feat: count live workers and worker slots per tenant

[grutt]

Description

Extend per-tenant allocated-resource counts with currently connected workers and their configured slot capacity, so cloud billing can SetUsage those gauges the same way as crons, scheduled runs, and webhooks.

Type of change

• New feature (non-breaking change which adds functionality)

What's Changed

• Count live non-operator workers (5s heartbeat, assigned dispatcher, active, not paused) per tenant
• Sum configured slot units for those same workers from v1_worker_slot_config
• Notify allocated-usage refresh on worker create, delete, pause, and active-status changes (not heartbeats)
• Cover the new counts in the allocated-resources repository test

Checklist

Changes have been:

• Tested (unit, integration, or manually with steps specified)
• Linted and formatted
• Documented (where applicable)
• Added to CHANGELOG (where applicable) -- see Keep a Changelog

Testing

CountAllocatedResourcesByTenant now asserts a live worker is counted and a stale/paused worker is not. Plan checked on a production shard: tenant-filtered worker+slot path uses Worker_tenantId_lastHeartbeatAt_idx.

---

🤖 AI Disclosure

• I acknowledge that an LLM was used in the creation of this Pull Request, in accordance with Hatchet's AI_POLICY.md.

• Details: Implemented the SQL, repository callbacks, generated sqlc, and tests.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hatchet-docs	Ready	Preview	Aug 27, 2026 3:00pm

[greptile-apps]

Greptile Summary

Extends tenant allocated-resource aggregation with live non-operator worker counts and configured slot capacity. It also emits allocated-resource change notifications after relevant worker mutations and expands repository coverage.

• Adds active-worker, worker-count, slot-count, and tenant-union CTEs.
• Extends generated sqlc rows with worker and slot totals.
• Adds worker lifecycle callback hooks after successful mutations.
• Tests that live workers are counted while stale and paused workers are excluded.

Confidence Score: 4/5

The PR appears safe to merge, with only a non-blocking comment clarification needed.

The worker predicates align with existing live-worker metering semantics, slot capacities are aggregated through the tenant-scoped worker relation, and the accepted concern is limited to how the query-plan constraint is documented.

Files Needing Attention: pkg/repository/sqlcv1/allocated_resources.sql, pkg/repository/sqlcv1/allocated_resources.sql.go

_{Reviews (1): Last reviewed commit: "Merge origin/main into feat/meter-alloca..." | Re-trigger Greptile}

[greptile-apps]

State Current Query Constraint

This comment contrasts the query with an alternative implementation rather than documenting the current tenant-first query-plan constraint, making the invariant less clear for future maintainers.

Suggested change

	-- Active workers are collected once and reused for both counts so the
	-- tenant-filtered path can use Worker_tenantId_lastHeartbeatAt_idx
	-- instead of scanning historical v1_worker_slot_config rows.
	-- Keep the tenant filter in active_workers so tenant-scoped queries use
	-- Worker_tenantId_lastHeartbeatAt_idx before joining worker slot rows.

Context Used: AGENTS.md (source)

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!