feat(server): W2-D agent assembly, session API, runAgent/resolveDecision and backend e2e

- Capabilities: all 14 registered (AO: inventory.query, download.plan/submit, organize.plan/execute, policy.preview/apply; agent: task.inspect/retry/cancel); startup asserts count == 14. Registry injects server-side authorize (automation_grant by action/scope/policy version/budget, one-shot approval bound to plan id + payload hash + revision + expiry; model claims ignored; automation switch), audit_event, server idempotency keys and a capability_invocation op ledger (migration 0021) used for crash reconciliation. download.plan creates draft plans only; download.submit promotes draft -> reserved after authorization. - Agent session API: POST/GET /agent/runs, GET /agent/runs/{id} (events by seq, messages, report), POST .../messages (wake), POST .../abort; automation-grants, approvals, capabilities invoke, tasks. - runAgent task: Pi or fake provider from settings (apiKey via secrets), events persisted to agent_run_event then published, resume + ledger reconciliation when pendingOperations > 0, waits for goal-linked downloads/imports and reports from business tables (submitted != imported). - Background triggers: decision.created -> bounded plan_only investigation (dedupe per decision); decision.updated(resolved) -> resolveDecision executes chosen option. - W2 closure: all OpenAPI routes registered in contracts api.ts (+ subset test); TMDB key from settings secret_ref with env fallback; PUT /sites/{id}/credentials documented as alias of /site-accounts; completion events documented as existing status events. - e2e (real PG + OpenWorkflow + testkit mocks): RSS->library, agent delegation with grant/deny, background continuation, worker kill -9 recovery (single qB add), auto investigation dedupe. Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-09-27 20:52:12 +08:00
1 parent 0dec0854df
commit 6280d52ec3
@@ -0,0 +1,147 @@
+import { TMDB_FIXTURES, type TmdbFixture } from "@umt/testkit";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import { INVESTIGATE_CAPABILITIES } from "../../src/modules/agent/runner.ts";
+import { investigateScript } from "./fixtures/scripts.ts";
+import { createE2E, type E2E, pgAvailable, runEvents, seedBase, waitFor } from "./harness.ts";
+
+// E5 后台唤醒:识别歧义产生 decision.created → outbox → resolveDecision → autoInvestigate 开启时入队有界 runAgent
+// (plan_only、白名单只读 + identity.resolve、小预算、trigger=event);同一 decision 只一个 run(dedupe_key)。
+
+const hasPg = await pgAvailable();
+
+const twin = (id: number): TmdbFixture => ({
+  type: "movie",
+  id,
+  original_title: "Twins",
+  original_language: "en",
+  origin_country: ["US"],
+  date: "1988-12-09",
+  titles: { "zh-CN": "龙兄鼠弟", "en-US": "Twins" },
+  popularity: 10 + id / 1000,
+  alternative_titles: [],
+  external_ids: {},
+});
+
+describe.skipIf(!hasPg)("E5 后台唤醒有界调查", () => {
+  let e: E2E;
+  let siteId: string;
+
+  beforeAll(async () => {
+    e = await createE2E("e5", {
+      fakeScripts: [investigateScript],
+      tmdbFixtures: [...TMDB_FIXTURES, twin(111), twin(112)],
+    });
+    siteId = (await seedBase(e)).siteId;
+    await e.settings({ agent: { autoInvestigate: true, investigateMaxToolCalls: 6 } });
+  }, 120_000);
+
+  afterAll(async () => {
+    await e?.close();
+  }, 60_000);
+
+  it("识别歧义 → 自动发起只读调查 → 结论回写待处理事项;重复事件只一个 run", async () => {
+    const rel = await e.db
+      .insertInto("source_release")
+      .values({ site_id: siteId, site_release_id: "990001", title: "Twins.1988.1080p.BluRay.x264-GRP" })
+      .returning("id")
+      .executeTakeFirstOrThrow();
+    await e.booted.ctx.engine.enqueue(
+      "resolveObservation",
+      { sourceReleaseId: rel.id },
+      { idempotencyKey: `e5:resolve:${rel.id}`, traceId: rel.id },
+    );
+    const decision = await waitFor(
+      async () =>
+        e.db
+          .selectFrom("decision")
+          .selectAll()
+          .where("kind", "=", "identity_ambiguous")
+          .where("subject_id", "=", rel.id)
+          .executeTakeFirst(),
+      60_000,
+      "识别歧义 decision",
+    );
+    const run = await waitFor(
+      async () =>
+        e.db
+          .selectFrom("agent_run")
+          .selectAll()
+          .where("dedupe_key", "=", `investigate:${decision.id}`)
+          .executeTakeFirst(),
+      60_000,
+      "自动调查 run",
+    );
+    expect(run).toMatchObject({ trigger: "event", mode: "plan_only", created_by: "system:auto-investigate" });
+    expect(run.goal).toContain(decision.id);
+    expect(run.scope).toMatchObject({
+      capabilities: [...INVESTIGATE_CAPABILITIES],
+      allowedKinds: ["read", "preview", "propose"],
+    });
+    expect(run.budget).toMatchObject({ maxToolCalls: 6 });
+
+    const finished = await waitFor(
+      async () => {
+        const r = await e.db.selectFrom("agent_run").selectAll().where("id", "=", run.id).executeTakeFirstOrThrow();
+        return ["succeeded", "failed", "budget_exceeded"].includes(r.status) ? r : null;
+      },
+      60_000,
+      "调查 run 结束",
+    );
+    expect(finished.status).toBe("succeeded");
+    const events = await runEvents(e, run.id);
+    const invoked = events
+      .filter((x) => x.type === "tool_call_start")
+      .map((x) => (x.payload as { capabilityId: string }).capabilityId);
+    expect(invoked).toEqual(["catalog.lookup", "inventory.query"]);
+    // 白名单外的写能力:被拒绝,没有执行
+    const rejected = events.filter((x) => x.type === "tool_call_rejected");
+    expect(rejected.map((x) => (x.payload as { toolName: string }).toolName)).toEqual(["download__submit"]);
+    expect(
+      await e.db.selectFrom("capability_invocation").select("id").where("agent_run_id", "=", run.id).execute(),
+    ).toHaveLength(0);
+    // 结论回写:待处理事项交还人工并带调查摘要
+    const d2 = await e.db.selectFrom("decision").selectAll().where("id", "=", decision.id).executeTakeFirstOrThrow();
+    expect(d2.status).toBe("awaiting_user");
+    expect(d2.agent_run_id).toBe(run.id);
+    expect((d2.evidence as { agentInvestigation?: { runId: string } }).agentInvestigation?.runId).toBe(run.id);
+
+    // 同一根因的重复事件(不同 outbox 键)与事项重新打开后的再次触发:仍只有一个 run
+    await e.db.updateTable("decision").set({ status: "open" }).where("id", "=", decision.id).execute();
+    for (const k of ["dup-1", "dup-2"]) {
+      await e.booted.ctx.outbox.emit(
+        e.db,
+        "decision.created",
+        { decisionId: decision.id, kind: "identity_ambiguous", severity: "warning", title: decision.title },
+        { idempotencyKey: `e5:${k}` },
+      );
+    }
+    await waitFor(
+      async () => {
+        const rows = await e.db
+          .selectFrom("outbox_event")
+          .select("status")
+          .where("idempotency_key", "in", ["e5:dup-1", "e5:dup-2"])
+          .execute();
+        return rows.length === 2 && rows.every((r) => r.status === "dispatched");
+      },
+      30_000,
+      "重复事件已投递",
+    );
+    // 等 resolveDecision 跑完
+    await waitFor(
+      async () => {
+        const runs = await e.booted.ctx.engine.list({ taskName: "resolveDecision" });
+        return runs.length >= 3 && runs.every((r) => r.status === "completed") ? runs : null;
+      },
+      30_000,
+      "resolveDecision 完成",
+    );
+    const runs = await e.db
+      .selectFrom("agent_run")
+      .select("id")
+      .where("dedupe_key", "=", `investigate:${decision.id}`)
+      .execute();
+    expect(runs).toHaveLength(1);
+    expect(await e.db.selectFrom("agent_run").select("id").execute()).toHaveLength(1);
+  }, 180_000);
+});