mirror of
https://bitbucket.org/siakitem/my-pi.git
synced 2026-10-26 10:45:22 +00:00
云服务部-叶林立andDevin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
7e6a8faae1
AutoReview： - 审查器回到 Codex Guardian 风格的对话式审查，默认跟随会话当前主模型 （每次审查实时读取 ctx.model）；provider/model 必须同时设置才固定审查 模型，只设其一视为配置错误并转人工。 - 结论只通过 approve / reject 两个工具给出，参数仅 reason；不强制 toolChoice，正文文本（即使是合法 JSON）一律忽略。 - approve 自动放行；reject 转人工并在弹窗显示理由（标注模型生成）。 超时、Provider 错误、截断、未调用工具、多次调用、参数无效等异常同样 转人工，并附固定的系统说明；用户按 Esc 中止时直接拒绝。 - 模型调用改为 modelRegistry.streamSimple(...).result()：Pi 1.3.0 面向 扩展的 ModelRegistry 没有 completeSimple，原写法在真实环境中每次都会 转人工。 - 对话记录保留可信度标注；工具结果在有更短的 pi-compaction 摘要时替换 为摘要；token 估算改为非 ASCII 字符按 1 token 计；预算可配置。 - 同一条用户消息内完全相同的请求只缓存 approve 结论。 - 默认 reasoning low、总超时 30 秒、最多 2 次尝试、maxOutputTokens 4000。 删除： - 未发布过的 TypeSafe Jev 分类器实现、DenialCircuitBreaker。 - 整个 annotator 机制（model-explanation、注册表、registerAnnotator、 provider-timeout）；annotators 配置键保留为已弃用并忽略，避免旧配置 被拒绝。 测试： - 新增 reviewer / transcript / 弹窗说明 / 默认配置测试；清理重复、恒真 或只测实现细节的旧测试。 - 使用 deepseek-v4.1-flash 实测 5 个场景各 5 次，25 次结论全部符合预期。 Generated with [Devin](https://devin.ai) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
56 lines
2.2 KiB
TypeScript
56 lines
2.2 KiB
TypeScript
import { describe, expect, it } from "vitest";
|
|
|
|
import {
|
|
type RequestedToolValidation,
|
|
validateRequestedTool,
|
|
} from "#src/handlers/permission-gate-handler";
|
|
|
|
// ── helpers ────────────────────────────────────────────────────────────────
|
|
|
|
function makeTools(names: string[]): { name: string }[] {
|
|
return names.map((name) => ({ name }));
|
|
}
|
|
|
|
const TOOLS = makeTools(["read", "bash", "edit"]);
|
|
|
|
// ── validateRequestedTool ──────────────────────────────────────────────────
|
|
|
|
describe("validateRequestedTool", () => {
|
|
describe("missing / unresolvable tool name", () => {
|
|
it("blocks when event has no name field", () => {
|
|
const result = validateRequestedTool({ type: "tool_call" }, TOOLS);
|
|
expect(result.status).toBe("block");
|
|
expect(
|
|
(result as Extract<RequestedToolValidation, { status: "block" }>)
|
|
.reason,
|
|
).toBeTruthy();
|
|
});
|
|
|
|
});
|
|
|
|
describe("unregistered tool", () => {
|
|
it("blocks when the tool name is not in the registered list", () => {
|
|
const result = validateRequestedTool({ name: "unknown-tool" }, TOOLS);
|
|
expect(result.status).toBe("block");
|
|
});
|
|
|
|
});
|
|
|
|
describe("registered tool (ok path)", () => {
|
|
it("returns ok with the raw tool name for a known tool", () => {
|
|
const result = validateRequestedTool({ name: "read" }, TOOLS);
|
|
expect(result).toEqual({ status: "ok", toolName: "read" });
|
|
});
|
|
|
|
it("returns the raw name as it appeared in the event (not normalised)", () => {
|
|
// If an alias mechanism were to normalise "Read" → "read",
|
|
// validateRequestedTool still returns the raw value from the event.
|
|
// Without aliases the raw name and registered name are the same; this
|
|
// asserts the contract that toolName comes from the event, not from the
|
|
// registration lookup's normalizedToolName field.
|
|
const result = validateRequestedTool({ name: "bash" }, TOOLS);
|
|
expect(result).toEqual({ status: "ok", toolName: "bash" });
|
|
});
|
|
});
|
|
});
|