The dedicated mc-forge droplet (159.203.170.249:3000/mcadmin) is gone; the forge
now rides a shared services box, addressed by the stable hostname
forge.mc.uvlava.com/applications. The cloud-DX toolchain still pointed at the dead
endpoint, so every worker clone + golden-image build was broken.
- scripts/lib/forge-remote.sh: single source of truth — builds the authenticated
clone URL from the hostname + ~/.vault/services-forge-token (relocation-proof;
no hardcoded IP). Exports MC_FORGE_GIT_REMOTE.
- cloud-bringup.sh / dist.sh: source the helper instead of the dead
mc_forge_creds + 159.203 URL. Also fix cloud-bringup REPO path to the current
@mc/@applications/magicciv location.
- settings.local.json autoMode trust block: name the new forge host + 'mc' DO
project (was 159.203 + 'mc:dev'), else cloud provisioning is denied as exfil.
- cloud-dx-do.md: document the new forge + token.
Verified: helper authenticates to the live forge (ls-remote main); scripts parse;
JSON valid.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Build the linux .so/wasm once on a worker and let sim/test/AI runners fetch the
prebuilt artifact (keyed by git sha) instead of recompiling — N workers share
one build. Adds the magicciv-artifacts DO Space, rclone in the golden image, and:
- dist:publish build + upload builds/<sha>/{.so,wasm}
- dist:fetch download the prebuilt .so for HEAD's sha
- dist:sync git pull -> fetch prebuilt if published, else build
- dist:models share RL .onnx via the Space (push/pull/ls)
Complements sccache (compile cache) by caching final outputs. Creds via
RCLONE_S3_* env over ssh, never on worker disk/argv; degrades to build-on-worker
when creds/cache absent.
Also hardens the dispatch layer (pre-existing, affected test/build/render too):
- pass -i ~/.ssh/id_mc_fleet on dispatch ssh (don't rely on agent-loaded key)
- guard _dist_first_host against an empty / "fleet down" inventory
- drop ssh -n on heredoc-stdin verbs (it redirected stdin from /dev/null)
Proven end-to-end on DO: publish built a 43.9MB .so + wasm; dist:sync fetched it
in 2.8s (no rebuild).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Incremental rebuilds accumulate snapshots (~$0.40/mo each). dist:prune keeps
the newest N (default 2: current + one rollback); dist:image reminds you to run it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Packer base image is now a var; ./run dist:image builds FROM the newest
mc-golden snapshot by default, so the idempotent provision.sh only redoes changed
work (~3-8 min vs ~20 cold). --cold rebuilds from stock Ubuntu to reset layer
cruft. Made the clone step idempotent (clone-or-fetch) so it works on a
pre-provisioned base. Directly addresses 'avoid unnecessary rebuilds'.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Golden image now installs the software-render stack (weston, libgl1-mesa-dri
llvmpipe, mesa-vulkan-drivers, vulkan-tools) so any worker renders proof scenes
via gl_compatibility/opengl3 with no GPU. New ./run dist:render <scene> <out.png>
wraps tools/capture-proof.sh against a worker (replaces the apricot SCREENSHOT_HOST).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Offload heavy compute from plum (M2 Air) to on-demand DO workers:
- dist:test — cargo test --workspace (nextest) on a worker (the main DX win)
- dist:build — cargo build + WASM on a worker; rsync the platform-independent
WASM back (native .so is linux-only, stays on the worker)
- dist:sync — git pull <ref> + rebuild gdext on live workers (no image rebuild)
- forge:down/up — snapshot+destroy / restore-from-snapshot (DO bills powered-off
droplets; only destroy stops it). ~$6/mo -> ~$0.30/mo idle; refreshes the
forge IP in ~/.vault/mc_forge_creds on restore.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ephemeral CPU Droplet fleet that horizontally scales the iteration loop:
- infra/terraform/test-fleet: cattle Droplets from a golden image (auto-discovered
by name via digitalocean_images), grouped under the mc:dev DO project, with a
mocked-provider test suite (no token/spend).
- infra/packer: golden-image builder reusing scripts/dev-setup/linux.sh.
- scripts/run/dist.sh: ./run dist:{check,up,sim,train,down} — shard sim/test
batches across workers via autoplay-batch AUTOPLAY_HOST+SEED_OFFSET.
GPU intentionally absent (workload is CPU-bound per docs/ai-production.md).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Verified file:line: the live GDScript events modules have NO era-based max_tier
cap (0 hits) — headless flat max_tier=10 is correct parity; an era cap would
invent a rule the game lacks (gold-plating, dropped). And natural events already
fire + apply terrain effects headless; only the fired list surfacing to
TurnResult is missing (processor.rs:1117 `let _fired =`), an observability nicety
not a system gap. Confirms the headless natural-events system is functionally
complete; narrows Gap-2's real remainder.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The live GDScript turn emitted `unit_healed` inline; the headless healing
phase recovered HP silently. The healing phase runs in the end-of-turn
`fn(&mut GameState)` registry (no event sink), so follow the FloraSuccession
buffer pattern: stash `(player, unit_id, applied_amount, col, row)` into a new
transient `GameState.pending_heal_events`, drain it in `step()` into
`TurnEvent::UnitHealed`. The buffered amount is the CLAMPED delta actually
applied (not the nominal heal rate). No wire surface — dispatch drops it; the
live UI consumes it via the kind-tagged `event_to_dict` dict.
Verified headless: mc-replay 19/0 (unit_healed_serde), mc-turn 289/0
(healing_buffers_unit_heal_event_with_applied_amount +
healing_buffers_clamped_amount_near_full_hp + event_collector_wiring).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
