direct customers by region for July 2026 and August 2026 month-end snapshots
Cost was recorded but no model name was. This happens when the answer came from an approved query — that path returns before the model name is stamped — so the cost figure is correct while the model behind it is not identified here. Cost was still priced per call at the right rate.
Basis: UK, APAC, Canada +4 more, 2026-07-01 to 2026-08-01. Governed filters applied. Figures as of 14:31 UTC — reused from an earlier identical question, not recomputed just now.
Per-step detail — each LLM call, each Omni request, each rejected draft —
is traced to stdout only and is not stored, so it cannot be shown here.
Query attempts counts run_query executions, not model turns.
WITH booking_snapshot_grain AS (
SELECT
omni_dbt__eds_calendar_day."DATE_DAY" AS "Snapshot Date",
omni_dbt__eds_all_accounts."DERIVED_ACCOUNT_ID" AS "Customer Account ID",
omni_dbt__eds_all_accounts."FINANCE_COUNTRY_GROUPING" AS "Region",
omni_dbt__eds_bookings."PRIMARY_KEY" AS "Booking Primary Key",
MAX(omni_dbt__eds_bookings."HAS_PAID_SUBSCRIPTION") AS "Has Paid Subscription"
FROM PROD."EDS_ALL_ACCOUNTS" AS omni_dbt__eds_all_accounts
INNER JOIN PROD."EDS_CALENDAR_DAY" AS omni_dbt__eds_calendar_day
ON omni_dbt__eds_calendar_day."DATE_DAY" >= omni_dbt__eds_all_accounts."VALID_FROM" AND omni_dbt__eds_calendar_day."DATE_DAY" <= omni_dbt__eds_all_accounts."VALID_TO"
LEFT JOIN PROD."EDS_BOOKINGS" AS omni_dbt__eds_bookings
ON omni_dbt__eds_all_accounts."DERIVED_ACCOUNT_ID" = omni_dbt__eds_bookings."DERIVED_ACCOUNT_ID" AND omni_dbt__eds_calendar_day."DATE_DAY" >= omni_dbt__eds_bookings."VALID_FROM" AND omni_dbt__eds_calendar_day."DATE_DAY" <= omni_dbt__eds_bookings."VALID_TO"
WHERE omni_dbt__eds_calendar_day."DATE_DAY" >= '2026-07-01'
AND omni_dbt__eds_calendar_day."DATE_DAY" < '2026-09-01'
AND omni_dbt__eds_calendar_day."DATE_DAY" = DATE_TRUNC('month', omni_dbt__eds_calendar_day."DATE_DAY" + INTERVAL '1 month') - INTERVAL '1 day'
AND omni_dbt__eds_all_accounts."FINANCE_ACCOUNT_TYPE" = 'Corporate'
AND omni_dbt__eds_all_accounts."FINANCE_COUNTRY_GROUPING" IS NOT NULL
AND COALESCE(omni_dbt__eds_all_accounts."IS_DEXT_DEMO", FALSE) = FALSE
AND omni_dbt__eds_all_accounts."FRANCHISE_NAME" IS NULL
AND (omni_dbt__eds_all_accounts."SALESFORCE_ACCOUNT_TYPE" != 'Reseller' OR omni_dbt__eds_all_accounts."SALESFORCE_ACCOUNT_TYPE" IS NULL)
GROUP BY
omni_dbt__eds_calendar_day."DATE_DAY",
omni_dbt__eds_all_accounts."DERIVED_ACCOUNT_ID",
omni_dbt__eds_all_accounts."FINANCE_COUNTRY_GROUPING",
omni_dbt__eds_bookings."PRIMARY_KEY"
),
direct_customer_snapshot AS (
SELECT
"Snapshot Date" AS "Snapshot Date",
"Region" AS "Region",
"Customer Account ID" AS "Customer Account ID"
FROM booking_snapshot_grain
GROUP BY
"Snapshot Date",
"Region",
"Customer Account ID"
HAVING MAX(CASE WHEN "Has Paid Subscription" = TRUE THEN 1 ELSE 0 END) = 1
)
SELECT
DATE_TRUNC('month', "Snapshot Date") AS "Month",
COUNT(DISTINCT CASE WHEN "Region" = 'UK' THEN "Customer Account ID" END) AS "UK",
COUNT(DISTINCT CASE WHEN "Region" = 'APAC' THEN "Customer Account ID" END) AS "APAC",
COUNT(DISTINCT CASE WHEN "Region" = 'Canada' THEN "Customer Account ID" END) AS "Canada",
COUNT(DISTINCT CASE WHEN "Region" = 'US' THEN "Customer Account ID" END) AS "US",
COUNT(DISTINCT CASE WHEN "Region" = 'France' THEN "Customer Account ID" END) AS "France",
COUNT(DISTINCT CASE WHEN "Region" = 'Other' THEN "Customer Account ID" END) AS "Other",
COUNT(DISTINCT "Customer Account ID") AS "Total Direct Customers"
FROM direct_customer_snapshot
GROUP BY DATE_TRUNC('month', "Snapshot Date")
ORDER BY "Month"
LIMIT 2
No feedback linked to this request.
Cost is the LLM completion spend LiteLLM priced for each call, summed per
request.
Chandler is running on the Codex CLI, which draws ChatGPT plan quota
rather than per-token API billing, so this figure is not money paid —
it is what the same traffic would have cost on the API, priced from the
token counts Codex reports. Read it as the size of the bill avoided.
Two known limits: embedding spend is not recorded, so
retrieval and matching cost is missing, and because the figure is one sum
per request it cannot be split by model within a request — a
request's classifier and agent calls can use different models while
llm_model holds only one name.
Only authenticated calls are logged, USAGE_LOG_ENABLED can
switch logging off, and log writes are fail-soft — this is not a complete
record of traffic. Questions are grouped into sessions: a session is
exact when the caller echoed its id back to us and otherwise inferred from
a 30-minute gap in that user’s activity, so a grouping is only as
good as the source shown on the session itself. A call with no
attributable user gets no session at all; those questions are listed
separately rather than dropped.