Automation & Reports
Turn recurring agent operations into scheduled work.
10 operational task templates · 9 cadence presets · 5 delivery channels · 13 report sections. Schedule checks, preserve run history, and deliver decision-ready reports.
10
operational task templates
9
cadence presets
5
delivery channels
13
report sections
Execution path
From a recurring rule to a reviewable result.
Automation is the operating layer for repeated Rylvo work: scheduled checks, evidence gathering, optional analysis, saved run history, and multi-channel delivery.
Set the operating rule
Choose a built-in task or report, its bots, cadence, timezone, and delivery channels.
Rylvo finds due work
Scheduled sweeps calculate the next due window and atomically claim work before execution.
The worker gathers evidence
The selected worker checks Rylvo data; reports assemble the chosen sources, sections, and filters.
AI analysis is optional
Add a BYOK model for a narrative analysis, or keep the deterministic worker summary and report data.
Results become reviewable
Runs and reports are saved, then notifications move through the configured delivery path.
Scheduled tasks
Ten real jobs for the work agents leave behind.
Choose a purpose-built worker instead of wiring a collection of generic cron jobs. Every task has a clear evidence source and product boundary.
AUTOMATION / SCHEDULED TASKS
TZ: ASIA/KOLKATA4 new edge cases filed
deterministicDeterministic sweep of recent traces. Findings enter the Edge Cases triage queue with their traces attached.
Delivered to
SlackEmailWebhookQuality & safety
Continuously exercise behavior, policy, and test coverage.
Trace Scan
Scan recent canonical traces for edge-case candidates and regressions.
Recent trace evidence
Red Team Run
Trigger configured adversarial testing and summarize the resulting findings.
Requires Red Team access
Guardrail Audit
Review enabled guardrails and their recent operational state.
Guardrail configuration
Test Suite Run
Run a selected test suite when the engine integration is configured; otherwise report recent run state.
Engine + cron configuration
Task results depend on the selected scope, available workspace data, adjacent feature access, and deployment configuration. Optional model analysis does not replace the underlying worker result.
Scheduling
Human schedules, without making people write cron.
Use a preset or describe the recurrence in calendar terms, then anchor it to the timezone where the work is actually reviewed.
Nine presets
Interval
Every N hours or days
Weekly
Selected weekdays + time
Monthly
Day of month + time
Weekly quality review
Conversation Quality
Schedules use IANA timezones. Monthly dates clamp to the available days in shorter months. Task and report sweeps determine the actual start window.
Scheduled reports
Build the report around the decision, not the dashboard.
Select the evidence, sections, audience, and destination once. Rylvo can generate it on demand or on its configured schedule.
13 selectable operational data sources
Assemble reports from the parts of Rylvo that match the audience and question.
Each source is filtered to the selected bot scope and report window, with bounded collection per source.
Delivery
Create a channel once. Reuse it everywhere.
Attach enabled notification channels to tasks and reports, test their configuration, and turn destinations off without rebuilding every schedule.
Slack
Incoming-webhook delivery with Slack-oriented report blocks.
HTML report email to configured addresses and resolved workspace recipients.
Webhook
Structured JSON for an HTTPS endpoint, with optional custom headers.
Discord
Discord webhook delivery using a destination-friendly message shape.
Microsoft Teams
Teams incoming-webhook delivery using a MessageCard payload.
Protected destinations
New webhook destinations are stored through the encrypted credential path; email address configuration is handled separately.
Destination-aware payloads
Slack, Discord, Teams, email, and generic webhooks receive payloads shaped for their destination.
Optional request signing
Webhook dispatch adds timestamped HMAC-SHA256 headers when the signing secret is configured in the deployment.
Reliability boundaries
Know what runs, what retries, and what stays visible.
Rylvo uses fixed execution and delivery policies so operators can understand the real timeout, retry, and dead-letter behavior.
Scheduled task execution
- Atomic claim before work starts
- 90-second execution timeout
- Up to 3 consecutive retries
- Fixed 5, 10, then 20 minute backoff
- Saved run history and last summary
Webhook delivery
- Durable outbox before network delivery
- Dispatcher checks pending work every 2 minutes
- Up to 5 delivery attempts
- Backoff with jitter
- Terminal failures move to dead-letter state
Report generation
- Atomic monthly allowance check
- Single- or multi-bot scope
- Parallel bounded data assembly
- 30-second assembly timeout
- Saved report and distribution status
Access and plans
Separate seeing the operation from changing it.
Automation follows the workspace permission model, while plan allowances control how many tasks, channels, report configurations, and generated reports an organization can use.
Viewer
View the Automation workspace and its current execution state.
Operator
Generate an existing report configuration on demand.
Admin and owner
Create and manage scheduled tasks, channels, and report configurations.
Automation starts on Lite
Capacity increases by plan. Red Team, Mission Control data, connectors, and other adjacent capabilities remain subject to their own plan and configuration requirements.
FAQ
The production details worth knowing
Scheduling precision, retries, AI, delivery, and plan boundaries—answered without hand-waving.
Can I schedule arbitrary custom code?
No. The customer-facing scheduler exposes ten built-in operational task templates. The internal custom task type is not accepted by the canonical task API and does not execute arbitrary code.
How precise are scheduled runs?
Schedules use the selected timezone and wall-clock target, then run in a due-work sweep. Task sweeps run about every five minutes and report sweeps about every fifteen minutes, so this is operational scheduling rather than second-level cron precision.
What happens when a scheduled task fails?
A task times out after 90 seconds and can retry up to three consecutive times with fixed backoff. Its run history, last result, next run, and failure count remain visible. After the retry threshold, it waits for the next scheduled cadence.
Is an AI model required?
No. Workers produce a deterministic summary and structured findings. A configured BYOK model can add narrative analysis or report formatting, with fallback to the underlying result if that analysis fails.
How reliable is webhook delivery?
Non-email delivery is persisted to a durable outbox. The dispatcher retries failed sends up to five attempts before dead-lettering them. HMAC signature headers are added when the deployment has a webhook signing secret configured.
Which plans include Automation?
Automation starts on Lite. Task, channel, report-config, and monthly report allowances vary by plan; specific workers can also depend on adjacent features such as Red Team, canonical traces, or configured connectors.
Turn recurring operations into a system.
Schedule real operational checks, preserve their run history, build decision-ready reports, and deliver results where the team already works.