multi-agent-mux/.agents/skills/lib_py/loop/policy.py

"""The completion decision. A pure function, deliberately.

No I/O, no clock, no globals: that is what lets the phase-0 scenario fixtures be
fed to it verbatim as test vectors. Those seven scenarios and their
`expected_policy.decision` survived three-reviewer scrutiny, so they are the
contract this implements rather than a fresh set of rules.

Fail-closed throughout: every gate must pass for `candidate_approved`. Anything
unproven holds.
"""
from __future__ import annotations

from lib_py.loop import attribution
from lib_py.loop.evidence import Decision

# Native states that mean "not doing anything right now". Idle alone proves
# nothing about completion; it only makes completion *possible* to consider.
QUIESCENT = ("idle", "done")

# The four identifiers an external collector must supply before a completion can
# be attributed to this attempt.
ATTRIBUTION_FIELDS = ("job_id", "attempt_id", "agent_session_uuid",
                      "pane_process_generation")


def _timeline_by_kind(scenario):
    out = {}
    for item in scenario.get("timeline") or []:
        out.setdefault(item.get("source_kind"), []).append(item)
    return out


def collect_observed_identity(scenario) -> dict:
    """Gather identifiers the *evidence* carries, for comparison with dispatch.

    Two sources, and the distinction matters. Association of a job or attempt to
    an observation is the external collector's business, so it lives in a
    ``collector_attribution`` envelope beside the payload -- never inside it,
    because a native Herdr payload must stay native (phase 0 forbids MAM fields
    there, and a fabricated one would be indistinguishable from a real reading).
    The report manifest is a filesystem fact and stays in that payload.

    Every occurrence is kept. Taking only the first would drop a later,
    contradicting observation, which is precisely what this exists to catch.
    """
    observed = {}
    for item in scenario.get("timeline") or []:
        envelope = item.get("collector_attribution") or {}
        for field in ATTRIBUTION_FIELDS:
            if envelope.get(field) is not None:
                observed.setdefault(field, []).append(envelope[field])
        payload = item.get("payload") or {}
        if item.get("source_kind") != "herdr":
            # A MAM-side source (an event, a filesystem fact) legitimately names
            # the job or attempt in its own payload. A native Herdr payload never
            # may: phase 0 forbids MAM fields there, so an identifier appearing
            # inside one is either fabricated or a contract violation, and either
            # way it is not evidence.
            for field in ATTRIBUTION_FIELDS:
                if payload.get(field) is not None:
                    observed.setdefault(field, []).append(payload[field])
        manifest = payload.get("report_manifest")
        if isinstance(manifest, dict) and manifest.get("attempt_id") is not None:
            observed.setdefault("manifest_attempt_id", []).append(manifest["attempt_id"])
    return observed


def decide(scenario) -> Decision:
    """Map one composite scenario to a decision.

    Order matters: transport doubt and staleness invalidate the evidence itself,
    so they are answered before anything is inferred from it.
    """
    dispatch = scenario.get("dispatch_context") or {}
    by_kind = _timeline_by_kind(scenario)
    herdr_items = by_kind.get("herdr") or []
    latest = herdr_items[-1]["payload"] if herdr_items else {}
    filesystem = scenario.get("filesystem") or {}

    # 1. Transport doubt: evidence gathered before a gap cannot be trusted.
    for item in herdr_items:
        payload = item.get("payload") or {}
        if payload.get("transport_healthy") is False or payload.get("error"):
            return Decision("hold_reconnect",
                            "herdr transport failed; cached observations are void")

    # 2. A signal from a previous attempt or process generation decides nothing.
    for item in herdr_items:
        payload = item.get("payload") or {}
        pane_generation = payload.get("pane_process_generation")
        if pane_generation and dispatch.get("pane_process_generation") and \
                pane_generation != dispatch["pane_process_generation"]:
            return Decision("reject_stale",
                            "signal belongs to a different process generation")
        pane_id = payload.get("pane_id")
        if pane_id and dispatch.get("pane_process_generation") and \
                not attribution.pane_matches_generation(
                    pane_id, dispatch["pane_process_generation"]):
            return Decision("reject_stale", "signal belongs to a different pane")

    # 3. A blocked agent is waiting for a human. Completion is forbidden.
    if latest.get("agent_status") == "blocked" or latest.get("matched_rule"):
        return Decision("hold_blocked", "agent is blocked on an interactive prompt")

    # 4. Attribution: the dispatch record must be well formed AND agree with the
    #    identifiers the evidence itself carries. Flags on the dispatch record
    #    alone are self-assertion, not correspondence.
    observed = collect_observed_identity(scenario)
    manifest_attempts = observed.pop("manifest_attempt_id", [])

    attributed = attribution.check(dispatch)
    if not attributed.ok:
        return Decision("hold", attributed.reason)

    # Absence of proof is not proof. Comparing only the identifiers that happen
    # to be present would let evidence carrying none at all be approved on the
    # dispatch record's own say-so, which is the self-assertion this check exists
    # to replace. Every identifier must be supplied, then must agree.
    for field in ATTRIBUTION_FIELDS:
        if not observed.get(field):
            return Decision("hold",
                            "no observed {} to attribute this completion to".format(field))
        for value in observed[field]:
            if value != dispatch.get(field):
                return Decision("hold", "attribution mismatch on {}: observed {!r}, "
                                        "dispatched {!r}".format(field, value,
                                                                 dispatch.get(field)))

    # The finalized report must belong to the attempt being judged. A missing
    # manifest attempt holds for the same reason: it proves nothing.
    if not manifest_attempts:
        return Decision("hold", "no report manifest attempt to associate with this attempt")
    for manifest_attempt in manifest_attempts:
        if manifest_attempt != dispatch.get("attempt_id"):
            return Decision("hold",
                            "report manifest belongs to attempt {!r}, not {!r}".format(
                                manifest_attempt, dispatch.get("attempt_id")))

    # 5. Still working, or a tool is running underneath a ready prompt.
    if latest.get("agent_status") not in QUIESCENT:
        return Decision("hold", "agent is not quiescent (status={!r})".format(
            latest.get("agent_status")), attribution_ok=True)
    for item in by_kind.get("harness") or []:
        if (item.get("payload") or {}).get("active"):
            return Decision("hold", "a tool is still executing", attribution_ok=True)

    # 6. The report must be finalized, not merely paused. A debounce alone is a
    #    candidate filter, never an authority to complete.
    if not filesystem.get("exists"):
        return Decision("hold", "no report present", attribution_ok=True)
    if not filesystem.get("manifest_committed"):
        return Decision("hold", "report is not finalized (no commit manifest)",
                        attribution_ok=True)
    if not filesystem.get("hash_verified"):
        return Decision("hold", "report manifest hash unverified", attribution_ok=True)

    return Decision("candidate_approved",
                    "attribution, quiescence and a verified manifest all hold",
                    attribution_ok=True)