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

@@ -33,6 +33,7 @@ import json
 import logging
 import os
 import sys
+import time
 
 import aiohttp
 from aiohttp import web
@@ -102,16 +103,31 @@ async def _pm2_run(*args: str, timeout: float = 10) -> tuple[int, bytes, bytes]:
 _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"
+# /operations polls every 5s; the usage endpoint 429s if hit that often (it's
+# meant for a status bar, not a poll loop). Cache the last good payload and only
+# re-fetch when older than this — Anthropic sees ~1 req/min regardless of how
+# many dashboards poll. The 5-hour window moves slowly, so 60s is plenty fresh.
+_USAGE_TTL_SEC = 60.0
+_usage_cache: dict = {"ts": 0.0, "data": None}
 
 
 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."""
+    OAuth token), cached for _USAGE_TTL_SEC. On a fetch failure (e.g. the
+    endpoint's own 429) serve the last good payload if we have one, so a
+    transient limit doesn't blank the dashboard."""
+    now = time.monotonic()
+    if _usage_cache["data"] is not None and (now - _usage_cache["ts"]) < _USAGE_TTL_SEC:
+        return web.json_response(_usage_cache["data"])
+
     try:
         with open(_CLAUDE_CRED_PATH) as f:
             token = json.load(f)["claudeAiOauth"]["accessToken"]
     except Exception as e:
+        if _usage_cache["data"] is not None:
+            return web.json_response(_usage_cache["data"])
         return web.json_response({"error": f"no claude credentials: {e}"}, status=502)
+
     headers = {
         "Authorization": f"Bearer {token}",
         "User-Agent": _USAGE_UA,
@@ -122,12 +138,17 @@ async def usage_status(request: web.Request) -> web.Response:
         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(
+                    raise RuntimeError(f"usage endpoint {r.status}")
-                        {"error": f"usage endpoint {r.status}"}, status=502)
+                data = await r.json()
-                return web.json_response(await r.json())
+    except Exception as e:  # never throw — serve stale if we have it
-    except Exception as e:  # never throw
+        if _usage_cache["data"] is not None:
+            return web.json_response(_usage_cache["data"])
         return web.json_response({"error": f"usage fetch failed: {e}"}, status=502)
 
+    _usage_cache["ts"] = now
+    _usage_cache["data"] = data
+    return web.json_response(data)
+
 
 async def pm2_status(request: web.Request) -> web.Response:
     """Return a trimmed ``pm2 jlist`` for the legal-ai background services."""

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

@@ -83,7 +83,7 @@ function SubscriptionUsagePanel({ usage }: { usage: SubscriptionUsage }) {
         <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} />
+          <UsageMeter label="שבועי · Sonnet" w={usage.seven_day_sonnet} />
         </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%