relay-codex 0.0.0-stage → 0.5.0

package/python/pool.py

@@ -0,0 +1,586 @@
+"""Cross-process coordination. Contains usage metadata, never tokens or prompts."""
+
+from __future__ import annotations
+import asyncio
+import contextlib
+import fcntl
+import json
+import os
+from pathlib import Path
+import sqlite3
+import time
+import uuid
+import subprocess
+import functools
+
+
+@contextlib.asynccontextmanager
+async def file_lock(path, *, shared=False):
+    """Nonblocking flock with cancellable waiting, never block the event loop."""
+    fd = os.open(path, os.O_CREAT | os.O_RDWR, 0o600)
+    try:
+        while True:
+            try:
+                fcntl.flock(
+                    fd, (fcntl.LOCK_SH if shared else fcntl.LOCK_EX) | fcntl.LOCK_NB
+                )
+                break
+            except BlockingIOError:
+                await asyncio.sleep(0.05)
+        yield
+    finally:
+        os.close(fd)
+
+
+def buckets(snapshot, model=None):
+    values = snapshot.get("rateLimitsByLimitId")
+    if not values:
+        return [snapshot.get("rateLimits") or {}]
+    return [
+        v
+        for k, v in values.items()
+        if k == "codex" or (model and v.get("normalModelSlug") == model)
+    ]
+
+
+def blocked(snapshot, model=None):
+    return snapshot.get("ordinaryUsageAllowed") is False or any(
+        b.get("rateLimitReachedType") or b.get("spendControlReached")
+        for b in buckets(snapshot, model)
+    )
+
+
+def reset_time(snapshot, model=None, now=None):
+    """For simultaneous exhausted windows, all must reset: take the latest.
+
+    A timestamp only schedules a fresh read; it never establishes recovery.
+    """
+    now = time.time() if now is None else now
+    windows = [
+        b.get(k) or {}
+        for b in buckets(snapshot, model)
+        for k in ("primary", "secondary")
+    ]
+    exhausted = [w for w in windows if w.get("usedPercent", 0) >= 100]
+    dates = [w.get("resetsAt") for w in exhausted]
+    if dates and all(isinstance(d, (int, float)) and d > now for d in dates):
+        return max(dates) + 2  # Small grace for server clock/propagation.
+    return None
+
+
+class Pool:
+    def __init__(self, state: Path, *, clock=time.time):
+        self.state, self.clock = state, clock
+        self.session = uuid.uuid4().hex
+        self.path = state / "pool.sqlite"
+        with self.db() as db:
+            db.executescript("""
+              CREATE TABLE IF NOT EXISTS usage(label TEXT PRIMARY KEY, checked REAL, snapshot TEXT);
+              CREATE TABLE IF NOT EXISTS blocks(label TEXT, scope TEXT, until REAL, reason TEXT,
+                PRIMARY KEY(label,scope));
+              CREATE TABLE IF NOT EXISTS sessions(id TEXT PRIMARY KEY,pid INTEGER,label TEXT);
+              CREATE TABLE IF NOT EXISTS threads(thread TEXT PRIMARY KEY,owner TEXT,pid INTEGER);
+              CREATE TABLE IF NOT EXISTS account_options(label TEXT PRIMARY KEY, options TEXT NOT NULL);
+              CREATE TABLE IF NOT EXISTS session_details(id TEXT PRIMARY KEY, birth TEXT, seen REAL, details TEXT NOT NULL);
+              CREATE TABLE IF NOT EXISTS reservations(owner TEXT PRIMARY KEY,label TEXT,until REAL);
+              CREATE TABLE IF NOT EXISTS intents(owner TEXT,thread TEXT,details TEXT,PRIMARY KEY(owner,thread));
+              CREATE TABLE IF NOT EXISTS events(id INTEGER PRIMARY KEY AUTOINCREMENT,at REAL,owner TEXT,kind TEXT,details TEXT);
+              CREATE TABLE IF NOT EXISTS history(id INTEGER PRIMARY KEY AUTOINCREMENT,at REAL,label TEXT,snapshot TEXT);
+              CREATE TABLE IF NOT EXISTS commands(id INTEGER PRIMARY KEY AUTOINCREMENT,owner TEXT,op TEXT,details TEXT,at REAL,status TEXT);
+              CREATE TABLE IF NOT EXISTS signals(name TEXT PRIMARY KEY,revision INTEGER);
+            """)
+        self.path.chmod(0o600)
+
+    @contextlib.contextmanager
+    def db(self):
+        db = sqlite3.connect(self.path, timeout=5)
+        try:
+            with db:
+                yield db
+        finally:
+            db.close()
+
+    @staticmethod
+    def alive(pid):
+        try:
+            os.kill(pid, 0)
+            return True
+        except ProcessLookupError:
+            return False
+        except PermissionError:
+            return True
+
+    def prune(self, db):
+        for table, key in [("sessions", "id"), ("threads", "thread")]:
+            for ident, pid in db.execute(f"SELECT {key},pid FROM {table}").fetchall():
+                row = db.execute(
+                    "SELECT birth FROM session_details WHERE id=?",
+                    (
+                        ident
+                        if table == "sessions"
+                        else (
+                            db.execute(
+                                "SELECT owner FROM threads WHERE thread=?", (ident,)
+                            ).fetchone()
+                            or [""]
+                        )[0],
+                    ),
+                ).fetchone()
+                birth_matches = not row or not row[0] or row[0] == process_birth(pid)
+                if not self.alive(pid) or not birth_matches:
+                    db.execute(f"DELETE FROM {table} WHERE {key}=?", (ident,))
+        db.execute(
+            "DELETE FROM reservations WHERE until<? OR owner NOT IN (SELECT id FROM sessions)",
+            (self.clock(),),
+        )
+
+    def assign(self, label):
+        with self.db() as db:
+            self.prune(db)
+            db.execute(
+                "INSERT OR REPLACE INTO sessions VALUES(?,?,?)",
+                (self.session, os.getpid(), label),
+            )
+            db.execute("DELETE FROM reservations WHERE owner=?", (self.session,))
+        self.touch(account=label)
+
+    def loads(self):
+        with self.db() as db:
+            self.prune(db)
+            rows = dict(
+                db.execute(
+                    "SELECT label,count(*) FROM sessions WHERE id<>? GROUP BY label",
+                    (self.session,),
+                )
+            )
+            for label, count in db.execute(
+                "SELECT label,count(*) FROM reservations WHERE owner<>? GROUP BY label",
+                (self.session,),
+            ):
+                rows[label] = rows.get(label, 0) + count
+            # Public loads still include this session for status and compatibility.
+            own = db.execute(
+                "SELECT label FROM sessions WHERE id=?", (self.session,)
+            ).fetchone()
+            if own:
+                rows[own[0]] = rows.get(own[0], 0) + 1
+            return rows
+
+    def reserve(self, label, expected_load=None, limit=0, ttl=120):
+        with self.db() as db:
+            db.execute("BEGIN IMMEDIATE")
+            self.prune(db)
+            row = db.execute(
+                "SELECT options FROM account_options WHERE label=?", (label,)
+            ).fetchone()
+            options = json.loads(row[0]) if row else {}
+            if maintenance_active(options) or not options.get("enabled", True):
+                return False
+            configured = options.get("max_sessions", 0)
+            limit = (
+                min(limit, configured) if limit and configured else limit or configured
+            )
+            current = db.execute(
+                "SELECT count(*) FROM sessions WHERE label=? AND id<>?",
+                (label, self.session),
+            ).fetchone()[0]
+            current += db.execute(
+                "SELECT count(*) FROM reservations WHERE label=? AND owner<>?",
+                (label, self.session),
+            ).fetchone()[0]
+            own = db.execute(
+                "SELECT label FROM sessions WHERE id=?", (self.session,)
+            ).fetchone()
+            compare = current + (1 if own and own[0] == label else 0)
+            if (
+                limit
+                and current >= limit
+                or expected_load is not None
+                and compare > expected_load
+            ):
+                return False
+            # Reserve before any network call. No inference is moved until login succeeds.
+            db.execute(
+                "INSERT OR IGNORE INTO sessions VALUES(?,?,?)",
+                (self.session, os.getpid(), ""),
+            )
+            db.execute(
+                "INSERT OR REPLACE INTO reservations VALUES(?,?,?)",
+                (self.session, label, self.clock() + ttl),
+            )
+        self.touch()
+        return True
+
+    def begin_maintenance(self, label):
+        with self.db() as db:
+            db.execute("BEGIN IMMEDIATE")
+            self.prune(db)
+            if (
+                db.execute("SELECT 1 FROM sessions WHERE label=?", (label,)).fetchone()
+                or db.execute(
+                    "SELECT 1 FROM reservations WHERE label=?", (label,)
+                ).fetchone()
+            ):
+                return False
+            row = db.execute(
+                "SELECT options FROM account_options WHERE label=?", (label,)
+            ).fetchone()
+            options = json.loads(row[0]) if row else {}
+            if maintenance_active(options):
+                return False
+            options.update(
+                maintenance=True,
+                maintenance_pid=os.getpid(),
+                maintenance_birth=process_birth(os.getpid()),
+            )
+            db.execute(
+                "INSERT OR REPLACE INTO account_options VALUES(?,?)",
+                (label, json.dumps(options)),
+            )
+        self.signal("accounts")
+        self.signal()
+        return True
+
+    def release_reservation(self):
+        with self.db() as db:
+            db.execute("DELETE FROM reservations WHERE owner=?", (self.session,))
+
+    def renew_reservation(self, ttl=120):
+        with self.db() as db:
+            db.execute(
+                "UPDATE reservations SET until=? WHERE owner=? AND until>?",
+                (self.clock() + ttl, self.session, self.clock()),
+            )
+
+    def touch(self, **details):
+        with self.db() as db:
+            row = db.execute(
+                "SELECT details FROM session_details WHERE id=?", (self.session,)
+            ).fetchone()
+            value = json.loads(row[0]) if row else {}
+            value.update(details)
+            db.execute(
+                "INSERT OR REPLACE INTO session_details VALUES(?,?,?,?)",
+                (
+                    self.session,
+                    process_birth(os.getpid()),
+                    self.clock(),
+                    json.dumps(value),
+                ),
+            )
+
+    def detail(self):
+        with self.db() as db:
+            self.prune(db)
+            rows = db.execute(
+                "SELECT s.id,s.pid,s.label,d.seen,d.details FROM sessions s LEFT JOIN session_details d ON s.id=d.id"
+            ).fetchall()
+        return [
+            {
+                **json.loads(r[4] or "{}"),
+                "id": r[0],
+                "pid": r[1],
+                "account": r[2],
+                "seen": r[3],
+            }
+            for r in rows
+        ]
+
+    def options(self, label):
+        with self.db() as db:
+            row = db.execute(
+                "SELECT options FROM account_options WHERE label=?", (label,)
+            ).fetchone()
+        value = {
+            "enabled": True,
+            "priority": 0,
+            "max_sessions": 0,
+            "models": [],
+            **(json.loads(row[0]) if row else {}),
+        }
+        if value.get("maintenance") and not maintenance_active(value):
+            value["maintenance"] = False
+        return value
+
+    def set_options(self, label, **values):
+        value = self.options(label)
+        value.update(values)
+        with self.db() as db:
+            db.execute(
+                "INSERT OR REPLACE INTO account_options VALUES(?,?)",
+                (label, json.dumps(value)),
+            )
+        self.signal("accounts")
+        self.signal()
+
+    def signal(self, name="wake"):
+        with self.db() as db:
+            db.execute(
+                "INSERT INTO signals VALUES(?,1) ON CONFLICT(name) DO UPDATE SET revision=revision+1",
+                (name,),
+            )
+
+    def revision(self, name="wake"):
+        with self.db() as db:
+            row = db.execute(
+                "SELECT revision FROM signals WHERE name=?", (name,)
+            ).fetchone()
+        return row[0] if row else 0
+
+    def intent(self, thread, details=None):
+        with self.db() as db:
+            if details is None:
+                row = db.execute(
+                    "SELECT details FROM intents WHERE owner=? AND thread=?",
+                    (self.session, thread),
+                ).fetchone()
+                return json.loads(row[0]) if row else None
+            db.execute(
+                "INSERT OR REPLACE INTO intents VALUES(?,?,?)",
+                (self.session, thread, json.dumps(details)),
+            )
+
+    def record(self, kind, **details):
+        # Callers use fixed reason codes and metadata, never raw exceptions/payloads.
+        with self.db() as db:
+            import settings
+
+            limit = settings.load(self.state)["event_limit"]
+            db.execute(
+                "INSERT INTO events(at,owner,kind,details) VALUES(?,?,?,?)",
+                (self.clock(), self.session, kind, json.dumps(details)),
+            )
+            db.execute(
+                "DELETE FROM events WHERE id NOT IN (SELECT id FROM events ORDER BY id DESC LIMIT ?)",
+                (limit,),
+            )
+
+    def command(self, owner, op, details=None):
+        with self.db() as db:
+            cursor = db.execute(
+                "INSERT INTO commands(owner,op,details,at,status) VALUES(?,?,?,?,?)",
+                (owner, op, json.dumps(details or {}), self.clock(), "pending"),
+            )
+        self.signal()
+        return cursor.lastrowid
+
+    def take_commands(self):
+        with self.db() as db:
+            db.execute("BEGIN IMMEDIATE")
+            rows = db.execute(
+                "SELECT id,op,details FROM commands WHERE owner=? AND status='pending' ORDER BY id",
+                (self.session,),
+            ).fetchall()
+            for row in rows:
+                db.execute(
+                    "UPDATE commands SET status='received' WHERE id=?", (row[0],)
+                )
+        return [(r[0], r[1], json.loads(r[2])) for r in rows]
+
+    def ack(self, command, status):
+        with self.db() as db:
+            db.execute("UPDATE commands SET status=? WHERE id=?", (status, command))
+
+    def claim(self, thread):
+        with self.db() as db:
+            db.execute("BEGIN IMMEDIATE")
+            self.prune(db)
+            row = db.execute(
+                "SELECT owner FROM threads WHERE thread=?", (thread,)
+            ).fetchone()
+            if row and row[0] != self.session:
+                return False
+            db.execute(
+                "INSERT OR REPLACE INTO threads VALUES(?,?,?)",
+                (thread, self.session, os.getpid()),
+            )
+            return True
+
+    def owner(self, thread):
+        with self.db() as db:
+            self.prune(db)
+            row = db.execute(
+                "SELECT owner FROM threads WHERE thread=?", (thread,)
+            ).fetchone()
+            return row[0] if row else None
+
+    def release_thread(self, thread):
+        with self.db() as db:
+            db.execute(
+                "DELETE FROM threads WHERE thread=? AND owner=?", (thread, self.session)
+            )
+
+    def close(self):
+        with self.db() as db:
+            db.execute("DELETE FROM sessions WHERE id=?", (self.session,))
+            db.execute("DELETE FROM threads WHERE owner=?", (self.session,))
+            db.execute("DELETE FROM reservations WHERE owner=?", (self.session,))
+
+    def save(self, label, snapshot):
+        with self.db() as db:
+            db.execute(
+                "INSERT OR REPLACE INTO usage VALUES(?,?,?)",
+                (label, self.clock(), json.dumps(snapshot)),
+            )
+            db.execute(
+                "INSERT INTO history(at,label,snapshot) VALUES(?,?,?)",
+                (self.clock(), label, json.dumps(snapshot)),
+            )
+            import settings
+
+            db.execute(
+                "DELETE FROM history WHERE at<?",
+                (self.clock() - settings.load(self.state)["history_days"] * 86400,),
+            )
+
+    def checked_at(self, label):
+        with self.db() as db:
+            row = db.execute(
+                "SELECT checked FROM usage WHERE label=?", (label,)
+            ).fetchone()
+        return row[0] if row else None
+
+    def cached(self, label, max_age=60, not_before=None):
+        with self.db() as db:
+            row = db.execute(
+                "SELECT checked,snapshot FROM usage WHERE label=?", (label,)
+            ).fetchone()
+        if (
+            row
+            and self.clock() - row[0] < max_age
+            and (not_before is None or row[0] >= not_before)
+        ):
+            return json.loads(row[1])
+        return None
+
+    def mark(self, label, model=None, *, snapshot=None, reason="quota", delay=300):
+        if snapshot is not None:
+            self.save(label, snapshot)
+        until = (
+            reset_time(snapshot or {}, model, self.clock())
+            if reason == "quota"
+            else None
+        )
+        until = until or self.clock() + delay
+        scope = model or "*"
+        with self.db() as db:
+            db.execute(
+                "INSERT OR REPLACE INTO blocks VALUES(?,?,?,?)",
+                (label, scope, until, reason),
+            )
+        return until
+
+    def deadline(self, label, model=None):
+        with self.db() as db:
+            rows = db.execute(
+                "SELECT until FROM blocks WHERE label=? AND (scope='*' OR scope=?)",
+                (label, model or "*"),
+            ).fetchall()
+        return max((row[0] for row in rows), default=0)
+
+    def invalidate(self, label):
+        with self.db() as db:
+            db.execute("DELETE FROM usage WHERE label=?", (label,))
+
+    def clear(self, label, model=None):
+        with self.db() as db:
+            db.execute(
+                "DELETE FROM blocks WHERE label=? AND (scope='*' OR scope=?)",
+                (label, model or "*"),
+            )
+
+    def reset_identity(self, label):
+        with self.db() as db:
+            for table in ("usage", "blocks", "history"):
+                db.execute(f"DELETE FROM {table} WHERE label=?", (label,))
+        self.signal()
+
+
+def process_birth(pid):
+    if pid == os.getpid():
+        return _own_birth()
+    return _birth(pid)
+
+
+def maintenance_active(options):
+    if not options.get("maintenance"):
+        return False
+    pid = options.get("maintenance_pid")
+    return (
+        pid is None
+        or Pool.alive(pid)
+        and process_birth(pid) == options.get("maintenance_birth")
+    )
+
+
+def _birth(pid):
+    try:
+        return subprocess.run(
+            ["ps", "-p", str(pid), "-o", "lstart="],
+            capture_output=True,
+            text=True,
+            timeout=2,
+        ).stdout.strip()
+    except (OSError, subprocess.SubprocessError):
+        return ""
+
+
+@functools.lru_cache(maxsize=1)
+def _own_birth():
+    return _birth(os.getpid())
+
+
+CAPACITY_MESSAGE = "Selected model is at capacity. Please try a different model."
+
+
+def failure_kind(info, message=None):
+    if info == "policyCapacityUnavailable":
+        return "capacity"
+    # Older/generic error envelopes can omit the enum. Match only the known
+    # terminal capacity message, never broad words like "policy" or "capacity".
+    if (
+        info in (None, "other", "serverOverloaded", "flexUnavailable")
+        and isinstance(message, str)
+        and message.strip().removeprefix("■").strip() == CAPACITY_MESSAGE
+    ):
+        return "capacity"
+    if isinstance(info, str):
+        if info == "usageLimitExceeded":
+            return "quota"
+        if info == "unauthorized":
+            return "auth"
+        if info in {
+            "rateLimitExceeded",
+            "internalServerError",
+            "serverOverloaded",
+            "flexUnavailable",
+        }:
+            return "transient"
+        return None
+    if isinstance(info, dict):
+        for key in [
+            "httpConnectionFailed",
+            "responseStreamConnectionFailed",
+            "responseStreamDisconnected",
+            "responseTooManyFailedAttempts",
+        ]:
+            if key in info:
+                status = (info[key] or {}).get("httpStatusCode")
+                if status == 401:
+                    return "auth"
+                if (
+                    status is None
+                    or isinstance(status, int)
+                    and (
+                        status in (408, 429)
+                        or 500 <= status < 600
+                        or status == 200
+                        and key
+                        in {
+                            "responseStreamDisconnected",
+                            "responseTooManyFailedAttempts",
+                        }
+                    )
+                ):
+                    return "transient"
+    return None

package/python/probe_native.py

@@ -0,0 +1,91 @@
+"""Verify installed native login controls using fake identities, with network denied.
+
+This does not verify server authorization, model generation, or quota failover.
+"""
+
+import asyncio
+import base64
+import json
+import os
+import tempfile
+import time
+import argparse
+
+from router import RPC, codex_path
+
+
+def synthetic_token(account):
+    def encode(value):
+        return base64.urlsafe_b64encode(json.dumps(value).encode()).decode().rstrip("=")
+
+    return ".".join(
+        [
+            encode({"alg": "none", "typ": "JWT"}),
+            encode(
+                {
+                    "exp": int(time.time()) + 3600,
+                    "email": f"{account}@example.invalid",
+                    "https://api.openai.com/auth": {
+                        "chatgpt_account_id": account,
+                        "chatgpt_user_id": account,
+                        "chatgpt_plan_type": "plus",
+                    },
+                }
+            ),
+            "fake-signature",
+        ]
+    )
+
+
+async def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--binary", default=None)
+    args = parser.parse_args()
+    with tempfile.TemporaryDirectory(prefix="codex-accounts-offline-probe-") as folder:
+        env = dict(os.environ)
+        env["CODEX_HOME"] = folder
+        for key in ("OPENAI_API_KEY", "CODEX_API_KEY", "CODEX_ACCESS_TOKEN"):
+            env.pop(key, None)
+        proc = await asyncio.create_subprocess_exec(
+            "/usr/bin/sandbox-exec",
+            "-p",
+            "(version 1) (allow default) (deny network*)",
+            args.binary or codex_path(),
+            "app-server",
+            "--stdio",
+            "-c",
+            'cli_auth_credentials_store="ephemeral"',
+            env=env,
+            stdin=asyncio.subprocess.PIPE,
+            stdout=asyncio.subprocess.PIPE,
+            stderr=asyncio.subprocess.DEVNULL,
+        )
+        rpc = RPC(proc)
+        pid = proc.pid
+        try:
+            await rpc.initialize()
+            for account in ("synthetic-one", "synthetic-two"):
+                result = await rpc.call(
+                    "account/login/start",
+                    {
+                        "type": "chatgptAuthTokens",
+                        "accessToken": synthetic_token(account),
+                        "chatgptAccountId": account,
+                        "chatgptPlanType": "plus",
+                    },
+                )
+                assert result.get("type") == "chatgptAuthTokens"
+                assert proc.pid == pid and proc.returncode is None
+            print(
+                "PASS: native app-server accepted two synthetic login changes in one process."
+            )
+            print("Network denied by OS sandbox. No real accounts or model calls used.")
+            print(
+                "Not verified: real authentication, thread continuation, or actual quota handoff."
+            )
+        finally:
+            await rpc.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

package/python/processes.py

@@ -0,0 +1,12 @@
+"""Reap detached children while their launcher remains alive."""
+
+import subprocess
+import threading
+
+
+def spawn(*args, **kwargs):
+    process = subprocess.Popen(*args, **kwargs)
+    threading.Thread(
+        target=process.wait, daemon=True, name="codex-child-reaper"
+    ).start()
+    return process

package/python/requirements.lock

@@ -0,0 +1,6 @@
+attrs==26.1.0
+jsonschema==4.26.0
+jsonschema-specifications==2025.9.1
+referencing==0.37.0
+rpds-py==2026.9.1
+websockets==15.0.1

package/python/requirements.txt

@@ -0,0 +1,2 @@
+websockets>=15,<16
+jsonschema>=4.25,<5