[Update] bulk update

common/.config/claude/statusline.py

@@ -90,25 +90,49 @@ def term_cols():
 
 
 # --- shared transcript usage ------------------------------------------------
+def _scan_usage(text):
+    """Last `message.usage` block among these JSONL lines, or {}."""
+    usage = {}
+    for line in text.splitlines():
+        line = line.strip()
+        if not line or '"usage"' not in line:
+            continue
+        try:
+            obj = json.loads(line)
+        except Exception:
+            continue
+        msg = obj.get("message")
+        if isinstance(msg, dict) and isinstance(msg.get("usage"), dict):
+            usage = msg["usage"]
+    return usage
+
+
+TRANSCRIPT_TAIL = 1 << 20  # 1 MiB
+
+
 def last_usage(data):
-    """Most recent `message.usage` block from the transcript, or {}."""
+    """Most recent `message.usage` block from the transcript, or {}.
+
+    Reads only the tail: this runs on every status refresh and transcripts grow to
+    tens of MB, so re-parsing the whole file each time is unbounded work for a value
+    that lives at the end. Falls back to a full scan if the tail holds no usage block
+    (a long stretch of tool results can push it past the window).
+    """
     try:
         path = data.get("transcript_path")
         if not path or not os.path.exists(path):
             return {}
-        usage = {}
-        with open(path, "r", errors="ignore") as fh:
-            for line in fh:
-                line = line.strip()
-                if not line or '"usage"' not in line:
-                    continue
-                try:
-                    obj = json.loads(line)
-                except Exception:
-                    continue
-                msg = obj.get("message")
-                if isinstance(msg, dict) and isinstance(msg.get("usage"), dict):
-                    usage = msg["usage"]
+        with open(path, "rb") as fh:
+            size = fh.seek(0, os.SEEK_END)
+            start = max(0, size - TRANSCRIPT_TAIL)
+            fh.seek(start)
+            text = fh.read().decode("utf-8", "ignore")
+        if start:
+            text = text.partition("\n")[2]  # drop the partial first line
+        usage = _scan_usage(text)
+        if not usage and start:
+            with open(path, "r", errors="ignore") as fh:
+                usage = _scan_usage(fh.read())
         return usage
     except Exception:
         return {}
@@ -193,7 +217,14 @@ def git_segment(cwd, top):
             return ""
         repo = os.path.basename(top) or "repo"
         branch = _git(cwd, "rev-parse", "--abbrev-ref", "HEAD").stdout.strip() or "?"
-        dirty = bool(_git(cwd, "status", "--porcelain").stdout.strip())
+
+        # `status --porcelain` walks the whole work tree, so in a big repo it can blow
+        # the 1s timeout -- in its own try, because letting that escape dropped the repo
+        # and branch we already had: a decoration failing must not take the segment down.
+        try:
+            dirty = bool(_git(cwd, "status", "--porcelain").stdout.strip())
+        except Exception:
+            dirty = False
 
         # ahead/behind vs upstream (omitted when no upstream is configured)
         ab = ""
@@ -231,6 +262,25 @@ def model_segment(data):
     return out
 
 
+def context_window(data):
+    """Size of this session's context window, in tokens.
+
+    Claude Code resolves the `[1m]` model suffix away before handing the status line
+    its JSON -- `model.id` is always the bare API id ("claude-fable-5") -- so sniffing
+    the id for "1m" could never detect a 1M session and every one of them was measured
+    against a 200k window (5x overstated %, ⚠compact from ~16% real usage). The payload
+    carries the real number; display_name ("Opus 4.8 (1M context)") is the fallback.
+    """
+    cw = data.get("context_window")
+    if isinstance(cw, dict):
+        for key in ("context_window_size", "max_tokens", "size"):
+            val = cw.get(key)
+            if isinstance(val, (int, float)) and val > 0:
+                return int(val)
+    name = ((data.get("model") or {}).get("display_name") or "").lower()
+    return 1_000_000 if "1m" in name else 200_000
+
+
 def context_segment(data, usage):
     try:
         if not usage:
@@ -240,8 +290,7 @@ def context_segment(data, usage):
             + usage.get("cache_creation_input_tokens", 0)
             + usage.get("cache_read_input_tokens", 0)
         )
-        model_id = (data.get("model") or {}).get("id", "")
-        window = 1_000_000 if "1m" in model_id.lower() else 200_000
+        window = context_window(data)
         pct = used / window * 100 if window else 0
         col = bucket(pct, 50, 80)
         out = "📝 " + color(f"{pct:.0f}%", col) + color(f" ({used / 1000:.0f}k)", "gray", dim=True)