corpus_sweep capture: budget for the marked ranges still filling

Only the bytes on disk were checked: at 296.6 GB with 2,300 ranges complete it
marked 100 more (~310 GB). The marked ranges not complete count at the stored
ranges' mean size, and no more are marked than fit under 300 GB.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

deploy/corpus_sweep.py

@@ -55,8 +55,16 @@ def corpus_bytes():
 
 def capture(n):
     used = corpus_bytes()
-    if used >= MAX_BYTES:
-        print(f'corpus at {used / 1e9:.0f} GB of {MAX_BYTES / 1e9:.0f}: nothing more marked')
+    # Marked ranges not complete yet take room too: at the stored ranges' mean size, less what they hold already
+    # (2026-10-13: 2,300 complete at 296.6 GB and 100 more marked would have made ~310 GB).
+    complete = sum(1 for _ in complete_ranges())
+    marked = int(psql('SELECT count(*) FROM ranges WHERE capture;')[0][0])
+    mean = used / complete if complete else 0
+    pending = max(0, marked - complete) * mean
+    room = MAX_BYTES - max(used, complete * mean + pending)
+    n = min(n, int(room // mean)) if mean else n
+    if n <= 0:
+        print(f'corpus at {used / 1e9:.1f} GB of {MAX_BYTES / 1e9:.0f} (with {marked - complete} ranges still filling, ~{(complete * mean + pending) / 1e9:.0f} GB): nothing more marked')
         return
     top = [line.split('\t')[1] for line in open(os.path.join(ROOT, 'groups.tsv')).read().splitlines()[:TOP]]
     names = ','.join("'" + g.replace("'", "''") + "'" for g in top)