Files
skills/orchestrator/kvb_orchestrator/sources.py
T
Kevin Veen-BirkenbachandClaude Opus 5.5 85b79de0d8 fix(orchestrator): settle queue jobs whose holder is gone instead of showing them running
The hub read queue.json alone, so an active entry whose hold had exited kept showing as running. client/queue_live.py probes each hold lock as queue.py's held() does (shared non-blocking flock, hard links probed too) and reads the whole queue-history.jsonl, cached by inode, size and mtime; an active entry with a history row started after its enqueue ends with that row's exit, one whose holder is gone becomes gone. The outcome lives in the hub-owned field hub_outcome; queue.py status is never rewritten. resources.py uses the same settled(), plan medians skip the queue kind, and a guard test keeps every client module in the Dockerfile COPY line. Also closes fd leaks in read_history and records.read_regular.

MEDUSA: 11 findings in resources.py (lines 73, 81, 262, 445 x2, 576, 227, 479, 524, 578) and crons.py:125 accepted as false positives after an independent review, operator pick 2026-11-12.

🎯 Fix: 92 %
💥 Break: 5 %
💸 Cost: 100 %
⏱️ Time: 100 %
🛡️ Security: 100 %
🏆 Score: 87
🚦 Gate: ✅ 90/5, cost and time <= 100, security 100 %

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 23:09:32 +00:00

58 lines
3.0 KiB
Python

from __future__ import annotations
import time
from .artifacts import enrich
from .council import HUMAN, agent, read_claims
from .crons import cron_state, queue_rows, timer_rows
from .errors import ToolError
from .fields import single
from .lifetimes import edge_times, lifetime, session_gone, subagent_rows
from .names import PSEUDO, SESSION_ID, Directory, title_sightings
from .records import epoch, iso_epoch
from .register import REGISTER, read_register
from .tree import assign
class Sources:
def subagents(self, query: dict[str, list[str]]) -> dict:
session = single(query, "session") or ""
if not SESSION_ID.fullmatch(session):
raise ToolError("session must be a session id")
now = time.time()
beats = [claim["beat"] for claim in read_claims(self.council) if claim["session"] == session and claim["beat"] is not None]
try:
path = next(iter(sorted(self.transcripts.root.glob(f"*/{session}.jsonl"))), None)
except OSError:
path = None
times = edge_times(path) if path is not None else (None, None, None)
life = lifetime(times, session_gone(max(beats, default=None), times[2], now))
return {"session": session, **life, "subagents": subagent_rows(self.transcripts.root, session)}
def directory(self, claims: list[dict], news: list[dict], now: float) -> Directory:
claimed = {claim["session"] for claim in claims}
sightings = title_sightings(self.transcripts.root, claimed, now) if self.names.due(now) else []
sightings += [(claim["session"], claim["agent"], iso_epoch(claim["since"]), "claim") for claim in claims if claim["agent"] and claim["session"] != "-"]
self.names.observe(sorted(sightings, key=lambda sighting: (sighting[2] is None, sighting[2] or 0.0)), now)
known = {entry["name"] for entry in self.names.listed()}
radio = []
for row in news:
name, ts = agent(row.get("from")), epoch(row.get("ts"))
if name and name != HUMAN and name not in known and ts is not None:
radio.append((f"{PSEUDO}{name}", name, ts, "radio"))
known.add(name)
self.names.observe(radio, now)
return Directory(self.names.listed())
def repositories(self) -> dict:
tasks = read_register(self.council / REGISTER)
reviews = self.reviews.listed()
scan = assign(self.tree.get(), tasks, read_claims(self.council), reviews)
history = self.history.rows()
return {**enrich(scan, tasks, reviews, self.touched, history, self.jobs.listed()), "cleanups": self.cleanup_results(), "gates": self.gate_view(scan)}
def schedules(self, claims: list[dict], now: float) -> list[dict]:
beats = {claim["session"]: claim["beat"] for claim in claims if claim["beat"] is not None}
rows = [cron_state(cron, beats, now) for cron in self.transcripts.schedules()]
return rows + timer_rows(self.timers) + queue_rows(self.council, self.history.rows)