Repository navigation
[SI] SingularityIteration-0.1.10.26-beta
Pre-release0.1.10.26
基于 0.1.10.25。交易所设计方案的第二阶段 X3:系统悬赏三个等级、国库与宝库循环、每周上限的发行、si_exchange 数据包、Java API 与事件。新规则先在离线测试 Harness1126 中与 0.1.10.25 的市场委托比较。详细说明见 docs/releases/0.1.10.26/UPDATE.md。
- 系统悬赏(
BountyBoard):普通(每天保持 3~5 件,市场缺少的商品,价值 4,000 × 0.5~1.5,3 天,奖励 1.15 倍)、高级(每周 2~3 件,数据包中的成套物资,7 天,1.3 倍,多人游戏中最后一天之前一人最多 50%)、特别(每天 5%,稀有物品,7 天)。取代每天 3 件、全部以新货币支付的市场委托(存档中的旧委托保留到到期,配置systemCommissionsPerDay删除)。离线一年:每周发行的货币 708,416 → 146,761,单件最大货币奖励 1,212,928 → 45,427,委托价值中位数 4,120 → 4,429。 - 奖励来源:国库库存(发布时预留)、宝库(交付给悬赏的稀有物品,之后作为奖励发出)、制作(每周上限:高级 16 个、特别 1 个、UU 物质 4,000 mB,发布时计入)、货币(支付时发行)。无法支付时普通与高级改为货币,特别不发布;到期时未支付的预留退回。悬赏需要的商品市场不出售。交付按比例立即支付(
first)或完成 / 到期时按贡献分配(contribution,最大余数法)。 - 期货机面板第 7 页“悬赏”(交付、奖励来源、份额、本周的发行余量与宝库);指南《期货交易机》三语更新;
/mio_icif bounty list / tables / post / cancel / errors / settings(权限 2)。 - 数据包
data/<命名空间>/si_exchange/:bounties、rewards、listings(开服即上市、不下市)、economy.json(配置中修改过的值 > 数据包 > 默认值)、commodities(si_futures 格式)。商品新字段tag、accept_components(出售箱与集中清算)。默认内容:高级 7、特别 4、普通 3 个表,三个奖励池。文档docs/datapacks/si_exchange.md。 - 公共 API 只增不改:
core.api.exchange(ExchangeApi、BountySpec等与事件MarketDayEvent、ExchangeTradeEvent、BountyPostingEvent、BountyCompletedEvent);API 指南第 22 节。 - 新配置 20 项(
bounty*);语言tools/lang_1126.py;新检查check_1126.py;离线测试Harness1126(31 项);GameTestsRound1126GameTests(7 项,未运行)。
0.1.10.25
基于 0.1.10.24。交易所设计方案(Q1~Q8 按推荐批准)的第一阶段 X1:买卖额度分离、两级卖出深度、每日集中清算、出售箱、管理命令、工业品。每项先在离线测试中与 0.1.10.24 的规则比较,只采用确认有改进的部分。详细说明见 docs/releases/0.1.10.25/UPDATE.md。
- 买卖额度分离与两级卖出深度:取消每名玩家每天 100 个(买卖合计)与每种商品每天 256 个的收购预算。买入改为单独的设置
buyItemsPerDay(默认 0 = 不限,由国库库存与进口限制)。卖出没有数量上限,买价按“自己今天卖出的量”(个人深度sellDepthPlayer256)与“所有人卖出且尚未转出的卖压”(服务器深度sellDepthServer1024,每天转出sellExportPerDay50%)连续下降(SellDepth)。离线(铁锭,价格 200):16 / 64 个得到之前的 97% / 89%,1,000 个由之前只能卖 100 个的 18,263 变为全部卖出 39,430;一名玩家每天最多 39,731,所有玩家(任意数量的账号)第一天最多 198,656、之后每天约 47,808(之前每天 42,214)。服务器深度未采用方案的 2,048(每天约 92,288)。 - 每日集中清算:出售可以交给集中清算,在交易日更替时与所有人的一起出售。每人的个人深度只作用于自己的数量,服务器深度作用于总量;单独出售时与立即出售所得相同,结果与提交顺序无关,可设最低平均价(百分比)。五人各卖 1,000 个铁锭:依次立即出售 39,430 → 793,集中清算各 22,069。方案中“所有人同一单价”未采用(一人 1,000 个可得 123,841,绕过个人深度)。
- 价值上限:多人游戏中市场每天最多付给每名玩家 50,000 货币的卖出所得(
sellAllowancePerDay,未用完的最多保留 3 天),单人游戏默认不启用;玩家之间的交易不计入。 - 出售箱:期货机面板新增第 6 页“出售”:18 格出售箱(Shift 点击,“放入”放入快捷栏以外所有可出售的物品;改名、附魔、损坏的物品不接受),每行显示立即出售与集中清算的所得,可立即出售或交给集中清算,列出自己的清算订单并可撤回;关闭面板时箱中物品退回。订单的数量上限 1,024 → 100,000(期货仍为 1,024)。卖单先与不低于市场买价的玩家买单成交,再卖给市场,最后与其余买单成交到限价。
- 管理命令(权限 2):
/mio_icif market simulate <物品> <数量> [玩家]、arbitrage [条数]、calibrate、ledger [天数]。 - 工业品与套利审计:新类别“工业品”(
si_futures/industrial.json,11 种:锡、青铜、铅、银、精炼铁、橡胶、电路板、高级电路板、高级合金、碳板、铱);数据包商品可设置depth、export_per_day、import_per_day、import_premium。审计(ArbitrageAudit)在 0.1.10.24 的配方中发现 38 种会自我增殖的物品,修正两个配方后为 0:空锡罐 → 锡粉改为 2 个对 1 个(IC2 1.12.2),石英楼梯 → 石英 6 → 4。原矿(铁、铜、金)不再进口,曲奇 20 → 11。GameTests:Round1125GameTests(9 项)。检查工具check_1125.py,离线测试verify-si1125(45 项)。
0.1.10.24
基于 0.1.10.23。玩家报告与素材更新共 3 项。详细说明见 docs/releases/0.1.10.24/UPDATE.md。
- 发酵机的热量(对齐 IC2 1.12.2
TileEntityFermenter.work(),按 IC2 2.8.222-ex112 的字节码核对):只在能开始一次运行的 tick 吸收热量(输入罐有一次运行的生物质、沼气罐有一次运行的空间),热源推入与本机抽取合计每 tick 最多 100 HU(FermenterHeatIntake)。此前 4000 HU 的储热在没有生物质或沼气罐已满时也会吸满(离线:空转 1 分钟吸收 4,000 HU,沼气罐满时多吸 4,000 HU),热源为此消耗燃料,之后的最初几次运行快于热源。热量条改为本次运行已获得的热量(IC2 的转换进度),工作灯不再因残留热量而常亮。稳定运行时每分钟的运行次数、沼气与肥料此前已与 IC2 一致,配方数值(20 mB → 400 mB、4000 HU、每 500 mB 一个肥料)与 IC2 1.12.2 相同,未改;IC2 1.7.10 的默认值(10 mB → 200 mB、8000 HU)未采用。 - 全屏面板与 JEI / EMI:期货机界面的
leftPos停在屏幕宽度的一半,JEI 把左半屏当作空闲区域显示书签并抢走点击。期货机与电网监控面板现向 JEI(GUI 额外区域)和 EMI(排除区域)声明整个窗口;EMI 也避开机器界面左侧的工具栏(与 JEI 相同)。 - 素材:用户的 11 个工具与武器 3D 模型及其纹理(等离子空气炮使用较新的文件)、43 张纹理与 2 个动画文件(量子核心为新动画)。65 处模型元素旋转(±30 度、双轴等,1.21.1 不加载)改为最近的允许角度。移除内置资源包「实验性风格」(
resourcepacks/si_experimental,680 个文件)、AppearanceStyle与装备控制台“外观”页的风格卡片,该页第一页改为“装甲”(护手覆盖、框架核心外装)。GameTests:Round1124GameTests(3 项)。检查工具check_1124.py,离线测试verify-si1124(64 项)。
0.1.10.23
基于 0.1.10.22。核心系统改进方案第二阶段之三:行为预测 P2、P4~P7。每项先离线与旧做法逐项比较,只采用确认有改进的部分。详细说明见 docs/releases/0.1.10.23/UPDATE.md。
- 诱导弹转向律(P4):比例导引不加重力补偿时比纯追踪多转约 7%;加上重力补偿后,满弓箭在同等命中率下每发少转约 1 度(预测诱导协同的 APN 少转 1.4 度)。仅对发射速度 ≥ 2.0 格/tick 且有重力的弹采用;低速弹、无重力弹,以及(无协同时)对飞行目标保持之前的纯追踪,因为那些情况下比例导引没有更好。
- 运动模型与数据链(P2):预测路径按游戏的移动规则计算(方块滑度、水中阻力、走上半砖、跳上一格台阶、掉落、被击退后的滑行、追击者在面前停下);离线 1 秒路径误差:半砖 3.77 → 0.03 格,走入水中 1.28 → 0.67,被击退 2.78 → 0.95。行为预测器 Mk.II 接收 8 个生物的寻路节点(新数据包
suit_path_nodes),沿转角的路径误差 4.04 → 1.36。 - 锁定保持(P5):锁定的目标躲到遮挡物后时保持锁定(Mk.I 1 秒、Mk.II 2 秒),在预测位置附近重新出现即恢复,传送的目标不会被误认。远处不构成威胁的生物每 4 tick 更新路径(计算量 63%,之字形行走者的路径偏差 0.18 → 0.58 格)。
- 优先级与分配(P6):行为预测器显示最危险的三个生物 #1~#3;诱导弹由轮流改为按“优先级 × 命中概率 × 剩余生命份额”贪心分配(扣除飞行中的弹)。离线:每轮击杀 0.51 → 1.00,每次击杀的过量伤害 5.25 → 6.03。
- 来袭弹警告与攻击倒计时(P7):威胁传感器显示来袭的箭与火球的方向、秒数与闪避方向(离线:警告命中的准确率 98.6%,按闪避方向躲开 77%,随机方向 44%);行为预测器的攻击标记显示倒计时(拉弓、装填、蓄力、膨胀、学习的近战间隔)。GameTests:
Round1123GameTests(6 项)。检查工具check_1123.py,离线测试verify-si1123(127 项)。
0.1.10.26
Based on 0.1.10.25. Phase 2 X3 of the Exchange Design Plan: three tiers of system bounties, Treasury and Vault circulation, weekly issuance caps, the si_exchange datapack, Java API, and events. The new rules were first compared against the 0.1.10.25 market commissions in the offline test Harness1126. See docs/releases/0.1.10.26/UPDATE.md for details.
- System Bounties (
BountyBoard): Normal (maintains 35 items daily, items lacking in the market, value 4,000 × 0.51.5, 3 days, 1.15x reward), Advanced (2~3 items weekly, item sets from datapacks, 7 days, 1.3x reward, in multiplayer a single player can contribute at most 50% before the final day), Special (5% daily, rare items, 7 days). Replaces the 3 daily market commissions paid entirely in newly issued currency (old commissions in saves remain until expiry, configsystemCommissionsPerDayremoved). Offline one-year simulation: weekly currency issuance 708,416 → 146,761, single maximum currency reward 1,212,928 → 45,427, median commission value 4,120 → 4,429. - Reward Sources: Treasury stock (reserved at posting), Vault (rare items delivered to bounties, later distributed as rewards), Crafting (weekly caps: Advanced 16, Special 1, UU Matter 4,000 mB, counted at posting), Currency (issued upon payment). If payment is impossible, Normal and Advanced convert to currency, while Special is not posted; unpaid reservations are refunded upon expiry. Items required by bounties are not sold by the market. Delivery pays immediately proportional to contribution (
first), or distributes proportionally upon completion / expiry (contribution, Largest Remainder Method). - Futures Machine Panel Page 7 "Bounties" (delivery, reward sources, shares, remaining issuance this week, and Vault); Guide "Futures Trading Machine" updated in three languages;
/mio_icif bounty list / tables / post / cancel / errors / settings(permission 2). - Datapack
data/<namespace>/si_exchange/:bounties,rewards,listings(listed on server start, never delisted),economy.json(modified values in config > datapack > default),commodities(si_futuresformat). New commodity fieldstag,accept_components(sell box and centralized clearing). Default contents: Advanced 7, Special 4, Normal 3 tables, three reward pools. Documentationdocs/datapacks/si_exchange.md. - Public API is additive only:
core.api.exchange(ExchangeApi,BountySpec, etc., and eventsMarketDayEvent,ExchangeTradeEvent,BountyPostingEvent,BountyCompletedEvent); API Guide Section 22. - 20 new configs (
bounty*); languagetools/lang_1126.py; new checkcheck_1126.py; offline testHarness1126(31 items); GameTestsRound1126GameTests(7 items, not run).
0.1.10.25
Based on 0.1.10.24. Phase 1 X1 of the Exchange Design Plan (Q1~Q8 approved as recommended): separation of buy/sell quotas, two-tier sell depth, daily centralized clearing, sell box, admin commands, industrial goods. Each item was first compared against the 0.1.10.24 rules in offline tests, and only confirmed improvements were adopted. See docs/releases/0.1.10.25/UPDATE.md for details.
- Separation of Buy/Sell Quotas and Two-Tier Sell Depth: Removed the previous daily caps of 100 items per player (buy + sell combined) and a 256-item buy budget per commodity. Buying now uses a separate setting
buyItemsPerDay(default 0 = unlimited, limited by treasury stock and imports). Selling has no quantity cap; the buy price decreases continuously based on "your own amount sold today" (personal depthsellDepthPlayer256) and "total sell pressure from all players not yet exported" (server depthsellDepthServer1024, exportingsellExportPerDay50% daily) (SellDepth). Offline (Iron Ingot, price 200): 16 / 64 items yield 97% / 89% of the previous price; 1,000 items went from 18,263 (previously capped at 100) to 39,430 for selling everything; a single player makes at most 39,731 daily, and all players (any number of accounts) make at most 198,656 on the first day and about 47,808 daily thereafter (previously 42,214 daily). The rejected server depth of 2,048 (about 92,288 daily) was not adopted. - Daily Centralized Clearing: Sales can be submitted to centralized clearing, sold together with everyone else's at the daily rollover. Personal depth applies only to your own quantity, server depth applies to the total; selling individually yields the same as immediate selling, results are independent of submission order, and a minimum average price (percentage) can be set. Five players each selling 1,000 Iron Ingots: immediate selling sequentially 39,430 → 793, centralized clearing 22,069 each. The proposal for "everyone same unit price" was not adopted (one person could get 123,841 for 1,000 items, bypassing personal depth).
- Value Cap: In multiplayer, the market pays at most 50,000 currency per player per day for sales (
sellAllowancePerDay, unused amounts retained for up to 3 days), disabled by default in singleplayer; player-to-player trades do not count. - Sell Box: New page 6 "Sell" on the Futures Machine panel: 18-slot sell box (Shift-click, "Place" puts all sellable items outside the hotbar; renamed, enchanted, or damaged items are rejected), each row shows proceeds for immediate selling and centralized clearing, can sell immediately or submit to centralized clearing, lists your own clearing orders and allows withdrawal; items in the box are returned when the panel is closed. Order quantity cap 1,024 → 100,000 (Futures remains 1,024). Sell orders first match with player buy orders at or above the market buy price, then sell to the market, then match with remaining buy orders down to the limit price.
- Admin Commands (permission 2):
/mio_icif market simulate <item> <quantity> [player],arbitrage [count],calibrate,ledger [days]. - Industrial Goods and Arbitrage Audit: New category "Industrial Goods" (
si_futures/industrial.json, 11 types: Tin, Bronze, Lead, Silver, Refined Iron, Rubber, Circuit Board, Advanced Circuit Board, Advanced Alloy, Carbon Plate, Iridium); datapack commodities can setdepth,export_per_day,import_per_day,import_premium. Audit (ArbitrageAudit) found 38 self-multiplying items in the 0.1.10.24 recipes, now 0 after fixing two recipes: Empty Tin Can → Tin Dust changed to 2 to 1 (IC2 1.12.2), Quartz Stairs → Quartz 6 → 4. Raw ores (Iron, Copper, Gold) are no longer imported, Cookie 20 → 11. GameTests:Round1125GameTests(9 items). Check toolcheck_1125.py, offline testverify-si1125(45 items).
0.1.10.24
Based on 0.1.10.23. 3 items total for player reports and asset updates. See docs/releases/0.1.10.24/UPDATE.md for details.
- Fermenter Heat (aligned with IC2 1.12.2
TileEntityFermenter.work(), verified against IC2 2.8.222-ex112 bytecode): Absorbs heat only on ticks when it can start a run (input tank has one run of biomass, gas tank has room for one run), heat sources pushing in and the machine extracting combined at most 100 HU per tick (FermenterHeatIntake). Previously, the 4,000 HU heat buffer would fill up even with no biomass or when the gas tank was full (offline: absorbs 4,000 HU idling for 1 minute, absorbs an extra 4,000 HU when gas tank is full), causing heat sources to consume fuel, and the first few subsequent runs were faster than the heat source. The heat bar now shows the heat acquired for this run (IC2's conversion progress), and the working light no longer stays on due to residual heat. Steady-state runs per minute, biogas, and fertilizer were already aligned with IC2; recipe values (20 mB → 400 mB, 4000 HU, 1 fertilizer per 500 mB) are the same as IC2 1.12.2 and unchanged; IC2 1.7.10 defaults (10 mB → 200 mB, 8000 HU) were not adopted. - Fullscreen Panels and JEI / EMI: The Futures Machine UI's
leftPoswas stuck at half the screen width, causing JEI to treat the left half-screen as free space, displaying bookmarks and stealing clicks. The Futures Machine and Grid Monitoring Panel now declare the entire window to JEI (GUI extra areas) and EMI (exclusion areas); EMI also avoids the toolbar on the left side of the machine UI (same as JEI). - Assets: User's 11 tool and weapon 3D models and their textures (Plasma Air Cannon uses the newer file), 43 textures and 2 animation files (Quantum Core has a new animation). 65 model element rotations (±30 degrees, dual-axis, etc., not loaded in 1.21.1) were changed to the nearest allowed angle. Removed the built-in resource pack "Experimental Style" (
resourcepacks/si_experimental, 680 files),AppearanceStyle, and the style card on the "Appearance" page of the equipment console; the first page is now "Armor" (gauntlet coverage, frame core exterior). GameTests:Round1124GameTests(3 items). Check toolcheck_1124.py, offline testverify-si1124(64 items).
0.1.10.23
Based on 0.1.10.22. Core system improvement plan phase 2 part 3: Behavioral Prediction P2, P4~P7. Each item was first compared offline item-by-item with the old approach, and only confirmed improvements were adopted. See docs/releases/0.1.10.23/UPDATE.md for details.
- Guided Projectile Steering Law (P4): Proportional navigation without gravity compensation turns about 7% more than pure pursuit; with gravity compensation, a full-draw bow turns about 1 degree less per shot at the same hit rate (APN with predictive guidance synergy turns 1.4 degrees less). Applied only to projectiles with launch speed ≥ 2.0 blocks/tick and gravity; low-speed projectiles, gravity-less projectiles, and (without synergy) flying targets retain the previous pure pursuit, as proportional navigation is not better in those cases.
- Motion Model and Datalink (P2): Predicted paths are calculated using the game's movement rules (block slipperiness, water drag, stepping onto slabs, jumping up 1 block, falling, sliding after knockback, pursuers stopping in front); offline 1-second path error: slabs 3.77 → 0.03 blocks, walking into water 1.28 → 0.67, after knockback 2.78 → 0.95. Behavior Predictor Mk.II receives pathfinding nodes of 8 mobs (new packet
suit_path_nodes), path error around corners 4.04 → 1.36. - Lock Retention (P5): When a locked target hides behind cover, the lock is retained (Mk.I 1 second, Mk.II 2 seconds), restoring when the target reappears near the predicted position; teleporting targets are not misidentified. Distant non-threatening mobs have their paths updated every 4 ticks (computation 63%, path deviation for zigzag walkers 0.18 → 0.58 blocks).
- Priority and Allocation (P6): The Behavior Predictor displays the top three most dangerous mobs #1~#3; guided projectiles change from round-robin to greedy allocation based on "priority × hit probability × remaining health share" (deducting projectiles in flight). Offline: kills per round 0.51 → 1.00, overkill damage per kill 5.25 → 6.03.
- Incoming Projectile Warning and Attack Countdown (P7): The Threat Sensor displays the direction, seconds, and dodge direction of incoming arrows and fireballs (offline: warning hit accuracy 98.6%, dodging in the suggested direction 77%, random direction 44%); the Behavior Predictor's attack marker displays a countdown (drawing bow, reloading, charging, swelling, learned melee intervals). GameTests:
Round1123GameTests(6 items). Check toolcheck_1123.py, offline testverify-si1123(127 items).