Configuration¶

The collector reads its settings from environment variables. Locally, copy .env.example to .env. docker compose up also works without a .env, using the defaults below.

Keys and access¶

Variable	Default	Notes
`REQLY_INGEST_KEY`	`demo-key`	Write key. SDKs and OTLP exporters send it as `X-Reqly-Key`; it also manages SLOs, specs and projects
`REQLY_READ_KEY`	`demo-read-key`	Read key the dashboard sends. It's compiled into the dashboard, so treat it as public to anyone who can open the dashboard. Must differ from the ingest key
`PUBLIC_DASHBOARD`	`true`	`true`: anyone with the dashboard can read (a public demo). `false`: sign-in required and the read key is refused
`REQLY_ADMIN_USERNAME` / `REQLY_ADMIN_PASSWORD`	`admin` / empty	First admin, created at start-up when there are no users. Password 12+ characters
`SESSION_TTL_HOURS`	`168`	How long a sign-in lasts
`CORS_ORIGINS`	`*` locally	Comma-separated dashboard origins allowed to call the collector. Restrict it in production
`FORWARDED_ALLOW_IPS`	`127.0.0.1`	Behind a proxy, the proxy's IP (or `*` if the collector is only reachable through it), so rate limits see real client IPs

The collector warns at start-up if the keys are the defaults or if the two keys are equal.

AI¶

Variable	Default	Notes
`GROQ_API_KEY`	empty	console.groq.com/keys, free tier. Empty: plain-text weekly report, Ask Reqly off
`GROQ_MODEL`	`openai/gpt-oss-120b`	Model for the weekly report
`ASK_MODEL`	`GROQ_MODEL`	Model for Ask Reqly; needs tool calling
`ASK_DAILY_LIMIT`	`200`	Ask Reqly questions per collector per day. `0` turns Ask off
`LLM_PRICES_FILE`	empty	YAML price table that overrides and extends the built-in one (LLM cost)
`CONSUMER_LIMIT_PER_DAY`	`1000`	Distinct API consumers kept per service per day; later new ones are `__other__`. `0` turns the cap off

Alerts and jobs¶

Variable	Default	Notes
`ALERTS_ENABLED`	`true`	Hourly anomaly check at :15
`ALERT_SLACK_WEBHOOK_URL`	empty
`ALERT_DISCORD_WEBHOOK_URL`	empty
`ALERT_WEBHOOK_URL`	empty	Generic JSON webhook (payload)
`ALERT_RENOTIFY_HOURS`	`6`	Reminder interval while an alert keeps firing
`LLM_COST_ALERT_MIN_USD`	`1.0`	The smallest extra LLM spend per route-hour, in USD, worth an alert. `0` turns LLM cost alerts off. The local Compose stack uses `0.4`
`DASHBOARD_URL`	`http://localhost:5173`	Linked from Slack alerts
`INSIGHTS_SCHEDULER_ENABLED`	`true`	`false` turns off the weekly insights report
`LATE_DATA_REFRESH_SECONDS`	`60`	How often late-arriving events are materialized into the aggregates

Database¶

Variable	Default	Notes
`DATABASE_URL`	set by Compose	`postgresql://reqly:<password>@<host>:5432/reqly`. Add `?sslmode=require` for a remote database
`POSTGRES_PASSWORD`	`localdev`	Local Compose only

The database must be TimescaleDB with the Toolkit: the timescale/timescaledb-ha:pg16 image includes both. The collector applies its migrations on start-up.

Dashboard¶

Variable	Default	Notes
`VITE_COLLECTOR_URL`	`http://localhost:8000`	Collector URL as seen from the browser, at build time
`VITE_READ_KEY`	`demo-read-key`	At build time
`REQLY_COLLECTOR_URL` / `REQLY_READ_KEY`	—	The same two settings at run time, for the prebuilt image

Local demo traffic¶

Variable	Default	Notes
`BACKFILL_WEEKS`	`8`	Weeks of synthetic history the load generator creates on first run
`BACKFILL_EVENTS_PER_HOUR`	`30`	Synthetic events per hour during the backfill

SDK settings (REQLY_SERVICE_NAME, REQLY_API_KEY, …) are listed on the Python and Node.js pages.