Archived
feat(loop-engine): 配方执行器 —— project / prune / say,每一笔之前刷读数
三条硬规矩:每一笔之前刷读数(后一笔要数前一笔刚写下的行,不刷新就恒为 0); 值没变的写静默跳过(配方不是模型,原地打转不必被看见); 出错交人不重试不退回模型(配方抛错是包的 bug,退回模型就是让它去凑)。 主键算不出来的那一行先不写 —— 空主键写下去是一行永远对不上的垃圾。 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LKMZDwMbyKzjw6Hembwvet
This commit is contained in:
@@ -9,6 +9,7 @@ import type { AgentAction, LoopEngineService, TurnOutcome } from '../../contract
|
||||
import { ForbiddenTerminalError, TURN_ACTION_CAP } from '../../contract.ts';
|
||||
import type {
|
||||
AgentPort,
|
||||
AnalyzeGateway,
|
||||
CardGateway,
|
||||
Clock,
|
||||
ListenerRegistrar,
|
||||
@@ -53,6 +54,7 @@ export type TurnRunnerDeps = {
|
||||
readonly skills: SkillGateway;
|
||||
readonly tasks: TaskContext;
|
||||
readonly clock: Clock;
|
||||
readonly analyze: AnalyzeGateway;
|
||||
};
|
||||
|
||||
/**
|
||||
|
||||
Reference in New Issue
Block a user