Parallel API 1.0.10 — Changes since 1.0.3

中文

• 新增服务端配置 parallelapi-takeover.toml，可按 Mod 命名空间、机器 ID 配置配方接管、并行倍率、虚拟资源和界面显示，也可设置排除项。
• 增加对兼容 Mekanism CachedRecipe 机器、AE Inscriber 系列及符合受支持配方接口的 AE 机器的配方接管。配方输入、产物、耗能按同一实际批次数结算，并保存进行中的批次。
• 增加精确虚拟物品、流体和能量缓存。大数值保留在内部计算中，旧接口只在传输边界转换。
• 增加统一并行显示；修复客户端启动时的槽位数量注入崩溃。
• 修复 AE 供能的固定请求上限，改为按实际批次需求请求能量，并按电网实际供能量结算。
• 修复 AE 物品数量重复显示及自动输出问题：合并实体与虚拟库存，不再额外限制为每次最多 64 个；仍受接收端和外部接口上限约束。

兼容说明： AE 原生电网接口使用有限精度的 double；机器内部缓存使用 Flux BigNumber。直接通过 Flux 大数接口输入时保留大数精度。配置只能启用符合已支持控制器协议的机器，未知控制器仍使用原有处理方式。

English

• Added the server-side parallelapi-takeover.toml configuration. Select compatible machines by mod namespace or machine ID, configure parallel multipliers, virtual resources and UI display, and define exclusions.
• Added recipe takeover for compatible Mekanism CachedRecipe machines, AE Inscribers and AE machines using supported recipe interfaces. Inputs, outputs and energy use the same actual batch size, and in-progress batches are persisted.
• Added exact virtual item, fluid and energy buffers. Large values remain exact in internal calculations and are converted only at legacy transfer boundaries.
• Added a unified parallelism display and fixed a client-startup crash caused by the slot-count mixin.
• Removed the fixed AE energy request ceiling. Energy requests now follow actual batch demand, and processing is charged only for energy extracted from the grid.
• Fixed duplicate AE item-count labels and automatic output. Physical and virtual inventories are combined, with no extra 64-item limit per export call; receiver and external API limits still apply.

Compatibility note: The native AE grid API uses finite-precision double values, while the machine-owned buffer uses Flux BigNumber. Direct Flux BigNumber input preserves large-number precision. Configuration enables only machines matching supported controller protocols; unknown controllers keep their original behavior.