thresholds:
mc_red = cat>0.25 or env<-1.0
mc_orange = (not mc_red) and (env<0 or cat>0.10)
mc_scale = 0.5 if mc_orange else 1.0
The two disagree (e.g. cat=0.05/env=-0.5 → label GREEN→1.0 but BLUE orange→0.5;
cat=0.28 → label ORANGE→0.5 but BLUE red→1.0). Rewrote `_derive_mc_scale` to mirror
begin_day exactly on the source fields. Per operator: there is genuine ambiguity over
which surface is "nominal" — we go with the SOURCE FIELDS + begin_day formula (the path
that actually drives BLUE's sizing); the ambiguity + decision are noted in the docstring.
TESTS — the old fixtures were complicit: they fed `{"status":"ORANGE"}`, a payload shape
BLUE never emits, so they "passed" against fiction. Replaced with BLUE's real payload
(`catastrophic_prob`/`envelope_score`) and added a parametrized formula test covering the
exact divergence cases the status-based code got wrong, plus missing/garbage-field neutral.
15 passed (live-HZ smoke deselected).
ARCHITECTURAL DEBT documented — VIOLET_BLUE_PARITY_STRUCTURAL_DIVERGENCE.md:
VIOLET imitates BLUE's computations in a DIFFERENT module/scope structure. Kernels are
WRAPPED (safe), but orchestration arithmetic (compose/regime/ob/strength/mc_scale) is
HAND-REPLICATED out of the monolithic NDAlphaEngine — making orderly, verifiable BLUE↔VIOLET
parity comparison and refactoring hard. Doc inventories every re-derivation with BLUE's
file:line authority, the drift risk, and mitigations (parity-pin tests; single canonical
surface per ambiguous factor; eventual DITAv2 Rust-backplane convergence). Pointer added to
live_blue_source.py's module docstring.
Also reviewed (no change needed): launcher launch_dolphin_violet.py is DARK-safe
(ObserveOnlyVenue + shadow gated default-OFF, no order path); shadow_live_factors.py glue
is correct; trade_slot_compare.py reimplements no BLUE algo. fb34431's launcher tests —
which the authoring agent never ran on the slow mount — pass (14 green).
violet-only; no shared-file edits.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#1 (validation) — VIOLET_V34B_LIVE_FACTOR_FIELD_VALIDATION.md documents the
ground truth from BLUE's own code (esf_alpha_orchestrator / adaptive_circuit_breaker
/ nautilus_event_trader): of the 8 sizing inputs, ONLY `posture` is a flat HZ key.
`esof_score` is in HZ but as a payload to parse. The other five — boost, beta,
mc_scale, ob_median_imbalance, ob_agreement_pct, dc_status — are BLUE-organ outputs
(ACB over DOLPHIN_FEATURES.exf_latest, MC flag derivation, OBFeatureEngine, the
per-asset signal generator) and are NOT present as scalars in any HZ map. This
inverts live_factors.py's flat-snapshot premise; its speculative alternate paths
(acb_boost / s_acb_boost / ("acb","boost") / …) match nothing real. The full live
sourcing of the five is a multi-organ sprint (V3.4c), not a HZ scrape.
#2 (sourcing adapter) — live_factor_source.py sources what BLUE actually publishes,
read-only and pure (callers pass already-fetched HZ blobs; no client, no I/O):
- posture ← engine_snapshot['posture'] (DOLPHIN_STATE_BLUE), default APEX
- esof_score ← DOLPHIN_FEATURES['esof_latest'] via BLUE's OWN parse_esof_payload
+ esof_score_from_payload (wrap, don't reimplement; staleness gate
honored when max_age_s supplied)
- boost/beta/mc_scale/ob_*/dc_status ← BLUE's neutral sentinels (1.0/0.0/1.0/None/
None/"NONE") until V3.4c — explicit, never silently faked.
Flows through the validated extract_live_sizing_factors normalizer. ORGAN_DERIVED_
FACTORS names the six deferred to V3.4c so the journal can mark them NEUTRAL.
9 new tests green (posture default/upper, esof dict+raw-JSON+staleness, neutral
integration, all-neutral-when-empty). violet-only; no shared-file edits; no soak.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_DEV_SPEC_AND_PLAN.md: authoritative consolidated plan (mission, doctrine,
V0->V6 ladder w/ status, code map, full sizing composition, next steps, vision,
TODOs, operational notes). Supersedes scattered plan files.
VIOLET_SUB_SPEC__L3_EXCHANGE_LEVERAGE.md: self-contained parallel-developable unit
(V3.5) for an independent agent — wrap prod/bingx/leverage.py conviction->exchange
mapping, V-TYPES + bit-identity gate, full file paths/tests/gates/acceptance. Zero
overlap with V3.4 (DecisionEngine<->Sizing integration, lead-owned).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Repo cwd /mnt/dolphinng5_predict (git root), no remote (local-only) -> agent must
run on this host in this dir; needs host-local eigenvalues data, live ClickHouse,
and BLUE runtime for bit-identity. Adds CH creds + python path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
§8b: authoritative orchestrator composition (5 multipliers: base x dc_lev_mult x
regime_size_mult[ACB x meta x MC] x market_ob_mult x esof; dynamic cap, STALKER 2.0)
+ operator factor-recall map (DC boost + OB-consensus = the two aside ACBv6).
§8c: vision roadmap — five-lanes=separation-of-concerns reframe, LONG-alpha grail,
FPGA-pure VIBRISS banditry, pure-dataflow-DAG->compile nirvana w/ bit-identity as
bridge, culminating in DISTRACK (memory-constant streaming distributions, sequenced
AFTER live testnet->mainnet).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Logs the open question (noise/denominator artifact vs real depeg micro-edge
deliberately self-limited) + how to settle it. VIOLET keeps BLUE's exclusion
regardless (follow BLUE).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
parity_harness.py: median-curve parity of V3a VioletBetSizer vs recorded
dolphin.trade_events (vel_div->leverage), restricted to short-signal domain.
GATE PASSES on prod host: pearson 0.9998, max_abs_err 0.238 (budget 1.0) over
23 bins -> base conviction sizer reproduces BLUE's central tendency. Per-trade
scatter is the deferred SC/ACB/OB/gold modulation layer (separate finding doc).
3 unit + 1 gate green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds dolphin-supervisord.service (installed + enabled) and dolphin_startup_check.sh:
- ExecStartPre waits for HZ (CRITICAL/blocks), CH+Prefect (WARN/degraded-ok)
- Logs to /tmp/dolphin_logs/startup.log + run_logs/dolphin_startup_<date>.log
- Writes machine-readable /tmp/dolphin_logs/startup_status.json on every start
- nautilus_trader remains autostart=false — BLUE must be started manually
SYSTEM BIBLE bumped to v7.1; §16.10 updated, §16.14 added.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add §16.10 corrected daemon start sequence (supervisord NOT auto-started on boot),
§16.12 critical supervisord.conf rules (no /tmp paths, OBF starvation → BLUE freeze,
pre-restart position check), §16.13 OOM recovery runbook with exact commands.
Incident context (2026-06-08):
- Previous agent set nautilus_trader to /tmp/blue_runtime_mirror/ — broken after OOM reboot
- OBF died during BLUE run, degraded gate for 285+ bars, BLUE stuck in RETRACT on LTCUSDT
- Fix: revert supervisord.conf to /mnt canonical paths, restart supervisord
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Sprint 2 (accounting + observability parity, PINK scope):
- Verified pink_clickhouse.py writes the 8 BLUE-legacy row families at
matching schema and that capital authority in pink_direct.step() is
solely kernel.account (no balance-poll overwrite in the hot loop).
- Report: prod/clean_arch/dita_v2/SPRINT2_ACCOUNTING_PARITY.md.
Sprint 3 offline groundwork (no exchange contact):
- Add _write_trade_exit_leg to pink_clickhouse.py: one BLUE-schema-faithful
trade_exit_legs row per exit leg, with isolated (non-cumulative) per-leg
deltas tracked via _leg_state (reset on ENTER). Closes the docstring gap.
- New offline suite test_pink_multi_exit_groundwork.py (3 passed):
* Flaw 4 — two-leg exit closes once, realized accrues per leg, closed
slot rejects further EXIT (no double-close).
* Overshoot invariant — a final EXIT requesting more than the remaining
size CLAMPS (size to 0, no oversell), retiring the Sprint 0 cumulative-
ratio risk empirically.
* trade_exit_legs delta + full BLUE column-set assertions.
- Persistence regression after edits: 10 passed.
BLUE untouched: no changes to dolphin.* / DOLPHIN_*_BLUE / nautilus_event_trader.py.
Live VST multi-leg run remains deferred pending explicit authorization.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
