trade_log, newest first. Telemetry streams (shadow exits, state flips, prefilter rejects) excluded — see the Logs tab for the full stream.
Sizes are data + index from information_schema; "physical" adds the free space inside each tablespace, which the volume still pays for. Unmeasured values read UNKNOWN. Nothing on this page deletes anything.
The enforced set is UNKNOWN. The estimate is recomputed every sweep; moving RETENTION_MODE is the operator's call after reading it.
| Account | Tier | Mode | Auto-Arm | Open | Sessions | Last seen | Created |
|---|
trade_log — every persisted application event (buys, sells, Smart Entry/Exit state changes, withdrawals, breaker trips, operator actions). Click a row to expand its payload.
| When | Action | Token | User | Summary |
|---|
Pre-launch kill switches (lib/feature-flags.js). Env-driven — change the variable and restart.
lib/tiers.js — what each plan grants. Read-only here; edits are a code change.
Whitelisted, non-secret variables as this process sees them. Unset = the code's default applies.
Presence only. Values never leave the server through this page.
Accounts with users.is_operator = 1. Grant/revoke from a user's detail panel.
Accounts with users.is_tester = 1: elite limits and every unreleased feature, but no access to this console. Grant/revoke from a user's detail panel.
Read-only funnel report: is Smart Entry actually evaluating candidates, and where is it rejecting them. Not a tuning verdict — a dropped fire rate can be a real gate change working as intended.
Candidates evaluated vs. gate-open ticks vs. actual fires, newest first. A funnel narrowing much faster than usual is the first sign something changed.
First unmet fire condition per tick (lib/candidate-paths.js describeBlocker). A reason that jumped is worth asking whether that gate change was intended.
The same ticks, classified by lib/reason-taxonomy.js instead — a different cut of the same funnel.
Active in the last 10 minutes, gate closed on their latest tick, oldest first — the "strong scores but never fires" case. Reads empty when nothing is currently armed.
Operator-only flags — none of these change any threshold or decision, they only say the pipeline may need a look.
Forward, same-candidate-stream shadow comparison — infrastructure only. No production variant winner is recommended here; there is not yet enough forward data to call one.
Each of the 5 strategy arms (V2, V5, V6, V8, V4_FIXED) plus the real production gate, evaluated against the same candidate ticks. Avg MFE/MAE are the 5-minute horizon, averaged only over evaluations with an outcome filled in yet — not treated as zero when missing.
CURRENT/EARLY/RECLAIM entry-timing arms, scoped to candidates that already pass the same hard safety gates as production. EARLY/RECLAIM thresholds are first-pass, unvalidated starting definitions — see lib/smart-entry-shadow.js.
Daily rollup from this harness's own recorded ticks — not real-time.
"Resolved entries" and "Evidence level" answer whether there is enough forward data yet — see the disclaimer below the table. Evidence stages are OPERATIONAL thresholds (INSUFFICIENT / EARLY SIGNAL / PRELIMINARY / USEFUL FORWARD EVIDENCE), never a statistical-significance or win-probability claim.
Read-only monitoring of the research harness itself — never affects trading. See lib/smart-entry-health.js for trading-pipeline health, a separate panel.