jgrusewski 4ebaea784a fix(factory): trial-count honesty — stable key, count-before-mark, LLM param guards ...

Fix 1 (CRITICAL): Proposal.__post_init__ coerces all param values to float so
{"window": 20} and {"window": 20.0} produce the same key(); key() uses :.10g
format for stable float rendering. Prevents the same hypothesis being counted
twice when analysts use int vs float params.

Fix 2 (CRITICAL): AnalystFleet.gather restructured to count-before-mark: fresh
proposals are collected without marking, add_trials(len(fresh)) fires FIRST, then
mark_seen loops. A crash between count and mark causes a transient over-count
(bar slightly high, safe) instead of a permanent under-count (bar low → false
discoveries reach capital).

Fix 3 (IMPORTANT): LlmAnalyst.propose rejects lines with empty kind (|| guard)
and drops non-finite param values (inf/nan) via math.isfinite; replaces the
contextlib.suppress pattern with explicit try/except + isfinite check.

Fix 4 (tests): test_cross_analyst_dedup, test_cross_cycle_dedup_does_not_recount,
test_llm_drops_empty_kind_and_nonfinite added to test_analyst_fleet.py.
New file tests/unit/test_proposal_key.py covers int==float key stability and
distinct-param separation. 224 passed, ruff clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-14 20:35:31 +02:00

..

test_binance_ccxt_helpers.py

fix(exec): clear remaining minors — zip strict, rate-budget margin test, mid-sizing rationale

2026-06-14 17:03:11 +02:00

test_binance_settings.py

feat(exec): crypto dep + BinanceSettings (safe defaults: testnet on, live off)

2026-06-14 16:45:01 +02:00

test_capacity.py

fix(exec): fail-safe risk gates — NaN/misconfig guards (no fail-open to max exposure)

2026-06-14 17:11:58 +02:00

test_charts.py

feat(cockpit): dependency-free inline-SVG nav sparkline

2026-06-14 11:47:19 +02:00

test_correlation.py

fix(factory): fail-safe diversification core — HRP uniform fallback, corr epsilon + empty guard, gate reason

2026-06-14 20:06:52 +02:00

test_crypto_exchange_dtos.py

feat(exec): crypto_exchange port (DTOs + Protocol)

2026-06-14 16:45:36 +02:00

test_crypto_fetch.py

fix(fetch): clean missing-key error + day-aligned funding array

2026-06-14 12:30:07 +02:00

test_cusum.py

feat: CUSUM structural-break detector ported (regime #6) — the edge-decay signal

2026-06-13 17:56:19 +02:00

test_div_gate.py

fix(factory): fail-safe diversification core — HRP uniform fallback, corr epsilon + empty guard, gate reason

2026-06-14 20:06:52 +02:00

test_dsr_scales_with_n.py

test(factory): lock DSR bar rises with trial count (FST immune system)

2026-06-14 19:54:57 +02:00

test_factory_settings.py

feat(factory): scipy dep + FactorySettings (safe diversification/governance defaults)

2026-06-14 19:52:26 +02:00

test_futures_fetch.py

feat(fetch): port futures fetcher into fxhnt (Databento, cost-gated, offline-testable)

2026-06-14 12:23:09 +02:00

test_gate.py

feat(cockpit): Phase-1 go/no-go gate (pure, reports-only)

2026-06-14 11:33:44 +02:00

test_gauntlet.py

feat: fxhnt foundation — hexagonal architecture + proven gauntlet + vertical slice

2026-06-09 12:21:47 +02:00

test_governance.py

feat(factory): promotion governance (caps + kill-switch budget)

2026-06-14 20:08:35 +02:00

test_hrp.py

fix(factory): fail-safe diversification core — HRP uniform fallback, corr epsilon + empty guard, gate reason

2026-06-14 20:06:52 +02:00

test_hurst.py

feat: Hurst exponent (R/S) ported — the persistence signal for regime classification

2026-06-13 19:09:08 +02:00

test_marginal.py

feat(factory): marginal Sharpe contribution (selection objective)

2026-06-14 19:54:08 +02:00

test_microprice.py

feat: SSOT feature-warehouse foundation + first feature TDD-ported (microprice)

2026-06-13 02:01:04 +02:00

test_ofi.py

feat: OFI features TDD-ported (Blocks F/G/H/CC) — the follow-the-money signal

2026-06-13 02:06:30 +02:00

test_periods.py

feat(cockpit): period-aware aggregation + rolling mean (raw, no normalization)

2026-06-14 11:45:42 +02:00

test_portfolio.py

feat: multi-strategy execution layer (domain netting + Broker port + ExecutionService)

2026-06-09 13:52:32 +02:00

test_proposal_key.py

fix(factory): trial-count honesty — stable key, count-before-mark, LLM param guards

2026-06-14 20:35:31 +02:00

test_reconciliation.py

fix(exec): capital-risk fixes — authoritative IOC sizing, accumulate all legs, θ=0 hold, true arrival mid, dry-run gate report, rate-budget window

2026-06-14 17:29:37 +02:00

test_regime.py

feat: regime layer — the foundation of the agentic strategy factory (ADR 0002)

2026-06-09 19:20:23 +02:00

test_risk_gates.py

fix(exec): fail-safe risk gates — NaN/misconfig guards (no fail-open to max exposure)

2026-06-14 17:11:58 +02:00

test_rounding.py

feat(exec): pure exchange-filter rounding (step/tick/min-notional)

2026-06-14 16:46:05 +02:00

test_smart_router.py

fix(exec): stable data-derived rebalance_id + rate-limit signal + cover/drift invariant tests

2026-06-14 17:39:29 +02:00

test_state_reader.py

fix(cockpit): DATE column for hypertable + sample-variance welford + test import order

2026-06-14 11:43:47 +02:00

test_target_weights.py

feat(exec): latest_target_weights — current per-symbol targets for execution

2026-06-14 16:46:45 +02:00

test_tca.py

fix(exec): capital-risk fixes — authoritative IOC sizing, accumulate all legs, θ=0 hold, true arrival mid, dry-run gate report, rate-budget window

2026-06-14 17:29:37 +02:00

test_validation.py

feat: López de Prado validation suite ported (DSR + PBO + permutation + FDR) — the honesty backbone

2026-06-13 02:24:41 +02:00