[slopfix] fix(self-host): repair dev compose generation, kitchen-sink serverful runner, and postgres/udb bootstrap

[MasterPtato]

No description provided.

[MasterPtato]

Stack for rivet-dev/rivet

• #5552 [slopfix] fix(universaldb): match fdb conflict semantics in …
• #5551 [slopfix] fix(universaldb,util,pegboard-envoy): index dedup …
• #5398 feat(kitchen-sink): add churnDb actor and seed driver for co …
• #5397 chore: Clean up
• #5382 [slopfix] docs(self-hosting): correct postgres config format …
• #5356 [slopfix] fix(pegboard): decrement kv_storage_used metric on …
• #5354 [slopfix] refactor(universaldb,ups): single/multi-node udb o …
• #5353 [slopfix] perf(universaldb): fold postgres drain_batch into …
• #5342 [SLOP(claude-opus-4-8-high)] perf(universaldb): batch leader …
• #5341 [slopfix] docs(self-hosting): promote postgres from experime …
• #5340 [slopfix] fix(self-host): repair dev compose generation, kit … 👈
• #5337 [SLOP(claude-opus-4-8-high)] feat(universaldb): graceful pos …
• #5336 [slopfix] test(universaldb): postgres leader failover + abor …
• #5335 [SLOP(claude-opus-4-8-high)] feat(ups): table-backed postgre …
• #5329 [SLOP(claude-opus-4-8-high)] feat(universaldb): postgres lea …
• #5327 [SLOP(claude-opus-4-8)] feat(pegboard-envoy): rate-limit env …
• #5326 [SLOP(claude-opus-4-8)] feat(util): add rate limiter primiti …
• #5322 [SLOP(claude-opus-4-8)] feat(rivetkit-core): add serde durat …
• #5321 Add observation for serde_bare
• #5320 Add observation and ser/de metrics for serde calls
• #5314 [SLOP(claude-opus-4-8-medium)] chore: add profiling cargo pr …
• #5306 [SLOP(claude-opus-4-8)] fix(kitchen-sink): serve built front …
• main

Get stack: forklift get 5340
Push local edits: forklift submit
Merge when ready: forklift merge 5340

_{change wpvkrymu}

[claude]

Review

Reviewed the diff (self-host compose regeneration/relocation, kitchen-sink Dockerfile, Prometheus registry namespace, runner-config bootstrap). Overall this is a solid "repair" PR — I traced the config removals against the Rust defaults and the metric-name change against the Grafana dashboard JSON, and both check out as correct fixes rather than regressions. Two doc/config drift issues are left behind by the directory move, though.

Issues found

1. Root CLAUDE.md "Docker dev environment" section now points at a path that no longer exists.
This PR moves the generated compose output from self-host/compose/dev* to self-host/dev* (confirmed via the PR's own renames and self-host/compose/template/src/main.ts's outputDir = path.join(__dirname, "../../../", templateName), which resolves to self-host/<name>, not self-host/compose/<name>). CLAUDE.md (repo root, lines 69-76) still says:

cd self-host/compose/dev
docker-compose up -d

and

- Do not edit `self-host/compose/dev*` configs directly. Edit the template in `self-host/compose/template/` and rerun...

Both are stale now — cd self-host/compose/dev will fail, and the caution note should say self-host/dev*. Since this file is load-bearing instructions for future agents/contributors, it's worth fixing in this PR (or a fast follow) rather than leaving it to drift.

2. Root .gitattributes linguist-generated patterns still target the old paths.
Lines 30-34 of .gitattributes mark self-host/compose/dev/**, self-host/compose/dev-host/**, etc. as linguist-generated=true. Since those directories moved to self-host/dev/**, self-host/dev-host/**, etc., these patterns no longer match anything, so GitHub will stop collapsing the generated compose/Grafana/vector configs in diffs and will start counting them in language stats. Should be updated to the new self-host/dev*/** paths in the same change.

Things that look correct (verified, not just assumed)

• engine/packages/metrics/src/registry.rs: adding the "rivet" namespace to the Prometheus registry looks like a genuine fix, not a breaking rename — the checked-in Grafana dashboards (e.g. dashboards/api.json) already query rivet_api_request_duration_bucket, etc., while the Rust metric is registered as bare "api_request_duration" (engine/packages/api-builder/src/metrics.rs). Before this change the exposed metric name and the dashboard queries didn't match; this fixes that. I didn't find any metric already registered with an explicit rivet_ prefix, so there's no double-prefixing risk.
• Removed guard.port, api_peer.port, cache.driver, clickhouse.secure from the generated config.jsonc files match the Rust defaults exactly (GUARD: u16 = 6420, API_PEER: u16 = 6421, CacheDriver::InMemory, secure: bool with #[serde(default)] → false), so this is a safe simplification, not a functional change.
• The new runner-config-init one-shot service and its curl retry loop look correct: valid shell quoting, correctly gated behind service_healthy/service_completed_successfully, and the RIVET_KITCHEN_SINK_MODE/RIVET_POOL env vars it feeds the runner are consumed by the existing kitchen-sink example code and rivetkit's env-vars.ts.
• Postgres max_connections=500 and mounting the volume at the /var/lib/postgresql parent (rather than /var/lib/postgresql/data) are reasonable dev-only changes; no PGDATA override is introduced so Postgres will still use .../data inside the volume.

Minor / non-blocking

• self-host/compose/template/src/services/edge/runner.ts is now an empty function kept only for its comment — pre-existing pattern, not introduced here, but worth eventually just inlining the comment at the call site if generateRunner never grows real logic.
• The runner-config-init script hardcodes 6420 and the Bearer dev admin token; consistent with the rest of this dev-only template (which already hardcodes admin_token: "dev" and 6420 elsewhere), so not a new issue, just flagging since it's a fresh piece of shell.

No test coverage changes are included, which is expected/fine for compose-template and docs plumbing like this — nothing here is unit-testable Rust/TS logic beyond what's already covered.