@stigmer/react 3.1.17 → 3.2.0

package/src/workflow/__tests__/task-presentation.test.ts

@@ -0,0 +1,479 @@
+// Unit tests for the headless task-presentation layer (T04): the built-in
+// per-kind preview lines across all 20 task kinds, the status-line
+// precedence, the disclosure taxonomy, defensive Struct parsing (summaries
+// are runner-dependent — malformed shapes must degrade, never break), and
+// the `registerTaskPresenter` override seam.
+
+import { describe, it, expect } from "vitest";
+import { WorkflowTaskKind } from "@stigmer/protos/ai/stigmer/agentic/workflow/v1/enum_pb";
+import type { JsonObject } from "@bufbuild/protobuf";
+import type { DerivedTaskState } from "../../internal/store/workflow-execution-event-store";
+import {
+  resolveTaskPreview,
+  registerTaskPresenter,
+  defaultDisclosureForKind,
+  valueSnippet,
+} from "../thread/task-presentation";
+
+function state(
+  overrides: Partial<DerivedTaskState> = {},
+): DerivedTaskState {
+  return {
+    taskName: "task",
+    taskKind: WorkflowTaskKind.set_vars,
+    status: "completed",
+    durationMs: 100,
+    costMicros: 0n,
+    tokensUsed: 0n,
+    attemptNumber: 1,
+    error: "",
+    childExecutionId: "",
+    agentSlug: "",
+    currentToolName: "",
+    messagesCount: 0,
+    toolCallsCount: 0,
+    inputSummary: null,
+    outputSummary: null,
+    approvalRequest: null,
+    approvalResolution: null,
+    ...overrides,
+  };
+}
+
+// ---------------------------------------------------------------------------
+// Status-line precedence (universal, kind-independent)
+// ---------------------------------------------------------------------------
+
+describe("status-line precedence", () => {
+  it("waiting_approval outranks any kind line", () => {
+    const { previewLine } = resolveTaskPreview(
+      state({
+        taskKind: WorkflowTaskKind.validate,
+        status: "waiting_approval",
+        outputSummary: { valid: false, errors: [{}] },
+      }),
+    );
+    expect(previewLine).toBe("Awaiting approval");
+  });
+
+  it("a failure shows the error's first line", () => {
+    const { previewLine } = resolveTaskPreview(
+      state({
+        taskKind: WorkflowTaskKind.validate,
+        status: "failed",
+        error: "[TypeError] expr.includes is not a function\nstack trace…",
+      }),
+    );
+    expect(previewLine).toBe("[TypeError] expr.includes is not a function");
+  });
+
+  it("skipped shows Skipped", () => {
+    const { previewLine } = resolveTaskPreview(state({ status: "skipped" }));
+    expect(previewLine).toBe("Skipped");
+  });
+
+  it("caps the line length (bounded header, not a body)", () => {
+    const { previewLine } = resolveTaskPreview(
+      state({ status: "failed", error: "x".repeat(500) }),
+    );
+    expect(previewLine.length).toBeLessThanOrEqual(80);
+    expect(previewLine.endsWith("…")).toBe(true);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Built-in per-kind lines — the full 20-kind table
+// ---------------------------------------------------------------------------
+
+describe("per-kind preview lines", () => {
+  it.each<{
+    label: string;
+    kind: WorkflowTaskKind;
+    overrides: Partial<DerivedTaskState>;
+    expected: string;
+  }>([
+    {
+      label: "set_vars from nested variables map",
+      kind: WorkflowTaskKind.set_vars,
+      overrides: {
+        inputSummary: { variables: { order_id: "o-1", total: "120" } },
+      },
+      expected: "set order_id, total",
+    },
+    {
+      label: "set_vars from top-level resolved variables",
+      kind: WorkflowTaskKind.set_vars,
+      overrides: { inputSummary: { a: "1", b: "2", c: "3", d: "4", e: "5" } },
+      expected: "set a, b, c +2 more",
+    },
+    {
+      label: "transform with engine and scalar output",
+      kind: WorkflowTaskKind.transform,
+      overrides: {
+        inputSummary: { engine: "JQ", expression: ".total" },
+        outputSummary: { result: 42 },
+      },
+      expected: "jq → 42",
+    },
+    {
+      label: "transform running (engine only, no output yet)",
+      kind: WorkflowTaskKind.transform,
+      overrides: { status: "running", inputSummary: { engine: "JSONATA" } },
+      expected: "jsonata",
+    },
+    {
+      label: "validate passing",
+      kind: WorkflowTaskKind.validate,
+      overrides: { outputSummary: { valid: true, errors: [] } },
+      expected: "valid",
+    },
+    {
+      label: "validate failing with error rows",
+      kind: WorkflowTaskKind.validate,
+      overrides: {
+        outputSummary: { valid: false, errors: [{ rule: "r1" }, { rule: "r2" }, { rule: "r3" }] },
+      },
+      expected: "3 errors",
+    },
+    {
+      label: "validate failing with one error (singular)",
+      kind: WorkflowTaskKind.validate,
+      overrides: { outputSummary: { valid: false, errors: [{ rule: "r1" }] } },
+      expected: "1 error",
+    },
+    {
+      label: "validate invalid without errors array",
+      kind: WorkflowTaskKind.validate,
+      overrides: { outputSummary: { valid: false } },
+      expected: "invalid",
+    },
+    {
+      label: "llm_call with model and text output",
+      kind: WorkflowTaskKind.llm_call,
+      overrides: {
+        inputSummary: { model: "claude-sonnet" },
+        outputSummary: { text: "The answer is 42." },
+      },
+      expected: 'claude-sonnet · "The answer is 42."',
+    },
+    {
+      label: "eval passing with score",
+      kind: WorkflowTaskKind.eval,
+      overrides: { outputSummary: { pass: true, score: 0.82 } },
+      expected: "pass · score 0.82",
+    },
+    {
+      label: "eval failing without score",
+      kind: WorkflowTaskKind.eval,
+      overrides: { outputSummary: { pass: false } },
+      expected: "fail",
+    },
+    {
+      label: "emit_event settled (CloudEvents type from output)",
+      kind: WorkflowTaskKind.emit_event,
+      overrides: { outputSummary: { type: "stigmer.ticket.classified" } },
+      expected: "emitted stigmer.ticket.classified",
+    },
+    {
+      label: "emit_event running (configured type from input)",
+      kind: WorkflowTaskKind.emit_event,
+      overrides: {
+        status: "running",
+        inputSummary: { event: { type: "stigmer.ticket.classified" } },
+      },
+      expected: "emitting stigmer.ticket.classified",
+    },
+    {
+      label: "notification delivered",
+      kind: WorkflowTaskKind.notification,
+      overrides: { outputSummary: { channel: "slack", delivered: true } },
+      expected: "sent via slack",
+    },
+    {
+      label: "notification delivery failed",
+      kind: WorkflowTaskKind.notification,
+      overrides: { outputSummary: { channel: "slack", delivered: false } },
+      expected: "delivery failed",
+    },
+    {
+      label: "human_input settled decision",
+      kind: WorkflowTaskKind.human_input,
+      overrides: { outputSummary: { outcome: "approve", reviewer: "suresh" } },
+      expected: "approve · by suresh",
+    },
+    {
+      label: "wait completed with structured duration",
+      kind: WorkflowTaskKind.wait,
+      overrides: { inputSummary: { duration: { seconds: 10 } } },
+      expected: "waited 10.0s",
+    },
+    {
+      label: "wait running",
+      kind: WorkflowTaskKind.wait,
+      overrides: {
+        status: "running",
+        inputSummary: { duration: { minutes: 1, seconds: 30 } },
+      },
+      expected: "waiting 1m 30s",
+    },
+    {
+      label: "wait until an absolute timestamp",
+      kind: WorkflowTaskKind.wait,
+      overrides: { inputSummary: { until: "2026-07-17T00:00:00Z" } },
+      expected: "waited until 2026-07-17T00:00:00Z",
+    },
+    {
+      label: "listen running",
+      kind: WorkflowTaskKind.listen,
+      overrides: { status: "running" },
+      expected: "waiting for signal",
+    },
+    {
+      label: "listen settled",
+      kind: WorkflowTaskKind.listen,
+      overrides: { status: "completed" },
+      expected: "signal received",
+    },
+    {
+      label: "agent_call running with live activity",
+      kind: WorkflowTaskKind.agent_call,
+      overrides: {
+        status: "running",
+        agentSlug: "blog-writer",
+        currentToolName: "web_search",
+        messagesCount: 7,
+        toolCallsCount: 3,
+      },
+      expected: "blog-writer · running web_search · 7 msgs · 3 tools",
+    },
+    {
+      label: "agent_call settled",
+      kind: WorkflowTaskKind.agent_call,
+      overrides: { agentSlug: "blog-writer", messagesCount: 12, toolCallsCount: 5 },
+      expected: "blog-writer · 12 msgs · 5 tools",
+    },
+  ])("$label", ({ kind, overrides, expected }) => {
+    const { previewLine } = resolveTaskPreview(state({ taskKind: kind, ...overrides }));
+    expect(previewLine).toBe(expected);
+  });
+
+  // Kinds that deliberately stay status-only: control flow (branch/fork
+  // detail needs graph topology the thread lacks — deferred), raise_error
+  // (the failure precedence carries the message), the invocation kinds
+  // (no verified output envelope — backend follow-up), and unspecified
+  // (the snapshot fallback).
+  it.each([
+    WorkflowTaskKind.switch_case,
+    WorkflowTaskKind.for_each,
+    WorkflowTaskKind.fork,
+    WorkflowTaskKind.try_catch,
+    WorkflowTaskKind.raise_error,
+    WorkflowTaskKind.http_call,
+    WorkflowTaskKind.grpc_call,
+    WorkflowTaskKind.activity_call,
+    WorkflowTaskKind.run_workflow,
+    WorkflowTaskKind.workflow_task_kind_unspecified,
+  ])("kind %d yields an empty line when completed cleanly", (kind) => {
+    const { previewLine } = resolveTaskPreview(
+      state({ taskKind: kind, outputSummary: { some: "data" } }),
+    );
+    expect(previewLine).toBe("");
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Defensive parsing — summaries are runner-dependent; never break
+// ---------------------------------------------------------------------------
+
+describe("defensive Struct parsing", () => {
+  it.each<{ label: string; kind: WorkflowTaskKind; input: JsonObject | null; output: JsonObject | null }>([
+    { label: "absent summaries", kind: WorkflowTaskKind.validate, input: null, output: null },
+    { label: "empty objects", kind: WorkflowTaskKind.set_vars, input: {}, output: {} },
+    {
+      label: "wrong-typed fields",
+      kind: WorkflowTaskKind.validate,
+      input: { variables: "not-an-object" },
+      output: { valid: "yes", errors: "none" },
+    },
+    {
+      label: "eval with non-boolean pass",
+      kind: WorkflowTaskKind.eval,
+      input: null,
+      output: { pass: "true", score: "high" },
+    },
+    {
+      label: "wait with unknown duration shape",
+      kind: WorkflowTaskKind.wait,
+      input: { duration: { fortnights: 2 } },
+      output: null,
+    },
+    {
+      label: "wait with non-numeric duration values",
+      kind: WorkflowTaskKind.wait,
+      input: { duration: { seconds: "ten" } },
+      output: null,
+    },
+  ])("degrades to empty on $label", ({ kind, input, output }) => {
+    const { previewLine } = resolveTaskPreview(
+      state({ taskKind: kind, inputSummary: input, outputSummary: output }),
+    );
+    expect(previewLine).toBe("");
+  });
+
+  it("handles an oversized variables map without an oversized line", () => {
+    const vars: Record<string, string> = {};
+    for (let i = 0; i < 100; i++) vars[`variable_number_${i}`] = "v";
+    const { previewLine } = resolveTaskPreview(
+      state({ taskKind: WorkflowTaskKind.set_vars, inputSummary: { variables: vars } }),
+    );
+    expect(previewLine.length).toBeLessThanOrEqual(80);
+    expect(previewLine.startsWith("set ")).toBe(true);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// valueSnippet
+// ---------------------------------------------------------------------------
+
+describe("valueSnippet", () => {
+  it.each<[JsonValueLike, string]>([
+    [null, "null"],
+    [42, "42"],
+    [true, "true"],
+    ["hello", '"hello"'],
+    ["line1\nline2", '"line1"'],
+    [[1, 2, 3], "[3 items]"],
+    [[1], "[1 item]"],
+    [{ a: 1, b: 2 }, "{2 fields}"],
+    // A single-field envelope unwraps to its scalar payload.
+    [{ result: 42 }, "42"],
+    [{ result: { nested: true } }, "{1 field}"],
+  ])("snippets %j as %s", (value, expected) => {
+    expect(valueSnippet(value)).toBe(expected);
+  });
+});
+
+type JsonValueLike = Parameters<typeof valueSnippet>[0];
+
+// ---------------------------------------------------------------------------
+// Disclosure taxonomy
+// ---------------------------------------------------------------------------
+
+describe("defaultDisclosureForKind", () => {
+  // T05 (DD-T05-5): every kind whose output CAN matter is a preview kind —
+  // the showBody gate keeps output-less cards as one-line rows, so preview
+  // disclosure costs nothing until the runner writes real output. Only
+  // genuinely body-less kinds (control flow, wait/listen, unspecified)
+  // remain summary.
+  it.each([
+    [WorkflowTaskKind.transform, "preview"],
+    [WorkflowTaskKind.validate, "preview"],
+    [WorkflowTaskKind.llm_call, "preview"],
+    [WorkflowTaskKind.eval, "preview"],
+    [WorkflowTaskKind.emit_event, "preview"],
+    [WorkflowTaskKind.notification, "preview"],
+    [WorkflowTaskKind.agent_call, "preview"],
+    [WorkflowTaskKind.set_vars, "preview"],
+    [WorkflowTaskKind.human_input, "preview"],
+    [WorkflowTaskKind.http_call, "preview"],
+    [WorkflowTaskKind.grpc_call, "preview"],
+    [WorkflowTaskKind.activity_call, "preview"],
+    [WorkflowTaskKind.run_workflow, "preview"],
+    [WorkflowTaskKind.wait, "summary"],
+    [WorkflowTaskKind.switch_case, "summary"],
+    [WorkflowTaskKind.fork, "summary"],
+    [WorkflowTaskKind.try_catch, "summary"],
+    [WorkflowTaskKind.for_each, "summary"],
+    [WorkflowTaskKind.listen, "summary"],
+    [WorkflowTaskKind.raise_error, "summary"],
+    [WorkflowTaskKind.workflow_task_kind_unspecified, "summary"],
+  ] as const)("kind %d defaults to %s", (kind, expected) => {
+    expect(defaultDisclosureForKind(kind)).toBe(expected);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Presenter registry (the platform-builder seam)
+// ---------------------------------------------------------------------------
+
+describe("registerTaskPresenter", () => {
+  it("overrides the kind line and restores the previous presenter on dispose", () => {
+    const dispose = registerTaskPresenter(WorkflowTaskKind.validate, {
+      previewLine: () => "custom check line",
+    });
+    try {
+      const withOverride = resolveTaskPreview(
+        state({ taskKind: WorkflowTaskKind.validate, outputSummary: { valid: true } }),
+      );
+      expect(withOverride.previewLine).toBe("custom check line");
+    } finally {
+      dispose();
+    }
+    const restored = resolveTaskPreview(
+      state({ taskKind: WorkflowTaskKind.validate, outputSummary: { valid: true } }),
+    );
+    expect(restored.previewLine).toBe("valid");
+  });
+
+  it("falls back to the built-in line when the override returns null", () => {
+    const dispose = registerTaskPresenter(WorkflowTaskKind.validate, {
+      previewLine: () => null,
+    });
+    try {
+      const { previewLine } = resolveTaskPreview(
+        state({ taskKind: WorkflowTaskKind.validate, outputSummary: { valid: true } }),
+      );
+      expect(previewLine).toBe("valid");
+    } finally {
+      dispose();
+    }
+  });
+
+  it("never lets an override outrank status lines (correctness surface)", () => {
+    const dispose = registerTaskPresenter(WorkflowTaskKind.validate, {
+      previewLine: () => "should not appear",
+    });
+    try {
+      const { previewLine } = resolveTaskPreview(
+        state({ taskKind: WorkflowTaskKind.validate, status: "waiting_approval" }),
+      );
+      expect(previewLine).toBe("Awaiting approval");
+    } finally {
+      dispose();
+    }
+  });
+
+  it("can override disclosure per kind", () => {
+    const dispose = registerTaskPresenter(WorkflowTaskKind.http_call, {
+      disclosure: () => "preview",
+    });
+    try {
+      const { disclosure } = resolveTaskPreview(
+        state({ taskKind: WorkflowTaskKind.http_call }),
+      );
+      expect(disclosure).toBe("preview");
+    } finally {
+      dispose();
+    }
+  });
+
+  it("restores a PREVIOUS override (not just the default) on dispose", () => {
+    const disposeA = registerTaskPresenter(WorkflowTaskKind.validate, {
+      previewLine: () => "A",
+    });
+    const disposeB = registerTaskPresenter(WorkflowTaskKind.validate, {
+      previewLine: () => "B",
+    });
+    try {
+      expect(
+        resolveTaskPreview(state({ taskKind: WorkflowTaskKind.validate })).previewLine,
+      ).toBe("B");
+      disposeB();
+      expect(
+        resolveTaskPreview(state({ taskKind: WorkflowTaskKind.validate })).previewLine,
+      ).toBe("A");
+    } finally {
+      disposeA();
+    }
+  });
+});

package/src/workflow/__tests__/useActiveTaskName.test.ts

@@ -19,6 +19,10 @@ function makeState(overrides: Partial<DerivedTaskState> = {}): DerivedTaskState
     currentToolName: "",
     messagesCount: 0,
     toolCallsCount: 0,
+    inputSummary: null,
+    outputSummary: null,
+    approvalRequest: null,
+    approvalResolution: null,
     ...overrides,
   };
 }

package/src/workflow/__tests__/useApprovalBoundary.test.ts

@@ -0,0 +1,158 @@
+// Unit tests for the approval-boundary watcher (S9): entering/exiting
+// crossings, the enabled gate (tracking continues while disabled — no
+// replayed crossings on enable), first-observation semantics, and callback
+// identity churn safety.
+
+import { describe, expect, it, vi } from "vitest";
+import { renderHook } from "@testing-library/react";
+import { useApprovalBoundary } from "../useApprovalBoundary";
+import type { DerivedTaskState } from "../../internal/store/workflow-execution-event-store";
+import { WorkflowTaskKind } from "@stigmer/protos/ai/stigmer/agentic/workflow/v1/enum_pb";
+
+function makeState(overrides: Partial<DerivedTaskState> = {}): DerivedTaskState {
+  return {
+    taskName: "test_task",
+    taskKind: WorkflowTaskKind.workflow_task_kind_unspecified,
+    status: "pending",
+    durationMs: 0,
+    costMicros: BigInt(0),
+    tokensUsed: BigInt(0),
+    attemptNumber: 1,
+    error: "",
+    childExecutionId: "",
+    agentSlug: "",
+    currentToolName: "",
+    messagesCount: 0,
+    toolCallsCount: 0,
+    inputSummary: null,
+    outputSummary: null,
+    approvalRequest: null,
+    approvalResolution: null,
+    ...overrides,
+  };
+}
+
+function statesOf(
+  ...entries: Array<[string, DerivedTaskState["status"]]>
+): ReadonlyMap<string, DerivedTaskState> {
+  return new Map(
+    entries.map(([taskName, status]) => [
+      taskName,
+      makeState({ taskName, status }),
+    ]),
+  );
+}
+
+describe("useApprovalBoundary", () => {
+  it("reports a task entering waiting_approval", () => {
+    const onCrossing = vi.fn();
+    const { rerender } = renderHook(
+      ({ states }) => useApprovalBoundary(states, true, onCrossing),
+      { initialProps: { states: statesOf(["gate", "running"]) } },
+    );
+    expect(onCrossing).not.toHaveBeenCalled();
+
+    rerender({ states: statesOf(["gate", "waiting_approval"]) });
+    expect(onCrossing).toHaveBeenCalledTimes(1);
+    expect(onCrossing).toHaveBeenCalledWith({ entered: ["gate"], exited: [] });
+  });
+
+  it("reports a task exiting waiting_approval", () => {
+    const onCrossing = vi.fn();
+    const { rerender } = renderHook(
+      ({ states }) => useApprovalBoundary(states, true, onCrossing),
+      { initialProps: { states: statesOf(["gate", "waiting_approval"]) } },
+    );
+    onCrossing.mockClear();
+
+    rerender({ states: statesOf(["gate", "running"]) });
+    expect(onCrossing).toHaveBeenCalledWith({ entered: [], exited: ["gate"] });
+  });
+
+  it("counts a first observation already waiting as entered (coalesced commits)", () => {
+    const onCrossing = vi.fn();
+    renderHook(() =>
+      useApprovalBoundary(statesOf(["gate", "waiting_approval"]), true, onCrossing),
+    );
+    expect(onCrossing).toHaveBeenCalledWith({ entered: ["gate"], exited: [] });
+  });
+
+  it("stays silent for non-boundary transitions", () => {
+    const onCrossing = vi.fn();
+    const { rerender } = renderHook(
+      ({ states }) => useApprovalBoundary(states, true, onCrossing),
+      { initialProps: { states: statesOf(["a", "running"]) } },
+    );
+
+    rerender({ states: statesOf(["a", "completed"], ["b", "running"]) });
+    expect(onCrossing).not.toHaveBeenCalled();
+  });
+
+  it("reports concurrent crossings in map (first-started) order", () => {
+    const onCrossing = vi.fn();
+    const { rerender } = renderHook(
+      ({ states }) => useApprovalBoundary(states, true, onCrossing),
+      {
+        initialProps: {
+          states: statesOf(
+            ["a", "waiting_approval"],
+            ["b", "running"],
+            ["c", "running"],
+          ),
+        },
+      },
+    );
+    onCrossing.mockClear();
+
+    rerender({
+      states: statesOf(
+        ["a", "running"],
+        ["b", "waiting_approval"],
+        ["c", "waiting_approval"],
+      ),
+    });
+    expect(onCrossing).toHaveBeenCalledWith({
+      entered: ["b", "c"],
+      exited: ["a"],
+    });
+  });
+
+  it("never fires while disabled, but keeps tracking — enabling later replays nothing", () => {
+    const onCrossing = vi.fn();
+    const { rerender } = renderHook(
+      ({ states, enabled }) => useApprovalBoundary(states, enabled, onCrossing),
+      {
+        initialProps: {
+          states: statesOf(["gate", "running"]),
+          enabled: false,
+        },
+      },
+    );
+
+    // The crossing happens while disabled (e.g. terminal-history replay).
+    rerender({ states: statesOf(["gate", "waiting_approval"]), enabled: false });
+    expect(onCrossing).not.toHaveBeenCalled();
+
+    // Enabling afterwards must not replay the stale crossing…
+    rerender({ states: statesOf(["gate", "waiting_approval"]), enabled: true });
+    expect(onCrossing).not.toHaveBeenCalled();
+
+    // …while a NEW crossing after enablement fires normally.
+    rerender({ states: statesOf(["gate", "running"]), enabled: true });
+    expect(onCrossing).toHaveBeenCalledWith({ entered: [], exited: ["gate"] });
+  });
+
+  it("a fresh callback identity alone never re-fires the last crossing", () => {
+    const first = vi.fn();
+    const second = vi.fn();
+    const states = statesOf(["gate", "waiting_approval"]);
+    const { rerender } = renderHook(
+      ({ cb }) => useApprovalBoundary(states, true, cb),
+      { initialProps: { cb: first } },
+    );
+    expect(first).toHaveBeenCalledTimes(1);
+
+    rerender({ cb: second });
+    expect(second).not.toHaveBeenCalled();
+  });
+});

package/src/workflow/__tests__/useExecutionAnnouncements.test.ts

@@ -19,6 +19,10 @@ function makeState(overrides: Partial<DerivedTaskState> = {}): DerivedTaskState
     currentToolName: "",
     messagesCount: 0,
     toolCallsCount: 0,
+    inputSummary: null,
+    outputSummary: null,
+    approvalRequest: null,
+    approvalResolution: null,
     ...overrides,
   };
 }