orchestrator/agent-team/tests/test_graph.py

"""Unit tests for agent_team.graph (Plane-2 P1 LangGraph wiring; §3.3, §7.1).

These exercise the P1 skeleton + human gate wiring:

* the pure node functions (intake/clarify-author/plan) in isolation,
* graph assembly + edge topology,
* the suspend-on-interrupt / resume-with-Command mechanic end to end,
* the thread_id-keyed driver seam (start/resume/get_state/pending_question),
* that the foundation contracts (PipelineState / Phase / TaskStatus /
  QuestionSet) are imported verbatim and round-trip through the wiring.

An in-memory checkpointer is injected (the SQLite checkpointer is the
production store, D9, not constructed in pre-deploy scaffolding).
"""

from __future__ import annotations

import pytest

# InMemorySaver is the modern name; fall back to MemorySaver on older langgraph.
try:  # pragma: no cover - import shim
    from langgraph.checkpoint.memory import InMemorySaver as _Saver
except ImportError:  # pragma: no cover - import shim
    from langgraph.checkpoint.memory import MemorySaver as _Saver

from agent_team import graph as graph_mod
from agent_team.graph import (
    CLARIFY,
    INTAKE,
    P1_PHASE_SEQUENCE,
    PLAN,
    build_checkpoint_serde,
    build_graph,
    build_sqlite_checkpointer,
    clarify_node,
    get_pipeline_state,
    intake_node,
    pending_question,
    plan_node,
    plan_phase,
    resume_task,
    start_task,
    thread_config,
)
from agent_team.task_model import Phase, PipelineState, TaskStatus
from agent_team.transport import QuestionSet


@pytest.fixture()
def compiled():
    """A graph compiled with a fresh in-memory checkpointer per test."""
    return build_graph(checkpointer=_Saver())


# --- Module surface / constants. -------------------------------------------


def test_node_name_constants_are_distinct() -> None:
    assert len({INTAKE, CLARIFY, PLAN}) == 3


def test_p1_phase_sequence_stops_at_plan() -> None:
    # P1 ends at an approved plan — no BUILD/VERIFY in the wired sequence (§7.1).
    assert P1_PHASE_SEQUENCE == (Phase.INTAKE, Phase.CLARIFY, Phase.PLAN)
    assert Phase.BUILD not in P1_PHASE_SEQUENCE
    assert Phase.VERIFY not in P1_PHASE_SEQUENCE


# --- Pure node behaviour. ---------------------------------------------------


def test_intake_node_activates_and_advances_to_clarify() -> None:
    out = intake_node(PipelineState(thread_id="t", current_phase=Phase.INTAKE.value))
    assert out["status"] == TaskStatus.ACTIVE.value
    assert out["current_phase"] == Phase.CLARIFY.value
    assert out["updated_at"]


def test_plan_node_lands_approved_plan_and_finishes() -> None:
    out = plan_node(PipelineState(thread_id="t", qa_history=[{"answer": "x"}]))
    assert out["status"] == TaskStatus.DONE.value
    assert out["current_phase"] == Phase.DONE.value
    assert out["plan"]["approved"] is True


def test_plan_phase_counts_qa_turns() -> None:
    state = PipelineState(qa_history=[{"answer": "a"}, {"answer": "b"}])
    plan = plan_phase(state)
    assert plan["qa_turns"] == 2
    assert plan["approved"] is True


def test_plan_phase_handles_empty_history() -> None:
    assert plan_phase(PipelineState())["qa_turns"] == 0


def test_clarify_node_suspends_rather_than_falling_through() -> None:
    # Called bare (no running graph), interrupt() refuses to return a value:
    # it raises because there is no runnable context to suspend into. This
    # confirms clarify_node genuinely suspends rather than falling through to
    # its post-interrupt return.
    with pytest.raises(RuntimeError):
        clarify_node(PipelineState(thread_id="t", transport="slack"))


# --- Graph assembly. --------------------------------------------------------


def test_build_graph_without_checkpointer_compiles() -> None:
    # An uncheckpointed graph still compiles (used only for straight-through
    # smoke paths); the driver requires a checkpointer for suspend/resume.
    assert build_graph() is not None


def test_build_graph_with_checkpointer_compiles(compiled) -> None:
    assert compiled is not None


def test_graph_nodes_present(compiled) -> None:
    nodes = set(compiled.get_graph().nodes)
    assert {INTAKE, CLARIFY, PLAN} <= nodes


# --- Suspend / resume end to end. ------------------------------------------


def test_start_task_suspends_on_human_gate(compiled) -> None:
    thread_id, state = start_task(compiled, transport="slack")
    # The task ran INTAKE then suspended at CLARIFY's interrupt().
    assert "__interrupt__" in state
    payload = pending_question(compiled, thread_id=thread_id)
    assert payload is not None
    assert payload["thread_id"] == thread_id
    assert payload["transport"] == "slack"
    assert payload["turn"] == 0
    assert payload["deadline"]


def test_start_task_seeds_task_description_into_state(compiled) -> None:
    # Regression: the intake description (Slack /new-task text, GitHub issue body)
    # must reach the graph state so the clarifier can reason about it. It is
    # seeded into the initial invoke and must persist through INTAKE into the
    # suspended CLARIFY snapshot (intake_node returns only a partial state).
    _thread_id, state = start_task(
        compiled, transport="slack", task="build a login form"
    )
    assert state.get("task") == "build a login form"


def test_start_task_seeds_slack_thread_ts_into_state(compiled) -> None:
    # One-thread-per-task: the root "📥 Task received" message ts must reach the
    # graph state so every later question/notification threads under it. Like
    # ``task`` it is seeded into the initial invoke and persists through INTAKE
    # into the suspended CLARIFY snapshot.
    _thread_id, state = start_task(
        compiled, transport="slack", slack_thread_ts="1700000000.ROOT"
    )
    assert state.get("slack_thread_ts") == "1700000000.ROOT"
    # It is also surfaced on the pending interrupt payload so the responder can
    # thread the question post.
    payload = pending_question(compiled, thread_id=_thread_id)
    assert payload["slack_thread_ts"] == "1700000000.ROOT"


def test_start_task_default_slack_thread_ts_is_empty(compiled) -> None:
    # A task with no root post (e.g. a GitHub-issue origin): slack_thread_ts is
    # empty so questions post top-level exactly as before.
    _thread_id, state = start_task(compiled, transport="slack")
    assert state.get("slack_thread_ts") == ""


def test_pending_question_carries_foundation_questionset(compiled) -> None:
    thread_id, _ = start_task(compiled, transport="slack")
    payload = pending_question(compiled, thread_id=thread_id)
    qset = payload["question_set"]
    # Verbatim foundation contract — not a redefinition.
    assert isinstance(qset, QuestionSet)
    assert qset.thread_id == thread_id
    assert qset.question_id == payload["question_id"]
    assert qset.turn == 0
    assert qset.questions  # non-empty question-set


def test_resume_drives_task_to_done(compiled) -> None:
    thread_id, _ = start_task(compiled, transport="slack")
    final = resume_task(compiled, thread_id=thread_id, answer={"text": "do the thing"})
    assert final["status"] == TaskStatus.DONE.value
    assert final["current_phase"] == Phase.DONE.value
    assert final["plan"]["approved"] is True


def test_answer_is_recorded_in_qa_history(compiled) -> None:
    thread_id, _ = start_task(compiled, transport="slack")
    answer = {"text": "scope is X"}
    final = resume_task(compiled, thread_id=thread_id, answer=answer)
    assert len(final["qa_history"]) == 1
    assert final["qa_history"][0]["answer"] == answer
    assert final["qa_history"][0]["turn"] == 0


def test_question_id_is_stable_across_resume(compiled) -> None:
    # The clarifier node re-executes on resume; the question_id must NOT change
    # between the id delivered at suspend (the ledger key) and the one recorded
    # in qa_history, or the §3.3.1 identity contract breaks.
    thread_id, _ = start_task(compiled, transport="slack")
    delivered = pending_question(compiled, thread_id=thread_id)["question_id"]
    final = resume_task(compiled, thread_id=thread_id, answer="ok")
    assert final["qa_history"][0]["question_id"] == delivered


def test_no_pending_question_after_completion(compiled) -> None:
    thread_id, _ = start_task(compiled, transport="slack")
    resume_task(compiled, thread_id=thread_id, answer="ok")
    assert pending_question(compiled, thread_id=thread_id) is None


def test_get_pipeline_state_reflects_suspend_then_done(compiled) -> None:
    thread_id, _ = start_task(compiled, transport="slack")
    mid = get_pipeline_state(compiled, thread_id=thread_id)
    # Suspended ON the clarifier gate: INTAKE already advanced the phase to
    # CLARIFY, and the clarifier's post-interrupt write (-> PLAN) has NOT yet
    # committed because the node is paused at interrupt(). Task is mid-flight.
    assert mid["current_phase"] == Phase.CLARIFY.value
    assert mid["status"] == TaskStatus.ACTIVE.value
    resume_task(compiled, thread_id=thread_id, answer="ok")
    done = get_pipeline_state(compiled, thread_id=thread_id)
    assert done["status"] == TaskStatus.DONE.value
    assert done["current_phase"] == Phase.DONE.value


# --- Thread isolation (§3.3.1 P1 exit criterion (d)). ----------------------


def test_two_tasks_suspend_and_resume_independently(compiled) -> None:
    t1, _ = start_task(compiled, transport="slack")
    t2, _ = start_task(compiled, transport="github")

    assert t1 != t2
    p1 = pending_question(compiled, thread_id=t1)
    p2 = pending_question(compiled, thread_id=t2)
    assert p1["transport"] == "slack"
    assert p2["transport"] == "github"
    assert p1["question_id"] != p2["question_id"]

    # Resume only t1; t2 must remain suspended on its own gate.
    f1 = resume_task(compiled, thread_id=t1, answer="answer-1")
    assert f1["status"] == TaskStatus.DONE.value
    assert pending_question(compiled, thread_id=t2) is not None

    f2 = resume_task(compiled, thread_id=t2, answer="answer-2")
    assert f2["status"] == TaskStatus.DONE.value
    assert f2["qa_history"][0]["answer"] == "answer-2"


def test_explicit_thread_id_is_honoured(compiled) -> None:
    tid, _ = start_task(compiled, thread_id="fixed-thread", transport="slack")
    assert tid == "fixed-thread"
    assert pending_question(compiled, thread_id="fixed-thread") is not None


# --- Durable resume across a fresh graph object (P1 exit criterion (a)). ----


def test_resume_works_on_a_new_graph_over_shared_checkpointer() -> None:
    # Simulates a process restart: a NEW compiled graph object built over the
    # SAME checkpointer must resume a task suspended by the first graph object.
    saver = _Saver()
    g1 = build_graph(checkpointer=saver)
    thread_id, _ = start_task(g1, transport="slack")

    g2 = build_graph(checkpointer=saver)  # "after restart"
    assert pending_question(g2, thread_id=thread_id) is not None
    final = resume_task(g2, thread_id=thread_id, answer="post-restart")
    assert final["status"] == TaskStatus.DONE.value
    assert final["qa_history"][0]["answer"] == "post-restart"


# --- Driver-seam helpers. ---------------------------------------------------


def test_thread_config_shape() -> None:
    assert thread_config("abc") == {"configurable": {"thread_id": "abc"}}


def test_start_task_mints_unique_thread_ids(compiled) -> None:
    t1, _ = start_task(compiled, transport="slack")
    t2, _ = start_task(compiled, transport="slack")
    assert t1 != t2


# --- Production checkpointer factory. --------------------------------------


def test_build_sqlite_checkpointer_missing_dep_raises_runtimeerror(
    monkeypatch, tmp_path
) -> None:
    # When the optional langgraph-checkpoint-sqlite package is absent, the
    # factory must fail loudly with a clear RuntimeError, never silently run
    # uncheckpointed. Force the ImportError path deterministically.
    import builtins

    real_import = builtins.__import__

    def _blocking_import(name, *args, **kwargs):
        if name == "langgraph.checkpoint.sqlite":
            raise ImportError("blocked for test")
        return real_import(name, *args, **kwargs)

    monkeypatch.setattr(builtins, "__import__", _blocking_import)
    with pytest.raises(RuntimeError, match="SQLite checkpointer"):
        build_sqlite_checkpointer(tmp_path / "state.db")


def test_build_sqlite_checkpointer_builds_when_dep_present(tmp_path) -> None:
    # If the optional package IS installed, the factory returns a checkpointer
    # over the DB path. Skip cleanly where it's absent (pre-deploy scaffolding).
    pytest.importorskip("langgraph.checkpoint.sqlite")
    cm = build_sqlite_checkpointer(tmp_path / "nested" / "state.db")
    assert cm is not None
    assert (tmp_path / "nested").is_dir()
    # Contract: the factory returns a CONTEXT MANAGER (SqliteSaver.from_conn_string
    # is a @contextmanager), so callers must enter it. Pin that here so a future
    # change that returns a bare/un-entered object is caught (review FIX).
    assert hasattr(cm, "__enter__") and hasattr(cm, "__exit__")
    with cm as saver:
        # The entered object is the real saver the graph compiles against.
        assert hasattr(saver, "get_next_version")


# --- Checkpoint serializer / QuestionSet msgpack allowlist (D9). ------------
# QuestionSet rides in the clarifier interrupt payload and so is msgpack-encoded
# into every checkpoint. The default serializer deserializes it but logs
# "Deserializing unregistered type agent_team.transport.base.QuestionSet ... will
# be blocked in a future version" on each load, and the coming LangGraph default
# turns that warning into a hard block (breaking durable resume). These pin that
# build_checkpoint_serde registers the type so it round-trips silently AND is
# already block-clean.

_QSET_MSGPACK_KEY = ("agent_team.transport.base", "QuestionSet")


def _capture_serde_warnings():
    """Attach a capturing handler to the serde logger; return (handler, buffer)."""
    import io
    import logging

    buf = io.StringIO()
    handler = logging.StreamHandler(buf)
    handler.setLevel(logging.WARNING)
    logger = logging.getLogger("langgraph.checkpoint.serde.jsonplus")
    logger.addHandler(handler)
    logger.setLevel(logging.WARNING)
    return logger, handler, buf


def _reset_serde_warning_dedup() -> None:
    """Clear the serializer's process-wide warn-once dedup set.

    jsonplus dedups "unregistered type" warnings across the process lifetime, so
    an earlier test (or the assertion below) could mask a regression. Clearing the
    set makes each assertion observe the live behavior, not a stale dedup.
    """
    from langgraph.checkpoint.serde import jsonplus as _jp

    _jp._warned_unregistered_types.clear()
    _jp._warned_blocked_types.clear()


def test_build_checkpoint_serde_allowlists_questionset() -> None:
    # The serde must carry QuestionSet on its msgpack allowlist (an explicit
    # collection, NOT the permissive ``True`` sentinel) — that is the registration
    # path the warning recommends and what makes resume forward-compatible.
    pytest.importorskip("langgraph.checkpoint.serde.jsonplus")
    serde = build_checkpoint_serde()
    allowed = serde._allowed_msgpack_modules
    assert allowed is not True  # not the warn-on-everything permissive default
    assert _QSET_MSGPACK_KEY in allowed


def test_questionset_round_trips_through_serde_without_warning() -> None:
    # The configured serializer must round-trip a QuestionSet AND emit no
    # "unregistered type" warning on deserialize.
    pytest.importorskip("langgraph.checkpoint.serde.jsonplus")
    _reset_serde_warning_dedup()
    logger, handler, buf = _capture_serde_warnings()
    try:
        serde = build_checkpoint_serde()
        qset = QuestionSet(
            thread_id="t-1",
            question_id="q-1",
            turn=0,
            questions=["What is in scope?"],
            context={"phase": Phase.CLARIFY.value},
        )
        encoded = serde.dumps_typed(qset)
        restored = serde.loads_typed(encoded)
    finally:
        logger.removeHandler(handler)

    assert restored == qset
    assert "unregistered type" not in buf.getvalue()


def test_questionset_round_trips_through_sqlite_checkpointer_without_warning(
    tmp_path,
) -> None:
    # End-to-end against the REAL production saver: drive the graph to the
    # clarifier suspend (which checkpoints a QuestionSet), force a checkpoint
    # load, and resume — asserting no "unregistered type" warning prints and the
    # task still completes. This is the runtime regression guard for the warning.
    pytest.importorskip("langgraph.checkpoint.sqlite")
    _reset_serde_warning_dedup()
    logger, handler, buf = _capture_serde_warnings()
    try:
        cm = build_sqlite_checkpointer(tmp_path / "state.db")
        with cm as saver:
            graph = build_graph(saver)
            thread_id, _ = start_task(graph, transport="slack")
            pending = pending_question(graph, thread_id=thread_id)
            assert isinstance(pending["question_set"], QuestionSet)
            # Force a fresh checkpoint deserialize (the warning's trigger point).
            get_pipeline_state(graph, thread_id=thread_id)
            resumed = resume_task(graph, thread_id=thread_id, answer="scope it")
    finally:
        logger.removeHandler(handler)

    assert resumed["status"] == TaskStatus.DONE.value
    assert "unregistered type" not in buf.getvalue()


# --- P2 review-loop wiring. -------------------------------------------------


def _p2_plan_stub(state: PipelineState) -> PipelineState:
    """Stand-in for the real planner: emit a plan and advance to REVIEW.

    Mirrors planner.plan_node's contract (sets ``plan`` + phase REVIEW) without a
    model call, so the P2 graph topology + the review loop can be driven in a
    unit test. The revision index tracks prior review rounds.
    """
    revisions = len(state.get("review_verdicts") or [])
    return PipelineState(
        plan={"phases": ["P1"], "revision": revisions},
        current_phase=Phase.REVIEW.value,
        status=TaskStatus.ACTIVE.value,
    )


def _p2_graph(review_text: str):
    """Compile a P2 graph whose review invoker returns ``review_text``."""
    from agent_team.nodes import review_loop

    review_loop.set_review_invoker(lambda prompt, **kw: review_text)
    return build_graph(
        checkpointer=_Saver(),
        live_plan_node=_p2_plan_stub,
        review_node=review_loop.bind_review_node(),
        route_review=review_loop.route_after_review,
    )


@pytest.fixture
def restore_review_invoker():
    """Save/restore the review-loop module-global invoker around a test."""
    from agent_team.nodes import review_loop

    saved = review_loop._review_invoker
    yield
    review_loop._review_invoker = saved


def test_build_graph_review_node_requires_route() -> None:
    from agent_team.nodes import review_loop

    with pytest.raises(ValueError, match="route_review"):
        build_graph(review_node=review_loop.review_node)


def test_p2_graph_approve_terminates(restore_review_invoker) -> None:
    # clarify(stub) -> plan(stub->REVIEW) -> review(APPROVE) -> END.
    graph = _p2_graph("VERDICT: APPROVE\nlooks solid")
    thread_id, _ = start_task(graph, transport="slack")
    final = resume_task(graph, thread_id=thread_id, answer="scope is X")

    # The review node advanced an APPROVED plan toward BUILD and the graph ended.
    assert final["current_phase"] == Phase.BUILD.value
    assert len(final["review_verdicts"]) == 1


def test_p2_graph_loops_then_escalates_on_persistent_changes(
    restore_review_invoker,
) -> None:
    # A reviewer that never approves loops plan<->review until the round cap,
    # then escalates (parks) rather than spinning. Default cap is 3 rounds.
    graph = _p2_graph("VERDICT: REQUEST CHANGES\nstill not ready")
    thread_id, _ = start_task(graph, transport="slack")
    final = resume_task(graph, thread_id=thread_id, answer="scope is X")

    assert final["current_phase"] == Phase.PARKED.value
    assert final["status"] == TaskStatus.PARKED.value
    assert len(final["review_verdicts"]) == 3  # looped to the cap, then escalated


# --- P3 build -> verify subgraph wiring (opt-in). ---------------------------


def _p3_plan_stub(state: PipelineState) -> PipelineState:
    """P2/P3 planner stub: emit an APPROVED, scoped plan and advance to REVIEW.

    Like ``_p2_plan_stub`` but carries a ``scope`` so the P3 BUILD node's
    trust-control-surface scan accepts the candidate diff, letting the
    build -> verify topology be driven end to end.
    """
    revisions = len(state.get("review_verdicts") or [])
    return PipelineState(
        plan={
            "title": "do it",
            "scope": ["src"],
            "phases": ["P1"],
            "revision": revisions,
        },
        current_phase=Phase.REVIEW.value,
        status=TaskStatus.ACTIVE.value,
    )


def _p3_diff() -> str:
    """A minimal in-scope unified diff the fake builder returns."""
    return "diff --git a/src/foo.py b/src/foo.py\n@@ -1 +1 @@\n-old\n+new\n"


def _p3_graph(review_text: str, *, ci_result_fetcher):
    """Compile a P3 graph: review -> build -> verify with injected seams.

    The diff builder is a fixed in-scope diff; the CI-result fetcher is injected
    so the test drives the verifier verdict (pass / fail / none) deterministically
    with no live CI.
    """
    from agent_team.nodes import review_loop
    from agent_team.nodes.build_verify_subgraph import (
        make_build_node,
        make_verify_node,
        route_after_verify,
    )
    from agent_team.nodes.verifier import VerifierConfig

    review_loop.set_review_invoker(lambda prompt, **kw: review_text)

    def fake_builder(*, plan, config):
        return _p3_diff()

    build_node = make_build_node(diff_builder=fake_builder)
    verify_node = make_verify_node(
        VerifierConfig(expected_run_id="r1", allowed_scope=["src"]),
        ci_result_fetcher=ci_result_fetcher,
    )
    return build_graph(
        checkpointer=_Saver(),
        live_plan_node=_p3_plan_stub,
        review_node=review_loop.bind_review_node(),
        route_review=review_loop.route_after_review,
        build_verify=(build_node, verify_node, route_after_verify),
    )


def test_build_graph_build_verify_requires_review_node() -> None:
    """build_verify without review_node is a wiring error (no 'build' route)."""
    from agent_team.nodes.build_verify_subgraph import (
        make_build_node,
        make_verify_node,
        route_after_verify,
    )
    from agent_team.nodes.verifier import VerifierConfig

    tuple_ = (
        make_build_node(diff_builder=None),
        make_verify_node(VerifierConfig(expected_run_id="")),
        route_after_verify,
    )
    with pytest.raises(ValueError, match="build_verify"):
        build_graph(build_verify=tuple_)


def test_p3_graph_authenticated_pass_routes_to_done(restore_review_invoker) -> None:
    """review(APPROVE) -> build -> verify(PASS via fake CI) -> DONE (PR terminus)."""

    def pass_fetcher(state):
        from agent_team.state_store import compute_content_hash

        diff_hash = compute_content_hash(_p3_diff().encode("utf-8"))
        return {"run_id": "r1", "conclusion": "success", "diff_hash": diff_hash}

    graph = _p3_graph("VERDICT: APPROVE\nlooks solid", ci_result_fetcher=pass_fetcher)
    thread_id, _ = start_task(graph, transport="slack")
    final = resume_task(graph, thread_id=thread_id, answer="scope is X")

    # The authenticated CI pass cleared the gate -> DONE terminus.
    assert final["current_phase"] == Phase.DONE.value
    assert final["status"] == TaskStatus.DONE.value


def test_p3_graph_inert_default_parks_at_verify(restore_review_invoker) -> None:
    """review(APPROVE) -> build -> verify(no CI result) -> BLOCK -> PARKED.

    With the INERT default (no authenticated CI result) the gate can never
    fabricate a pass, so an approved plan still parks at VERIFY. This is the
    production-safe behavior the opt-in subgraph ships with.
    """
    graph = _p3_graph(
        "VERDICT: APPROVE\nlooks solid", ci_result_fetcher=lambda state: None
    )
    thread_id, _ = start_task(graph, transport="slack")
    final = resume_task(graph, thread_id=thread_id, answer="scope is X")

    assert final["current_phase"] == Phase.PARKED.value
    assert final["status"] == TaskStatus.PARKED.value


def test_p3_graph_route_constants_mirror_subgraph_by_value() -> None:
    """graph.py's P3 route ids match the subgraph module by value (no cycle)."""
    from agent_team.nodes import build_verify_subgraph as bvs

    assert graph_mod.APPROVED_ROUTE == bvs.APPROVED_ROUTE
    assert graph_mod.BUILD_ROUTE == bvs.BUILD_ROUTE
    assert graph_mod.PARKED_ROUTE == bvs.PARKED_ROUTE


# --- Module import hygiene. -------------------------------------------------


def test_module_imports_without_optional_sqlite_dep() -> None:
    # The module-level import of graph must not pull in the optional SQLite
    # checkpointer (that import is deferred into build_sqlite_checkpointer).
    assert hasattr(graph_mod, "build_graph")
    assert hasattr(graph_mod, "build_sqlite_checkpointer")