Read the service's own watch
Read what Trillion's own watch saw last: whether every store answered its calls in the last ten minutes, and whether what landed on main went live within the hour
Read what Trillion's own watch saw last: whether every store answered its calls in the last ten minutes, and whether what landed on main went live within the hour
With {}, read the latest saved service-watch snapshot: { said, read_at, watches: { <watch>: { failing, said, evidence, since, last_worked } } }. With build_commit (exactly 40 hexadecimal characters), read just that retained build record from the same service owner, without enumerating keys or running a build. Returns said, read_at, watches: {}, build_commit, build_status (found, absent or unreadable), build, and build_read coverage with explicit byte limits and omissions. KV is eventually consistent; absence does not prove a build never ran. The record's outcome and diagnostics do not by themselves prove the live deployment. Screenshot bytes are not returned. Existing scheduled checks continue to use {}.
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to read what Trillion's own watch saw last: whether every store answered its calls in the last ten minutes, and whether what landed on main went live within the hour”
What your AI calls: run_tool {"tool_id":"tool:read-service-watch"}