Liquidity operations cockpit
Provider ledgers, live EWMA forecasts, and human coordination
Connecting
Historical Context
— warming up —
Live evidence
judge-verifiableEvery value below is sourced live from /v1/telemetry/snapshot — the same payload that drives the cards above. Nothing is hardcoded.
- Last event id
- #0
- Sim time
- —
- Window
- —
- Has evidence
- warming up
- Transactions
- —
- Drain / min
- —
- Consistency
- —
- As of
- —
Historical CTE — backend SQL
-- historical.shared_cash: returns the aggregated rollup that the
-- backend stores under historical_analytics.shared_cash in the snapshot.
WITH raw AS (
SELECT id, sim_time, delta_bdt
FROM shared.shared_cash_movement
WHERE agent_id = :agent_id
AND sim_time >= :cutoff
AND sim_time <= :as_of
ORDER BY sim_time DESC, id DESC
LIMIT :row_cap
), balance_points AS (
SELECT id, sim_time,
(SELECT balance_bdt FROM shared.shared_cash_ledger
WHERE agent_id = :agent_id)
- COALESCE(
sum(delta_bdt) OVER (
ORDER BY sim_time DESC, id DESC
ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING
), 0
) AS balance_after
FROM raw
), daily AS (
SELECT balance_after,
row_number() OVER (
PARTITION BY (sim_time AT TIME ZONE 'UTC')::date
ORDER BY sim_time DESC, id DESC
) AS daily_rank
FROM balance_points
)
SELECT count(*) AS transaction_count,
avg(-delta_bdt) FILTER (WHERE delta_bdt < 0) AS average_outflow,
avg(delta_bdt) FILTER (WHERE delta_bdt > 0) AS average_inflow,
sum(delta_bdt) AS net_delta,
extract(epoch FROM (max(sim_time) - min(sim_time))) / 60.0
AS span_minutes,
stddev_pop(abs(delta_bdt)) AS amount_stddev,
avg(abs(delta_bdt)) AS average_absolute_delta,
(SELECT avg(balance_after)
FROM daily WHERE daily_rank = 1) AS average_daily_balance,
(SELECT balance_bdt
FROM shared.shared_cash_ledger
WHERE agent_id = :agent_id) AS current_balance
FROM raw;Received — · connecting
Live test scenario
Inject synthetic transactions
Uses the normal ledger, EWMA, anomaly, alert, and SSE pipeline. The salary-period option is a simple calendar-aware heuristic—not solved seasonality—and every result still requires human review.
Forecast inputs
Live EWMA remains the prediction engine; committed history adds context.
- Historical window
- Loading…
- Live window
- 12 minutes
- Historical records
- Loading…
- Prediction model
- EWMA + historical context
Shared physical cash
—
Awaiting snapshot
bKash e-money
—
TTE calculating
Nagad e-money
—
TTE calculating
Rocket e-money
—
TTE calculating
Live provider balance trajectory
Every point is a persisted balance returned by the simulation engine.
Waiting for provider balance updates — no synthetic chart data is shown.
EWMA forecast evidence
Forecasts appear after enough real balance deltas have entered the EWMA window.