Overview
One cron-scheduled task keeps deep-research reports fresh for every person a user is tracking. Users add people via Create a deep research person; the cron re-enriches them on a schedule so reports don’t go stale.deep-research-daily
| Schedule | Daily at 06:00 UTC |
| Cron | 0 6 * * * |
| File | trigger/deep-research-daily.ts |
deep-research-enrich for every tracked person whose current status is not running. Skipping running prevents concurrent runs for the same person. Each child enrichment hits the web, crawls current data, re-runs the LLM synthesis, and updates the report in place.
What an agent should know. Polling Get the latest deep research report for a tracked person might return a
status: "running" at any time — it could be a freshly-queued user-initiated run, or this cron re-enriching. Either way, the polling pattern is the same: wait for status === "completed" and read the report field.Rate considerations. A single user tracking 500 people produces 500 child enrichment runs at 06:00 UTC. Each burns deep-research credits and can take ~60 s–5 min. Scale accordingly — the task doesn’t currently throttle itself.Credit impact. Daily refresh is a recurring credit cost. Users who don’t need daily freshness can stop tracking a person via Delete a deep research person to drop them from the cron’s fan-out.