feat(fleet): fan-out parc — vue d'ensemble + OTA groupé, rapport par device

- fleet.py : fan_out() séquentiel isolant les échecs par tablette (§9), et
  device_overview() (résumé santé partiel, sous-requêtes isolées).
- API : GET /api/fleet/overview (enrichi des labels), POST /api/fleet/update-check,
  POST /api/fleet/update-install (audité).
- UI : bouton « Vue parc » → tableau d'ensemble (état, RAM, MAJ, RAG, profils,
  sessions, crashes) + vérification OTA sur tout le parc.
- tests : +6 (isolation des échecs, overview partiel, endpoint). 25/25.
- validé live : overview + update-check sur tablette réelle (parse_call_bundle OK).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
alf 2026-06-18 22:52:11 +02:00
parent 7f999a06f0
commit 4b38efca4f
4 changed files with 224 additions and 0 deletions

View File

@ -22,6 +22,7 @@ from pydantic import BaseModel
from starlette.responses import Response from starlette.responses import Response
from ..adb import Adb, AdbError from ..adb import Adb, AdbError
from ..fleet import device_overview, fan_out
from ..provider import ProviderClient from ..provider import ProviderClient
from ..store import BadPassword, Store, StoreLocked from ..store import BadPassword, Store, StoreLocked
@ -176,6 +177,27 @@ def create_app(adb: Adb | None = None, store: Store | None = None) -> FastAPI:
def sessions(serial: str, profile_id: str | None = None): def sessions(serial: str, profile_id: str | None = None):
return [s.model_dump() for s in _guard(lambda: client(serial).sessions(profile_id))] return [s.model_dump() for s in _guard(lambda: client(serial).sessions(profile_id))]
# ---- flotte (fan-out parc, rapport par device §9) ---------------------
@app.get("/api/fleet/overview")
def fleet_overview():
report = fan_out(adb, lambda s: device_overview(adb, s))
labels = store.device_labels() if store.is_unlocked else {}
for r in report:
r["label"] = labels.get(r["serial"])
return report
@app.post("/api/fleet/update-check")
def fleet_update_check():
return fan_out(adb, lambda s: client(s).update_check())
@app.post("/api/fleet/update-install")
def fleet_update_install():
report = fan_out(adb, lambda s: client(s).update_install())
store.audit("fleet_update_install", actor=_OPERATOR,
detail={"devices": [r["serial"] for r in report if r["ok"]]},
now=_now_ms())
return report
# UI web locale servie à la racine — APRÈS les routes /api pour ne pas les # UI web locale servie à la racine — APRÈS les routes /api pour ne pas les
# masquer. `html=True` → sert index.html sur "/". # masquer. `html=True` → sert index.html sur "/".
if _WEB_DIR.is_dir(): if _WEB_DIR.is_dir():

68
kazeia_central/fleet.py Normal file
View File

@ -0,0 +1,68 @@
"""Opérations de flotte (fan-out) pour Kazeia-central.
§9 du contrat : toute action parc se fait par **fan-out séquentiel par device**, et
les commandes adb **échouent silencieusement par tablette** on capture le résultat
OU l'erreur de CHAQUE device, jamais un échec global qui masquerait les autres. Le
rapport renvoyé est une liste `{serial, ok, result|error}` exploitable telle quelle.
Séquentiel (pas parallèle) volontairement : conforme au contrat, et un `adb` partagé
sature vite ; pour un parc de quelques tablettes le coût est négligeable.
"""
from __future__ import annotations
from typing import Any, Callable
from .adb import Adb
from .provider import ProviderClient
def fan_out(adb: Adb, fn: Callable[[str], Any], *,
serials: list[str] | None = None) -> list[dict[str, Any]]:
"""Exécute `fn(serial)` sur chaque tablette prête (ou la sous-liste `serials`),
en isolant les échecs. Renvoie un rapport par device."""
ready = adb.ready_devices()
if serials is not None:
wanted = set(serials)
ready = [d for d in ready if d.serial in wanted]
report: list[dict[str, Any]] = []
for d in ready:
try:
report.append({"serial": d.serial, "ok": True, "result": fn(d.serial)})
except Exception as e: # adb/provider peut échouer pour CE device seulement
report.append({"serial": d.serial, "ok": False, "error": str(e)})
return report
def device_overview(adb: Adb, serial: str) -> dict[str, Any]:
"""Résumé santé d'une tablette pour la vue parc. Chaque sous-requête est isolée :
une section indisponible ne fait pas tomber tout le résumé (renvoi partiel)."""
c = ProviderClient(adb, serial)
def safe(f, default=None):
try:
return f()
except Exception:
return default
state = safe(lambda: c.state().model_dump())
updates = safe(lambda: c.updates().model_dump())
rag = safe(lambda: c.rag_status().model_dump())
profiles = safe(lambda: c.profiles(), [])
sessions = safe(lambda: c.sessions(), [])
crashes = safe(lambda: c.crashes(), [])
g = lambda d, k: (d or {}).get(k)
return {
"serial": serial,
"online": state is not None,
"uptime_s": g(state, "uptime_seconds"),
"mem_available_mb": g(state, "mem_available_mb"),
"update_phase": g(updates, "phase"),
"app_update_available": g(updates, "app_update_available"),
"rag_ready": g(rag, "ready"),
"rag_docs": g(rag, "doc_count"),
"profiles": len(profiles),
"sessions": len(sessions),
"crashes": len(crashes),
}

View File

@ -72,6 +72,7 @@
<span class="sub" id="sub">console de flotte</span> <span class="sub" id="sub">console de flotte</span>
<div class="right"> <div class="right">
<span id="lockchip"></span> <span id="lockchip"></span>
<button onclick="fleetView()">▦ Vue parc</button>
<button onclick="loadDevices()">↻ Parc</button> <button onclick="loadDevices()">↻ Parc</button>
</div> </div>
</header> </header>
@ -173,6 +174,51 @@ async function loadDevices() {
} }
} }
// ---- vue parc (fan-out) --------------------------------------------------
async function fleetView() {
selected = null; loadDevices();
const d = $("detail");
d.innerHTML = `<div class="muted">interrogation du parc…</div>`;
try {
const rows = await api("/fleet/overview");
const yn = (v) => v == null ? "—" : (v ? "oui" : "non");
const body = rows.map(r => {
const o = r.result || {};
const who = r.label ? `👤 ${r.label}` : `<span class="serial">${r.serial}</span>`;
if (!r.ok) return `<tr><td>${who}</td><td colspan="9" class="err">${r.error}</td></tr>`;
return `<tr>
<td>${who}</td>
<td>${dot(o.online)}${o.online?"en ligne":"hors-ligne"}</td>
<td>${o.uptime_s ?? "—"}</td>
<td>${o.mem_available_mb ?? "—"}</td>
<td>${o.update_phase ?? "—"}${o.app_update_available?" ⬆":""}</td>
<td>${o.rag_ready?dot(true):dot(false)}${o.rag_docs ?? 0}</td>
<td>${o.profiles ?? 0}</td>
<td>${o.sessions ?? 0}</td>
<td>${o.crashes ? `<span class="err">${o.crashes}</span>` : 0}</td></tr>`;
}).join("");
d.innerHTML =
`<div style="margin-bottom:12px;display:flex;gap:8px;align-items:center">
<button onclick="fleetUpdateCheck()">Vérifier les MAJ (tout le parc)</button>
<span id="fleetmsg" class="muted"></span>
</div>
<table>
<tr><th>tablette</th><th>état</th><th>uptime(s)</th><th>RAM(Mo)</th><th>MAJ</th>
<th>RAG</th><th>profils</th><th>sessions</th><th>crashes</th></tr>
${body || `<tr><td colspan="9" class="muted">aucune tablette prête</td></tr>`}
</table>`;
} catch (e) { d.innerHTML = `<div class="err">vue parc : ${e.message}</div>`; }
}
async function fleetUpdateCheck() {
$("fleetmsg").textContent = "vérification en cours…";
try {
const rep = await api("/fleet/update-check", {method:"POST"});
const ok = rep.filter(r => r.ok).length;
$("fleetmsg").textContent = `${ok}/${rep.length} tablette(s) ont accepté la vérification`;
} catch (e) { $("fleetmsg").textContent = e.message; }
}
function card(title, body, cls="") { return `<div class="card ${cls}"><h3>${title}</h3>${body}</div>`; } function card(title, body, cls="") { return `<div class="card ${cls}"><h3>${title}</h3>${body}</div>`; }
function kv(k, v) { return `<div class="kv"><span class="k">${k}</span><span class="v">${v ?? "—"}</span></div>`; } function kv(k, v) { return `<div class="kv"><span class="k">${k}</span><span class="v">${v ?? "—"}</span></div>`; }
function dot(on) { return `<span class="dot ${on ? "on":"off"}"></span>`; } function dot(on) { return `<span class="dot ${on ? "on":"off"}"></span>`; }

88
tests/test_fleet.py Normal file
View File

@ -0,0 +1,88 @@
"""Tests du fan-out flotte : isolation des échecs par device (§9) + overview."""
import nacl.pwhash
from fastapi.testclient import TestClient
from kazeia_central.adb import Adb, AdbError
from kazeia_central.adb.client import AdbDevice
from kazeia_central.api.app import create_app
from kazeia_central.fleet import device_overview, fan_out
from kazeia_central.store import Store
class _ReadyAdb(Adb):
"""adb factice : liste de tablettes prêtes, content_query cannée."""
def __init__(self, serials, canned=None, fail_paths=()):
super().__init__()
self._serials = serials
self._canned = canned or {}
self._fail = set(fail_paths)
def ready_devices(self):
return [AdbDevice(s, "device", {"model": "Pad3"}) for s in self._serials]
def content_query(self, path, *, serial=None, where=None, expect_rows=False, _retry=True):
if path in self._fail:
raise AdbError(f"{path} indispo")
return self._canned.get(path, [])
_CANNED = {
"state": [{"pid": 1, "uptime_seconds": 100, "pss_mb": 2, "ion_mb": 3, "mem_available_mb": 500}],
"updates": [{"phase": "UPTODATE", "app_update_available": 0}],
"rag_status": [{"enabled": 1, "ready": 1, "doc_count": 18}],
"profiles": [{"id": "p1"}, {"id": "p2"}],
"conversations": [{"profile_id": "p1", "session_id": "s1"}],
"crashes": [],
}
def test_fan_out_isolates_failures():
adb = _ReadyAdb(["D1", "D2", "D3"])
def fn(serial):
if serial == "D2":
raise RuntimeError("boom sur D2")
return f"ok-{serial}"
rep = fan_out(adb, fn)
assert [r["serial"] for r in rep] == ["D1", "D2", "D3"]
assert rep[0] == {"serial": "D1", "ok": True, "result": "ok-D1"}
assert rep[1]["ok"] is False and "boom sur D2" in rep[1]["error"]
assert rep[2]["ok"] is True # D3 non affecté par l'échec de D2
def test_fan_out_respects_serials_subset():
adb = _ReadyAdb(["D1", "D2", "D3"])
rep = fan_out(adb, lambda s: s, serials=["D1", "D3"])
assert [r["serial"] for r in rep] == ["D1", "D3"]
def test_fan_out_no_ready_devices():
assert fan_out(_ReadyAdb([]), lambda s: s) == []
def test_device_overview_full():
o = device_overview(_ReadyAdb(["D1"], _CANNED), "D1")
assert o["online"] and o["uptime_s"] == 100 and o["mem_available_mb"] == 500
assert o["rag_docs"] == 18 and o["profiles"] == 2 and o["sessions"] == 1 and o["crashes"] == 0
assert o["update_phase"] == "UPTODATE"
def test_device_overview_partial_when_state_down():
# /state KO → online False, mais le reste reste compté (renvoi partiel)
o = device_overview(_ReadyAdb(["D1"], _CANNED, fail_paths=["state"]), "D1")
assert o["online"] is False and o["uptime_s"] is None
assert o["profiles"] == 2 and o["rag_docs"] == 18
def test_fleet_overview_endpoint_merges_labels(tmp_path):
store = Store(tmp_path / "c.db",
kdf_ops=nacl.pwhash.argon2id.OPSLIMIT_MIN,
kdf_mem=nacl.pwhash.argon2id.MEMLIMIT_MIN)
c = TestClient(create_app(adb=_ReadyAdb(["D1"], _CANNED), store=store))
c.post("/api/unlock", json={"password": "pw"})
c.put("/api/devices/D1/label", json={"label": "Mme D."})
rep = c.get("/api/fleet/overview").json()
assert rep[0]["ok"] and rep[0]["label"] == "Mme D."
assert rep[0]["result"]["profiles"] == 2