Merge pull request 'feat(operations): show real claude.ai subscription usage % on /operations' (#244) from worktree-operations-usage-ui into main

mcp-server/src/legal_mcp/court_fetch_service/server.py

@@ -34,6 +34,7 @@ import logging
 import os
 import sys
 
+import aiohttp
 from aiohttp import web
 
 _pkg_root = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
@@ -93,6 +94,41 @@ async def _pm2_run(*args: str, timeout: float = 10) -> tuple[int, bytes, bytes]:
     return proc.returncode or 0, out, err
 
 
+# claude.ai subscription usage — the 5-hour / weekly utilization % the Claude
+# Code status bar shows, from the (undocumented) OAuth usage endpoint. Host-only:
+# the OAuth token lives in the CLI credentials file on the host, never in the
+# container. Read-only (no auth), like /pm2. The claude-code User-Agent is
+# REQUIRED — without it the request lands in an aggressively rate-limited bucket.
+_CLAUDE_CRED_PATH = "/home/chaim/.claude/.credentials.json"
+_OAUTH_USAGE_URL = "https://api.anthropic.com/api/oauth/usage"
+_USAGE_UA = "claude-code/2.1.177"
+
+
+async def usage_status(request: web.Request) -> web.Response:
+    """Proxy the claude.ai subscription usage % (host-only — needs the local
+    OAuth token). Returns the endpoint's JSON, or a 502 with an error string."""
+    try:
+        with open(_CLAUDE_CRED_PATH) as f:
+            token = json.load(f)["claudeAiOauth"]["accessToken"]
+    except Exception as e:
+        return web.json_response({"error": f"no claude credentials: {e}"}, status=502)
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "User-Agent": _USAGE_UA,
+        "anthropic-beta": "oauth-2025-04-20",
+    }
+    try:
+        timeout = aiohttp.ClientTimeout(total=15)
+        async with aiohttp.ClientSession(timeout=timeout) as session:
+            async with session.get(_OAUTH_USAGE_URL, headers=headers) as r:
+                if r.status != 200:
+                    return web.json_response(
+                        {"error": f"usage endpoint {r.status}"}, status=502)
+                return web.json_response(await r.json())
+    except Exception as e:  # never throw
+        return web.json_response({"error": f"usage fetch failed: {e}"}, status=502)
+
+
 async def pm2_status(request: web.Request) -> web.Response:
     """Return a trimmed ``pm2 jlist`` for the legal-ai background services."""
     try:
@@ -235,6 +271,7 @@ def build_app() -> web.Application:
     app = web.Application(client_max_size=64 * 1024 * 1024)
     app.router.add_get("/health", health)
     app.router.add_get("/pm2", pm2_status)
+    app.router.add_get("/usage", usage_status)
     app.router.add_post("/pm2/control", pm2_control)
     app.router.add_post("/fetch", fetch)
     return app

web-ui/src/app/operations/page.tsx

@@ -33,12 +33,68 @@ import {
   type OperationsSnapshot,
   type PipelineStats,
   type AgentRun,
+  type SubscriptionUsage,
+  type UsageWindow,
 } from "@/lib/api/operations";
 
 function mb(bytes: number): string {
   return `${Math.round((bytes || 0) / 1024 / 1024)}MB`;
 }
 
+// claude.ai subscription usage — the fuel gauge for every agent/drain.
+// Bar turns amber > 75% and red > 90%; the halacha drain auto-pauses at 100%
+// and resumes on its own when a window resets.
+function usageResetLabel(iso: string | null): string {
+  if (!iso) return "—";
+  const d = new Date(iso);
+  return `איפוס ${d.toLocaleTimeString("he-IL", { hour: "2-digit", minute: "2-digit" })}`;
+}
+
+function UsageMeter({ label, w }: { label: string; w?: UsageWindow | null }) {
+  const util = w?.utilization;
+  const active = typeof util === "number";
+  const pct = active ? Math.min(100, Math.max(0, util as number)) : 0;
+  const fill = pct >= 90 ? "bg-danger" : pct >= 75 ? "bg-warn" : "bg-gold";
+  return (
+    <div>
+      <div className="flex items-baseline justify-between text-[13px] text-ink-soft mb-1.5">
+        <span>{label}</span>
+        {active ? (
+          <b className="text-navy tabular-nums text-[15px] font-bold">{Math.round(pct)}%</b>
+        ) : (
+          <b className="text-ink-muted text-sm font-semibold">—</b>
+        )}
+      </div>
+      <div className="h-[7px] overflow-hidden rounded-full bg-rule-soft">
+        <span className={`block h-full rounded-full ${fill}`} style={{ width: `${pct}%` }} />
+      </div>
+      <div className="mt-1.5 text-[11.5px] text-ink-muted">
+        {active ? usageResetLabel(w?.resets_at ?? null) : "לא פעיל כרגע"}
+      </div>
+    </div>
+  );
+}
+
+function SubscriptionUsagePanel({ usage }: { usage: SubscriptionUsage }) {
+  if (!usage) return null; // undocumented endpoint unreachable — show nothing
+  return (
+    <Card className="border-rule">
+      <CardContent className="px-5 py-4">
+        <div className="grid gap-6 sm:grid-cols-3">
+          <UsageMeter label="חלון 5-שעות" w={usage.five_hour} />
+          <UsageMeter label="שבועי · כל המודלים" w={usage.seven_day} />
+          <UsageMeter label="שבועי · Opus" w={usage.seven_day_opus} />
+        </div>
+        <div className="mt-3 border-t border-rule-soft pt-2.5 text-[11.5px] text-ink-muted">
+          המנוי המשותף שמפעיל את כל הסוכנים והדריינרים. הפס נצבע בענבר מעל 75%
+          ובאדום מעל 90%. דריינר-ההלכות משתהה אוטומטית כשחלון מגיע ל-100% וממשיך
+          לבד כשהוא מתאפס.
+        </div>
+      </CardContent>
+    </Card>
+  );
+}
+
 // mockup 02: every region opens with a navy section heading (h2, 18px) — the
 // page reads as a sequence of titled sections rather than a stack of cards.
 function SectionHeader({ children }: { children: React.ReactNode }) {
@@ -777,6 +833,9 @@ export default function OperationsPage() {
           </div>
         ) : (
           <>
+            <SectionHeader>מכסת claude.ai</SectionHeader>
+            <SubscriptionUsagePanel usage={data.subscription_usage} />
+
             <SectionHeader>סוכנים פעילים</SectionHeader>
             <LiveAgentsPanel />
 

web-ui/src/components/training/learning-panel.tsx

@@ -106,7 +106,7 @@ function ProposalReview({ pairId }: { pairId: string }) {
   return (
     <div className="rounded-md bg-gold-wash/30 border border-gold/40 p-3 space-y-3 text-[0.8rem]">
       <p className="text-ink-muted text-[0.75rem]">
-        הצעת הדיסטילציה (סגנון/שיטה בלבד — INV-LRN5). בחר/י מה לאמץ; ייכתב לערוצים שהכותב צורך (שער-יו"ר).
+        הצעת הדיסטילציה (סגנון/שיטה בלבד — INV-LRN5). בחר/י מה לאמץ; ייכתב לערוצים שהכותב צורך (שער-יו&quot;ר).
       </p>
       {data.overall_assessment && (
         <p className="italic text-ink-muted">{data.overall_assessment}</p>

web-ui/src/lib/api/operations.ts

@@ -42,9 +42,26 @@ export type PipelineStats = {
   by_status: Record<string, number>; // raw counts, for the curious
 };
 
+/** One claude.ai usage window (5-hour / weekly / weekly-per-model). */
+export type UsageWindow = {
+  utilization: number | null; // 0-100; null when the window is inactive
+  resets_at: string | null;
+};
+
+/** claude.ai subscription usage — the same %s the Claude Code status bar shows,
+ *  via the (undocumented) OAuth usage endpoint proxied by the host bridge.
+ *  null when the endpoint is unreachable. */
+export type SubscriptionUsage = {
+  five_hour?: UsageWindow | null;
+  seven_day?: UsageWindow | null;
+  seven_day_opus?: UsageWindow | null;
+  seven_day_sonnet?: UsageWindow | null;
+} | null;
+
 export type OperationsSnapshot = {
   services: OpsService[];
   services_error: string | null;
+  subscription_usage: SubscriptionUsage;
   pipelines: {
     court_fetch: PipelineStats & { recent: CourtFetchJob[] };
     metadata_extraction: PipelineStats;

web/app.py

@@ -6525,6 +6525,21 @@ async def _ops_pm2_services() -> dict:
         return {"services": [], "error": f"לא ניתן להגיע לשירות-המארח: {e}"}
 
 
+async def _ops_subscription_usage() -> dict | None:
+    """Proxy the host court-fetch-service /usage — the claude.ai subscription
+    utilization % (5-hour / weekly). Host-only (the OAuth token lives on the
+    host). Returns the endpoint JSON, or None if unavailable (undocumented
+    endpoint — the dashboard shows nothing when this is None)."""
+    try:
+        async with httpx.AsyncClient(timeout=18.0) as client:
+            r = await client.get(f"{_COURT_FETCH_SERVICE_URL}/usage")
+        if r.status_code == 200:
+            return r.json()
+    except Exception:
+        pass
+    return None
+
+
 async def _ops_pm2_control(name: str, action: str) -> dict:
     """Proxy a mutating pm2 action to the host bridge (Bearer-authenticated)."""
     secret = os.environ.get("COURT_FETCH_SHARED_SECRET", "").strip()
@@ -6636,6 +6651,7 @@ async def operations_snapshot():
         ]
 
     pm2 = await _ops_pm2_services()
+    subscription_usage = await _ops_subscription_usage()
     controls = await db.get_drain_controls()
     bursts = await db.get_drain_bursts()
     for svc in pm2["services"]:
@@ -6652,6 +6668,7 @@ async def operations_snapshot():
     return {
         "services": pm2["services"],
         "services_error": pm2["error"],
+        "subscription_usage": subscription_usage,
         "pipelines": {
             "court_fetch": {
                 **_norm_pipeline(