teleo-infrastructure/tests/test_apply_worker.py
Fawaz 4de98b29fd
Some checks are pending
CI / lint-and-test (push) Waiting to run
feat(kb): apply-worker — auto-land approved proposals (stage 2 automation) (#36)
* feat(kb): apply-worker to auto-land approved proposals (stage 2 automation)

Event-driven worker that turns a HUMAN-approved kb_stage proposal into canonical
state, so an approval in Telegram surfaces in Leo's identity without Leo applying
its own work.

- Fires only on status='approved' (never pending_review) -> proposer != applier
  holds; the human approval stays the trigger. No auto-approve anywhere.
- Reuses scripts/apply_proposal.py verbatim as the sole apply path (same txn,
  rowcount=1 guard, FK stamp). Connects as the narrow kb_apply role, never
  superuser, never inside the hermes harness.
- Render hook (--render-cmd / KB_APPLY_RENDER_CMD) is inert until the SOUL
  renderer (PR2) is deployed; applying still works, rendered SOUL just lags.
- Ships INERT: report-only unless --enable / KB_APPLY_WORKER_ENABLED=1. systemd
  oneshot service + 5min timer, both shipped disabled.
- 10 unit tests; candidate query validated read-only vs prod (0 applyable today).

* fix(kb): apply-worker --max-per-tick cap + poison-pill retry ceiling

Fixer draft-exit items:
- --max-per-tick=1 (default): an enabled worker lands applies one-at-a-time
  and observably instead of draining the whole approved queue in one tick.
- --max-attempts=3 ceiling with a persisted failure-count state file: a
  deterministically-failing approved proposal is treated as a poison pill and
  skipped after N consecutive failures, instead of retrying every tick forever.
  State persists on disk because the worker runs oneshot per timer tick.

Both are inert until the worker is enabled; it still ships disabled.
2026-07-05 18:23:56 -04:00

199 lines
6.7 KiB
Python

"""Unit tests for scripts/apply_worker.py.
Pure tests: query building, candidate parsing, render-hook expansion, and the
report-only enablement gate -- no live database. Runnable under pytest or
standalone (`python3 tests/test_apply_worker.py`).
"""
import argparse
from pathlib import Path
import sys
try:
import pytest
except ImportError: # standalone fallback so the file runs without pytest installed
import contextlib
import types
pytest = types.SimpleNamespace()
@contextlib.contextmanager
def _raises(exc):
try:
yield
except exc:
return
raise AssertionError(f"expected {exc.__name__} to be raised")
pytest.raises = _raises
REPO_ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(REPO_ROOT / "scripts"))
import apply_worker as w # noqa: E402
# --- candidate query ------------------------------------------------------ #
def test_candidate_query_filters_on_approved():
sql = w.build_candidate_query()
# The structural guarantee: worker can only ever act on approved proposals.
assert "status = 'approved'" in sql
# Never selects pending_review (no auto-approve possible).
assert "pending_review" not in sql
def test_candidate_query_requires_apply_payload():
sql = w.build_candidate_query()
assert "payload ? 'apply_payload'" in sql
def test_candidate_query_lists_worker_types_only():
sql = w.build_candidate_query()
for t in w.WORKER_TYPES:
assert f"'{t}'" in sql
# a non-applyable type must not appear
assert "reject_claim" not in sql
# --- candidate parsing ---------------------------------------------------- #
def test_parse_candidates_skips_blank_lines():
out = '{"id": "a", "proposal_type": "add_edge", "agent_id": null}\n\n'
rows = w.parse_candidates(out)
assert len(rows) == 1
assert rows[0]["id"] == "a"
# --- render hook ---------------------------------------------------------- #
def test_render_command_none_when_unset():
assert w.build_render_command("", "agent-1") is None
def test_render_command_none_without_agent():
assert w.build_render_command("python3 render_soul.py --agent-id {agent_id}", None) is None
def test_render_command_expands_agent_id():
cmd = w.build_render_command("python3 render_soul.py --agent-id {agent_id}", "abc")
assert cmd == ["python3", "render_soul.py", "--agent-id", "abc"]
# --- enablement gate ------------------------------------------------------ #
def _args(**over):
base = dict(
enable=False, limit=20, max_per_tick=1, max_attempts=3,
failure_state_file=None, applied_by="kb-apply",
apply_script="apply_proposal.py", render_cmd="",
secrets_file="x", container="teleo-pg", db="teleo",
host="127.0.0.1", role="kb_apply",
)
base.update(over)
return argparse.Namespace(**base)
def test_report_only_gate_does_not_apply(monkeypatch=None):
calls = []
w.ap.load_password = lambda _f: "pw"
w.fetch_candidates = lambda a, p: [{"id": "p1", "proposal_type": "revise_strategy", "agent_id": "ag"}]
w.apply_one = lambda a, pid: calls.append(pid)
rc = w.run(_args(enable=False))
assert rc == 0
assert calls == [] # disabled worker performs NO writes
def test_enabled_worker_applies_and_renders():
applied, rendered = [], []
w.ap.load_password = lambda _f: "pw"
w.fetch_candidates = lambda a, p: [{"id": "p1", "proposal_type": "revise_strategy", "agent_id": "ag"}]
w.apply_one = lambda a, pid: applied.append(pid)
w.render_one = lambda a, agent_id: rendered.append(agent_id) or "ok"
rc = w.run(_args(enable=True))
assert rc == 0
assert applied == ["p1"]
assert rendered == ["ag"] # revise_strategy triggers a render
def test_enabled_worker_skips_render_for_add_edge():
applied, rendered = [], []
w.ap.load_password = lambda _f: "pw"
w.fetch_candidates = lambda a, p: [{"id": "e1", "proposal_type": "add_edge", "agent_id": None}]
w.apply_one = lambda a, pid: applied.append(pid)
w.render_one = lambda a, agent_id: rendered.append(agent_id) or "ok"
rc = w.run(_args(enable=True))
assert rc == 0
assert applied == ["e1"]
assert rendered == [] # add_edge changes the graph, not the rendered self
# --- per-tick cap + poison-pill ceiling ----------------------------------- #
def _cands(*ids):
return [{"id": i, "proposal_type": "add_edge", "agent_id": None} for i in ids]
def test_partition_caps_to_max_per_tick():
split = w.partition_candidates(_cands("a", "b", "c"), {}, max_attempts=3, max_per_tick=1)
assert [c["id"] for c in split["to_apply"]] == ["a"] # only 1 applied per tick
assert split["poisoned"] == []
def test_partition_skips_poison_pill_over_ceiling():
split = w.partition_candidates(_cands("a", "b"), {"a": 3}, max_attempts=3, max_per_tick=5)
assert [c["id"] for c in split["poisoned"]] == ["a"] # a is poisoned, never retried
assert [c["id"] for c in split["to_apply"]] == ["b"] # b still eligible
def test_enabled_worker_caps_applies_at_one_per_tick():
applied = []
w.ap.load_password = lambda _f: "pw"
w.fetch_candidates = lambda a, p: _cands("e1", "e2", "e3")
w.apply_one = lambda a, pid: applied.append(pid)
rc = w.run(_args(enable=True, max_per_tick=1))
assert rc == 0
assert applied == ["e1"] # capped to 1 even with 3 approved candidates
def test_failure_state_roundtrip_and_ceiling(tmp_path=None):
import tempfile
d = tempfile.mkdtemp()
path = str(Path(d) / "failures.json")
w.save_failure_state(path, {"p1": 2})
assert w.load_failure_state(path) == {"p1": 2}
# p1 at 2 with max_attempts 3 is still eligible; at 3 it poisons
split = w.partition_candidates(_cands("p1"), {"p1": 3}, max_attempts=3, max_per_tick=1)
assert [c["id"] for c in split["poisoned"]] == ["p1"]
def test_apply_failure_increments_persisted_count():
import tempfile
path = str(Path(tempfile.mkdtemp()) / "failures.json")
w.ap.load_password = lambda _f: "pw"
w.fetch_candidates = lambda a, p: _cands("bad")
def _boom(a, pid):
raise RuntimeError("deterministic failure")
w.apply_one = _boom
rc = w.run(_args(enable=True, failure_state_file=path, max_per_tick=1))
assert rc == 1
assert w.load_failure_state(path).get("bad") == 1 # count persisted for next tick
if __name__ == "__main__":
import traceback
failures = 0
for name, fn in sorted(globals().items()):
if name.startswith("test_") and callable(fn):
try:
fn()
print(f"PASS {name}")
except Exception: # noqa: BLE001
failures += 1
print(f"FAIL {name}")
traceback.print_exc()
sys.exit(1 if failures else 0)