`exd telemetry summary`

Per-flag eval counts and variant distribution, over a time window.

Synopsis

exd telemetry summary
    [--flag <key>]...
    [--namespace <slug>]
    [--environment <env>]
    [--since <duration> | --from <ts> --to <ts>]
    [--compare-to <duration>]
    [--source <uri>]...
    [--format human|json]
    [--thresholds <path>]
    [--engine <name>]
    [--limit <n>]
    [--explain]
    [--fail-on-error]

See common flags for the shared options.

Description

"How many evaluations happened, broken down by variant?" The default analytical entry point. Without any filter, summarizes every flag in every flag namespace across the window.

Use cases

Post-rollout sanity check. "Did the 10% rollout actually serve 10%?" Pair with --expected (via srm) for a statistical verdict; summary alone gives you the raw counts:
```
exd telemetry summary --flag onboarding-banner --since 24h
```
Week-over-week trend. Compare last 7 days against the prior 7:
```
exd telemetry summary --since 7d --compare-to 7d
```
Find under-evaluated flags. A flag with zero traffic triggers T008 (cold flag); a flag with anomalously low traffic triggers T007:
```
exd telemetry summary --namespace marketing --since 24h
```
Catch never-returned variants. A declared variant that never appears in records triggers T009:
```
exd telemetry summary --flag onboarding-banner --since 30d
```
CI gate on bucketing consistency. Two evaluations for the same unit_id_hash and manifest_version returning different variants trigger T010:
```
exd telemetry summary --flag onboarding-banner --since 1h --fail-on-error
```

Subcommand-specific flags

Flag	Notes
`--compare-to <duration>`	A second time window of equal length, ending where the primary window starts. Adds `delta_total` (signed integer) and `delta_pct` (signed float) to each per-flag row.

Result fields (`--format json`)

Per flag in result.flags:

Field	Type	Notes
`flag_key`	string	Flag key.
`namespace`	string	Namespace slug.
`total_evaluations`	integer	Sum across the window.
`variants`	map	Variant key → count.
`evaluation_reasons`	map	Reason key → count (e.g. `rule_matched`, `fallthrough`, `default_variant`).
`delta_total`	integer	Present iff `--compare-to`.
`delta_pct`	float	Present iff `--compare-to`.

Diagnostics

May emit:

T007 evaluation-rate anomaly
T008 cold flag (zero traffic)
T009 variant never returned
T010 bucketing inconsistency
T011 private-attribute leakage (every command)
T012 unknown evaluation reason (every command)

Example

$ exd telemetry summary --flag onboarding-banner --since 24h
flag                  total      variants
onboarding-banner    198,432    treat_a 91,008 (45.9%)  treat_b 90,521 (45.6%)  control 16,903 (8.5%)

diagnostics: (none)

Exit codes

See telemetry exit codes.

Synopsis​

Description​

Use cases​

Subcommand-specific flags​

Result fields (--format json)​

Diagnostics​

Example​

Exit codes​

See also​