feat(cellular): seed dev wildlife and tune budding playtests
Foundation / checks (client) (push) Successful in 5s
Foundation / checks (checks) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (postgres) (push) Successful in 4s
Foundation / checks (network) (push) Successful in 5s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 4m27s
Foundation / postgres-tools (push) Successful in 30s
Foundation / server-image (push) Successful in 4m16s
Foundation / image (push) Successful in 4m54s

This commit is contained in:
dingo
2026-10-26 12:01:57 -04:00
parent 18a64a29cd
commit ff121c3ed1
32 changed files with 738 additions and 78 deletions
+4
View File
@@ -35,6 +35,10 @@ requirements. See [the architecture](architecture/SYSTEM_OVERVIEW.md),
## Recommendations versus experiments
The maintainer's 2026-10-26 test request permits opt-in randomized wildlife roots
only in disposable DEV openings ([ADR-0180](adr/0180-dev-starting-wildlife.md)).
This fixture exception does not change the player-origin requirement for product worlds.
The maintainer clarified structural inheritance, birth-time control choice and
NPC-parent continuity on 2026-10-19 in
[ADR-0163](adr/0163-structural-inheritance-and-offspring-succession.md). These are
+11
View File
@@ -1,5 +1,16 @@
# Project status and next steps
## DEV startup wildlife and Bud controls - 2026-10-26
[ADR-0180](adr/0180-dev-starting-wildlife.md) implements opt-in, nursery-funded
random-genome roots at a fresh rules-13 opening, with deterministic durable
recovery and no replenishment on restart. The DEV trial prepares region 14 with
32 cells, 25% faster funded growth and the existing per-channel budding fee knobs.
Bud disables on resource shortfall and explains all five channels on hover; the
server projection includes the inherited plan and fee. See [tuning and promotion
limits](engineering/CELLULAR_PLAYTEST_TUNING.md) and [validation](engineering/VALIDATION.md).
Source/configuration preparation is distinct from image promotion and deployment.
## Photosynthesis producers (ADR-0136 step C) - 2026-10-19
[ADR-0175](adr/0175-photosynthesis-producers.md) adds genome schema 5's photosynthesis
+41
View File
@@ -0,0 +1,41 @@
# ADR-0180: Opt-in starting wildlife for disposable playtests
Accepted scoped experiment, 2026-10-26, requested by the maintainer in chat.
The vision requires player-origin lineages. Random founding wildlife conflicts with
that rule. Resolve this only for explicitly opted-in disposable DEV openings; this
does not change the product rule or authorize a World0 operation.
The optional revamp tuning key `playtest.starting_wildlife_cells` defaults to zero
and accepts 0–255, below the region's organism cap. Provisioning a nonzero count
also requires `--playtest-wildlife-first-id`, an operator-audited unused identity
range below `--life-allocator-start`. As with ADR-0123, the operator owns the
world-wide identity audit. Never use this range in another region or restored branch
of the same world. The DEV example reserves 900000003584–900000003615 for region 14.
Seeding is an explicit input to the bundle 14 -> 15 opening upgrade. It requires an
empty region at tick 1, spends nursery resources including construction heat, and
fails atomically on inadequate funding or placement. Roots are unowned test
organisms, with no fabricated player ancestry. Each begins fully grown, undamaged,
with zero spare matter and 30 seconds of full demand as energy, using the same
founder provisioning formula as players. Genome variation uses seeded, bounded
reference indices around the founder, with no acquired modules or free mutations.
The existing format-11 archive and organism records store the entire result.
Recovery first tries the historical upgrade, then, only for an empty tick-1 source,
reconstructs the fixture from the target's count and first identity and requires
byte equality. No per-tick rule, checkpoint encoding, wire version or golden
changes. Older readers can decode these checkpoints but cannot qualify this new
opening operation; retain the writing build. Existing non-fixture upgrades retain
their exact meaning. Restart and repeated provision never replenish wildlife.
Alternatives: silently seed product worlds (violates the vision); inject on every
worker start (duplicates stock and breaks history); add a general administrator
spawn protocol (unnecessary scope). Costs: an explicit exception to the no-biology
upgrade convention, extra replay qualification, and finite nursery consumption.
This is fixture coverage, not ecological balance or production qualification.
The Bud HUD consumes the server's current offspring-plan cost plus the configured
fee. Internal-cell litter displays a conservative maximum because its random draw
occurs at action time. No new simulation or wire version is needed for correcting
this derived presentation. Growth and fee changes remain new-region trial inputs.
+2
View File
@@ -1,5 +1,7 @@
# Architecture decision records
- [ADR-0180: DEV starting wildlife](0180-dev-starting-wildlife.md) - opt-in, nursery-funded random-genome roots for disposable playtest openings, exact recovery and no restart replenishment; a scoped exception to player-origin biology.
- [ADR-0179: Cellular gameplay revamp](0179-cellular-gameplay-revamp.md) - maintainer-accepted direction for rules 13: a 49-mutation catalog with behaviour-weighted offers, mutation splits and stay-or-become reproduction, multicellular organisms and emergence, matter as energy with a region spent pool, per-channel collection, membrane points and heritable body traits; two locked-requirement questions open; implementation pending.
- [ADR-0178: Durable liquid regions and session v13](0178-durable-liquid-region.md) - rules 12 on owner format 10 (currents, moving food, liquid motion), raw bundle 14 and schema 58 with a 13 -> 14 upgrade from the Git tuning file, the `cellular-v9` opening and session v13 (motion, currents, brake).
+6
View File
@@ -111,3 +111,9 @@ observations only; optical liquid shading is not a thermal field. Body temperatu
ambient temperature and directional heat guidance remain pending authoritative
phase-3 state and a negotiated perception contract. Existing public library and
wire interfaces are unchanged.
The [DEV playtest tuning](../engineering/CELLULAR_PLAYTEST_TUNING.md) Bud control
uses the current server-projected charge, including the inherited plan and fee.
Insufficient funding disables it, with a hover tooltip listing required and held
energy, carbon, nitrogen, phosphorus and other matter. Internal-cell litter shows
the maximum possible charge. The server remains authoritative at execution time.
@@ -0,0 +1,57 @@
# Cellular playtest opening controls
These trial settings live in
`deploy/kubernetes/region/dev/tuning/cellular-revamp.tuning`. They are read by
`evo-world-clone provision --opening cellular-v10 --revamp-tuning FILE`; the owner
then retains its recorded rules. Editing the file does not retune an existing
region. Open a fresh disposable region for a changed trial. See [ADR-0180](../adr/0180-dev-starting-wildlife.md).
| Key | Current DEV value | Meaning |
| --- | --- | --- |
| `playtest.starting_wildlife_cells` | 32 | Initial random-genome cells; 0 disables; max 255 and below the organism cap. Omission defaults to 0 for older files. |
| `body.growth_per_tick_centi_mille` | 625 | 6.25 per thousand per tick at growth speed 100, up from 5: 25% faster when funded. |
| `reproduction.bud_fee_e_pj` | 250000 | Additional energy fee per bud action. |
| `reproduction.bud_fee_c_fg` | 100000 | Additional carbon fee. |
| `reproduction.bud_fee_n_fg` | 30000 | Additional nitrogen fee. |
| `reproduction.bud_fee_p_fg` | 10000 | Additional phosphorus fee. |
| `reproduction.bud_fee_o_fg` | 60000 | Additional other-matter fee. |
| `reproduction.bud_fee_scale_percent` | 100 | Percentage of the listed fee charged; 0 disables it. |
The fee controls already existed; this change activates and documents them and
corrects the displayed cost to include them. Offspring tissue and starting reserve
remain funded biological costs. Faster growth spends the same resources in less
time; insufficient nutrients can still hold it back.
The DEV manifest prepares world 100, region 14, with a new stable genesis commit
and `--playtest-wildlife-first-id 900000003584` below
`--life-allocator-start 1000000000000`. This declares a reserved fixture range;
operators must audit its uniqueness across the world's other retained regions.
Regions 1–13 remain stored. Counts other than 32 need a correspondingly audited
range and enough nursery stock. The fixture spends the nursery rather than
minting resources; player founders draw its remainder.
Each cell uses founder-6 capabilities and full growth, no damage or spare matter,
30 seconds of its own full demand as reserve energy, and a seeded genome with
75–125-point body, collection, burn, engulfment and movement indices. It has no
acquired modules, player account or fabricated ancestry. Randomness is stable
for world, region and identity. Normal autonomous wildlife control takes over.
Recovery and repeated provisioning retain the saved population, including deaths;
they do not top it up.
Bud uses the current expressed parent and inherited offspring plan, plus the fee;
internal-cell litter conservatively shows its largest possible charge. Hover a
grey Bud button to inspect every required/available channel. A resource-ready
button still submits an intention: growth, placement, regional capacity and
changes before execution can affect the authoritative outcome.
The region-14 configuration requires an image built from this change before
manual Argo sync. The currently pinned image is not claimed to support the new
opening flag. Build promotion, deployment, live playtesting and identity auditing
against production stores were not performed by these local checks. No World0
operation is involved.
Focused durable verification in a new private local PostgreSQL cluster:
```sh
bash tools/ci/postgres-check.sh --starting-wildlife-only "$PWD/build/playtest-postgres/evo-postgres-tests"
```
+71
View File
@@ -10178,3 +10178,74 @@ Environmental limits and initial failures:
The AGENTS.md-required pre-push session-window refresh is desired Git configuration
only. No deployment, live reachability, World0 mutation or world reset is claimed.
## 2026-10-26 - DEV starting wildlife and resource-aware Bud
Scope: ADR-0180's explicit disposable DEV fixture, cellular opening/recovery,
server-projected budding requirements, native Bud control, and new-region tuning.
Existing unrelated drafts and untracked files were preserved. No golden, per-tick
simulation version, checkpoint encoding or wire layout changed.
Executed:
- GCC 14.2 Release with NDEBUG, configured from the release preset with client and
network enabled in build/playtest-release: complete CTest suite 122/122 passed.
Final focused UI-model/owner checks passed 2/2 after a test-only optional-access
cleanup and formatting-only correction in the affected revamp module.
- MSVC 19.44 client preset build succeeded; client CTest 10/10 passed. The final
UI-model cleanup was rebuilt and its test passed again. TMP/TEMP were directed
to ignored build/playtest-windows-tmp for local bindings-file checks.
- Clang 19 ASan+UBSan, existing asan-preset build/client-asan: focused tuning,
owner, UI-model and native panel checks 4/4 passed; final UI-model cleanup 1/1.
- PostgreSQL 18.6, GCC 14 Release with the existing prepared dependency:
tools/ci/postgres-check.sh --starting-wildlife-only against
build/playtest-postgres/evo-postgres-tests passed in a fresh private local
cluster. It published an empty genesis and all prerequisite upgrades, seeded
32 funded roots, replaced the lease, recovered byte-identical state, published
a real wildlife tick and replayed it identically without adding more cells.
- Parser/owner coverage includes optional count default zero, count bounds,
deterministic genome/placement, ordinary founder endowment rules, exact
conservation, refusal on inadequate nursery stock, invalid identities and
repeated seeding, and restored continuation. Cost checks include the fee;
UI checks cover all five shortfall channels and exact affordability equality.
- Native panels exercise 640x480, 960x640 and 1280x800, including the production
disabled Bud hover tooltip, refusal to click while short, and funded clicking.
A cost-aware fixture exposed Vitals gaining horizontal overflow after restore;
fixed content width and a reserved vertical scrollbar resolved it.
- Repository format-check passed with clang-format 19. A pre-existing formatting
violation in the affected cellular_revamp.cpp was corrected. git diff --check
passed. Rendered deployment policy passed (75 source objects, 140 rendered
objects, 17 overlays); its 50-case fail-closed self-test passed.
Limits and initial failures:
- The Windows debug build stopped with MSVC C1902 (program-database manager
mismatch) before compiling the changed code; the Windows Release client build
and Linux Release/sanitizer builds succeeded. No debug pass is claimed.
- Initial compiler and static-analysis failures (a missing draft initializer and
optional-access checks), plus the native layout/hover fixture failures, were
corrected before the successful checks above. An initial sanitizer compile
overlapped a header formatting write; the stable-source rerun passed.
- These are local fixtures, not live gameplay, ecological balance, hardware
performance, or production-store identity-audit evidence. DEV region 14 needs
a compatible image promotion before manual Argo sync. No deployment, live
reachability, World0 change, reset or pruning was performed. The AGENTS.md
pre-push session-window refresh is desired Git configuration only.
Final static analysis: the analysis preset in build/playtest-analysis built the
changed lifecycle, tuning, persistence, bootstrap and client targets, plus their
regressions, with clang-tidy 19 and warnings as errors. Its focused CTests passed
5/5 and the isolated PostgreSQL opening/recovery check passed again. A pre-existing
unchecked optional access in the touched UI-model test was rewritten through its
existing checked helper. Changed C++ files also passed clang-format 19 dry-run.
The actual QUIC-enabled evo-server also built successfully in build/tidy-postgres.
A separate clang-tidy-19 run on its main translation unit, with an explicit
cellular_bridge.hpp header filter and warnings as errors, passed. It required a
pre-existing optional-feedback access to be made local before checking/reading.
The bridge header passed the Linux clang-format 19 dry-run (minor formatter
versions differ on an existing multiline lambda's wrapping).
Publication limit: the scoped commit was created locally, but the configured HTTPS
remote rejected git push with an authentication failure. Git Credential Manager
is configured; no stored credential was deleted or changed. A user-authorized retry uses credentials supplied outside the repository. Unrelated drafts and untracked files
remain outside the commit.