Coder/deepseek-harness
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: identify and freeze messages at creation

Browse Source
This commit is contained in:
_Kerman
2026-07-28 13:55:59 +08:00
parent c49c0ba497
commit fbf87e660c
345 changed files with 5220 additions and 2901 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
docs/core-data-structures/tools.zh.md
View File

@@ -215,7 +215,7 @@ interface ToolRunContext extends ToolExecution {
* the agent loop. Contexts retain their individual source and metadata and
* are emitted in call order.
*/
deferContext(context: UserMessageData): void
deferContext(context: UserMessage): void
/**
* Mark a successful final result as terminal for the current agent turn.
* The marker rides this execution's own result (`concludesTurn` exists only
@@ -329,7 +329,7 @@ interface ToolExecutionSuccess {
readonly content: ContentBlock[]
readonly error?: never
readonly meta?: JsonValue
readonly additionalContexts?: UserMessageData[]
readonly additionalContexts?: UserMessage[]
/** The agent loop stops after committing this successful result batch. */
readonly concludesTurn?: true
}
@@ -343,7 +343,7 @@ interface ToolExecutionFailure {
readonly value?: never
readonly content: ContentBlock[]
readonly meta?: JsonValue
readonly additionalContexts?: UserMessageData[]
readonly additionalContexts?: UserMessage[]
readonly concludesTurn?: never
}
```
@@ -380,9 +380,9 @@ type PreToolDecision =
* next request, or block by turning corrective feedback into an error result.
*/
type PostToolDecision =
| { kind: 'accept'; content?: ContentBlock[]; value?: never; additionalContexts?: UserMessageData[] }
| { kind: 'accept'; value: JsonValue; content?: never; additionalContexts?: UserMessageData[] }
| { kind: 'block'; feedback: ContentBlock[]; additionalContexts?: UserMessageData[] }
| { kind: 'accept'; content?: ContentBlock[]; value?: never; additionalContexts?: UserMessage[] }
| { kind: 'accept'; value: JsonValue; content?: never; additionalContexts?: UserMessage[] }
| { kind: 'block'; feedback: ContentBlock[]; additionalContexts?: UserMessage[] }
```
调用 `next()` 获取默认决策，或直接返回一个决策以短路。前置策略可以 deny 或 ask；只有 `allowed-once` 才继续执行，而未授权、缺少审批通道或服务、或无 agent 的请求都会变为拒绝。Guard 仍可施加最终拒绝。参数不可被改写，因为历史记录、审计、UI 和执行必须保持一致。