Configuration

The complete environment reference for the Metrics Exporter — MongoDB connection, caching, timezone, per-group metric toggles, logging, and the optional health check.

The exporter is configured entirely through environment variables, set in config/.env under the same names the container reads (documented below); config/exporter.compose.yml passes each one straight through.

MongoDB connection

Name	Purpose	Default	Required
`MONGODB_URI`	Connection string (single node or replica set)	`mongodb://mongodb:27017/`	No
`MONGODB_DATABASE`	Database to scrape	`IntelliAsk`	No
`MONGO_READ_PREFERENCE`	Replica read strategy	`secondaryPreferred`	No

MONGO_READ_PREFERENCE accepts primary, primaryPreferred, secondary, secondaryPreferred, or nearest.

Database name is case-sensitive

MongoDB database names are case-sensitive. Production clusters use IntelliAsk; if a deployment uses a differently-named database, set MONGODB_DATABASE explicitly to match.

Caching & collection

The exporter caches computed metrics in a background thread so scrapes are cheap.

Name	Purpose	Default	Required
`METRICS_CACHE_ENABLED`	Enable the background refresh + cache	`true`	No
`METRICS_CACHE_TTL`	Cache refresh interval (seconds)	`60`	No

Keep your Prometheus scrape_interval ≥ METRICS_CACHE_TTL so every scrape gets fresh data without triggering redundant recomputation. On very large databases, raising METRICS_CACHE_TTL (e.g. 120–300) reduces MongoDB load.

Timezone

Name	Purpose	Default	Required
`METRICS_TIMEZONE`	IANA timezone for daily/weekly/monthly calendar boundaries	`UTC`	No

Only affects the calendar boundaries of the daily/weekly/monthly unique-user metrics (e.g. Asia/Tokyo, Europe/London). Timestamps are stored in UTC; an invalid value logs a warning and falls back to UTC.

Metric-group toggles

Each metric group can be enabled or disabled independently. Disabling the expensive groups (rating, tool) dramatically reduces query cost on large databases.

Name	Group	Default	Notes
`ENABLE_BASIC_METRICS`	Message / error / conversation counts	`true`
`ENABLE_TOKEN_METRICS`	Input/output token usage	`true`
`ENABLE_USER_METRICS`	User counts & activity (daily/weekly/monthly)	`true`
`ENABLE_MODEL_METRICS`	Per-model breakdowns	`true`
`ENABLE_TIME_WINDOW_METRICS`	5-minute activity windows	`true`
`ENABLE_RATING_METRICS`	Chat ratings & feedback	`true`	⚠️ expensive
`ENABLE_TOOL_METRICS`	Tool / plugin usage	`true`	⚠️ expensive
`ENABLE_FILE_METRICS`	File upload count	`true`

See the full list of metrics each group produces in the Metrics catalog.

Health check (optional)

Name	Purpose	Default	Required
`INTELLIASK_URL`	IntelliAsk web UI URL to probe for the `intelliask_status_code` metric	(unset)	No

When set, the exporter sends a HEAD request to that URL and exposes the HTTP status as intelliask_status_code (-1 if unreachable within ~5 s).

Logging

Name	Purpose	Default	Required
`LOGGING_LEVEL`	Log verbosity: `debug`/`info`/`warning`/`error`/`critical`	`info`	No
`LOGGING_FORMAT`	Python logging format string	`%(asctime)s - %(levelname)s - %(message)s`	No

Performance tuning for large databases

When	Do this
10M+ messages	`ENABLE_RATING_METRICS=false`
Large tool arrays	`ENABLE_TOOL_METRICS=false`
Frequent scrapes	`ENABLE_TIME_WINDOW_METRICS=false`
High MongoDB load	Raise `METRICS_CACHE_TTL` (e.g. `120`–`300`)
Replica set available	`MONGO_READ_PREFERENCE=secondaryPreferred`

Applying changes

After editing any EXPORTER_* value in config/.env, recreate the service:

intelliask up exporter