feat(subagent): v0.7.5 — 适配 kimi-code 0.36.0 子代理模型池(双写兼容/池管理/force/前置校验)+ 热力图详情按需加载 + models.dev 刷新(GLM-5.3)
Release / Version consistency (push) Canceled after 0s
Release / Build (macos-latest) (push) Canceled after 0s
Release / Build (ubuntu-latest) (push) Canceled after 0s
Release / Build (windows-latest) (push) Canceled after 0s
Release / Attach macOS install script (push) Canceled after 0s

This commit is contained in:
KimiSwitch Dev committed 2026-08-15 23:21:22 +08:00
1 parent 89c71fbfe6
commit 9a3875d990
19 files changed
+152041 -140593

No files matched your search

+19 -4
View File
@@ -134,13 +134,28 @@ export interface AgentSettings {
}
/**
* Kimi Code `[secondary_model]` section of config.toml: the secondary model
* bound to newly spawned subagents (experimental feature).
* `model` points at an alias defined in `[models]`; the remaining fields are
* a patch applied on top of the referenced entry (subagents only).
* Kimi Code `[secondary_model]` section of config.toml: the subagent model
* pool (experimental feature).
*
* v2 engine (default) reads the pool keys: `default_model` + `[models]` table
* (alias → description) + `force`. `model` is ignored by the v2 validator when
* pool keys are present, but is kept in sync with `default_model` so the v1
* engine (KIMI_CODE_LEGACY_FLAG=1) resolves the same alias.
*
* v1 engine reads only the recipe key `model` plus the patch fields; pool keys
* are ignored by it. Pool resolution order: `models` table → `default_model` →
* `model`.
*/
export interface SecondaryModelConfig {
/** v2: default subagent alias in the pool (a key of `models`). */
default_model?: string;
/** v2: alias → description table for the subagent model pool. */
models?: Record<string, string>;
/** v2: force subagents onto the default alias (excludes `models`). */
force?: boolean;
/** Legacy v1 recipe key — written in sync so KIMI_CODE_LEGACY_FLAG=1 works. */
model?: string;
/** v1 patch fields, applied on top of the referenced entry (v1 only). */
default_effort?: string;
max_output_size?: number;
max_context_size?: number;