Scheduling & Workflows
Configure when the threat engine runs and which durable workflows it triggers — from continuous earnings-season scanning to a single end-of-day rebalance.
How scheduling works
Telumin runs on a durable workflow engine — not serverless functions that cold-start. This means your pipeline can run multi-hour missions, maintain state across executions, retry deterministically, and resume after a crash exactly where it left off.
You configure cadence per watchlist. Each watchlist can have its own cadence, priority, broker target, and notification preferences. Critical signals (stop hits, thesis breaks) bypass the cadence and trigger immediately.
Pipeline Cadences
Pick the cadence that matches the velocity of your thesis.
Continuous (event-driven)
ACTIVE THESISPipeline runs on every inbound filing, wire, or earnings event. Trailing stops are evaluated tick-by-tick against streaming market data. For positions where a 1-hour delay is unacceptable.
Best for: Active concentrated long book, earnings season, M&A names, post-IPO positions
Market-hours sweep
STANDARDPipeline runs at 09:35, 12:00, and 15:55 US/Eastern (or your configured exchange). Drift checks, rebalance evaluation, and trailing-stop adjustments batch into three predictable windows.
Best for: Core long book, RIA model portfolios, multi-watchlist thesis monitoring
End-of-day signal digest
DAILYPipeline runs once after the close. Compiles the day’s signals, drift, fills, and any Skeptic vetoes into one digest delivered to your inbox or Slack.
Best for: Quarterly-rebalance books, family-office discipline log, principal daily digest
Scheduled rebalance only
STRATEGICPipeline runs on a fixed schedule (weekly, monthly, quarterly). Drift alerts are computed on cadence; the Adjudicator rules; action proposals reach you in the dashboard at the next window.
Best for: Long-horizon books, model portfolios, low-touch thesis monitoring
Durable Workflows
Pipeline output triggers durable workflows. Workflows are deterministic, durable, and resume-on-failure. The four core workflows:
Drift Monitor
Computes drift vs stated thesis break-conditions on each cycle. When a condition is challenged, the three-role debate fires and the action queue surfaces a proposal.
Stop-Loss Watcher
Monitors price data tick-by-tick per position. When the threshold is crossed, the watcher alerts immediately — you decide what to do.
Thesis Guard
When a cleared signal contradicts a Day-1 condition, the workflow alerts the operator and (if configured) closes the position automatically.
Earnings Distillation
On earnings day, the workflow consumes the call transcript, extracts deltas vs. consensus, and emits a structured signal into the watchlist.
Notification Delivery
Cleared signals fan out to every channel you’ve enabled. Delivery is per-watchlist and per-severity.
Telumin Plus & Alert Tray
Real-time push to the dashboard. Severity colours; Stop / Thesis Drift never suppressed by quiet hours.
MCP tools
POST cleared signals to your endpoint. Used for Slack, internal portfolio systems, LP portals, and reporting.
Email digest
EOD or weekly digest of the watchlist’s signals, fills, and Skeptic vetoes. Distribution lists supported.
API pull
Your agent queries cleared signals through the MCP connector on its own cadence — in claude.ai, Claude Code, or your own stack.
Priority & Escalation
Stops & Thesis Drift fast-track
Trailing-stop hits and Skeptic-cleared Thesis Drift signals bypass the watchlist’s normal cadence. Regardless of whether you’re on a daily or weekly schedule, these are processed immediately, written to the trade record, and surfaced through every channel. The pipeline is designed so the operator can be away from the terminal and still trust the discipline.
Infrastructure
Workflows run on a durable workflow engine with persistent compute. No serverless cold-starts, no half-finished rebalances on crash.
Scheduling Recommendations
Start with EOD, escalate as positions need it
End-of-day cadence gives you full discipline coverage with low noise. Promote a watchlist to market-hours or continuous only when you have an active thesis that needs faster feedback (earnings, M&A, post-IPO).
On-demand runs are always available
Regardless of cadence, you can trigger a one-shot pipeline run for any watchlist from the terminal or via an MCP tool call. The on-demand result is logged independently and doesn’t affect the next scheduled run.
Schedule your first watchlist
Follow the quickstart and pick a cadence. Promote later if your thesis needs it.