Daily Report

Once a day, Baldur rolls everything it did to keep your services healthy into one short digest — so you can tell at a glance whether yesterday was calm or a near-miss, without opening a dashboard.

What is it?

Most monitoring shouts at you in the moment: an alert fires, a dashboard turns red, and you go look. A daily report is the calm opposite: a scheduled summary that arrives once a day and tells you what happened over the last 24 hours and what was handled automatically. Think of it as the end-of-shift handover note an on-call engineer leaves for the next person: here's what broke, here's what recovered, here's what still needs eyes.

In Baldur's terms, the Daily Report gathers a full day of self-healing activity — circuit-breaker trips, the items it auto-processed (archived, expired, recovered, purged), and errors — and rolls it into a single digest, kept as a running history you can look back through.

Why it matters

Self-healing is mostly invisible when it works. Baldur quietly retries a flaky call, trips a circuit before a slow dependency drags everything down, returns a fallback so a user never sees an error, and if none of that ever surfaces, you have no idea whether the system is coasting or quietly catching fires all day.

The Daily Report makes the invisible visible. A digest answers the question every operator actually has (is everything OK?) without anyone opening a dashboard. On a quiet day it collapses to a single "all quiet" line (and a day with nothing recorded at all sends no digest), so a report that suddenly runs long is itself the signal that yesterday was busy. For a small team with nobody watching screens, it's the cheapest possible proof that your safety net is doing its job.

How it works in Baldur

Once a day, Baldur builds a report for the previous day from the activity it recorded, stores it, and keeps a rolling history (about three months by default) that you can query by date.

The report is adaptive, to stay readable:

Every digest opens with a core summary line, so a delivered report always tells you it ran.
Detail sections (auto-processing of archived/expired/recovered items, circuit-breaker activity, errors, and the like) appear only when they have something to report. A day with activity on record but nothing that needs attention collapses to a single line: "All quiet — 0 processed, 0 alerts."
A day with nothing recorded at all (no activity and no pending backlog) is skipped outright: no digest is sent and nothing is stored for that date.
The report's severity is derived from its contents: a clean day is info, task failures make it warning, a critical alert makes it critical.

How you get the report depends on your tier — you either pull it on demand or have it delivered:

What you observe	When it happens
`baldur report` (or the report API / admin console) returns the day's digest and a multi-day trend	Any time — the report is generated and stored in every tier
An "All quiet — 0 processed, 0 alerts" one-liner	Something was recorded that day, but nothing needed attention
No digest (and no stored report) for a date	Nothing was recorded that day — generation skips it entirely
Expanded detail sections	Those events actually occurred that day
The digest pushed to your Slack automatically each day	PRO — see Tier behavior below
A "what you're missing" insights block inside the report	OSS only — see Tier behavior below

The report ships disabled by default; you turn it on once you want the daily digest.

Configuration

The Daily Report's knobs — whether it's on, when it runs, how long history is kept, and how often the insights block appears — are treated as advanced settings in v1.0 rather than part of the stable operator environment-variable allowlist, so they may change before they're promoted. See the API Reference for the current settings; the report is off by default.

Reading a stored report needs no extra configuration — the CLI talks to the same data the report API serves:

baldur report                 # list recent reports
baldur report --date today    # show one day's report (or any YYYY-MM-DD)

Tier behavior

The Daily Report runs in every tier, but how you get it and what it contains scope to the features you have active.

In OSS: Baldur generates the report from your real activity, keeps the rolling history, and you read it on demand — baldur report on the command line, or the report API / admin console. OSS already captures failed work and replays it when a dependency recovers, so the digest shows that activity — the dead-letter queue section and the auto-replay line are both OSS. What OSS does not do is tune that recovery for you, so the report also carries a "what you're missing" insights block: drawn entirely from your own production numbers, it estimates the impact the PRO features would have had — for example N circuit-breaker trips with no automatic degradation, N operations captured in the dead-letter queue and replayed at a fixed batch size rather than one adapted to the recovering dependency, or N drift warnings you had to resolve by hand. It's a directional estimate from your data, not a synthetic demo, and it appears on a cadence you control.
With PRO active: the same report is delivered to Slack automatically each day — Baldur's notification transports ship with PRO, so in OSS the report is generated and stored but pushing it to a channel is a PRO capability. The "what you're missing" block disappears, because you are no longer missing it. The "Automated Actions" section — which on OSS carries the auto-replayed dead-letter batches — fills out with the rest of what Baldur did while you were away: canary rollouts and rollbacks, emergency-level changes, governance blocks. The same daily digest shifts from "here's what you're missing" to "here's what was handled for you."

Sections by tier

Every section the digest can contain belongs to a tier, and a section appears only when its producing feature is part of the tier you're running — so the digest never shows a section for a capability you don't have. The sections the Slack digest can carry today:

Section	What it covers	Tier
`auto_processing`	Items auto-processed yesterday — archived, expired, recovered, purged	OSS
`alerts`	Config-drift warnings	OSS
`circuit_breaker`	Circuit-breaker transitions (opened / closed)	OSS
`errors`	Task failures and critical alerts	OSS
`custom`	Any custom counters you record	OSS
`shadow_pro`	The OSS-only "what you're missing" insights block	OSS
`dlq`	Dead-letter queue activity — new, auto-resolved, pending	OSS
`automated_actions`	Heading for the actions Baldur took for you	OSS
`auto_replay`	Dead-letter batches auto-replayed	OSS
`canary`	Canary rollouts completed and rolled back	PRO
`emergency`	Emergency-level changes	PRO
`governance`	Governance policy blocks	PRO

Features that arrive after v1.0 add their own sections to the digest as they ship.