You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 417ce9e
Browse filesBrowse the repository at this point in the historyBrowse files
docs: retire the 350k figure the compact window no longer uses
Follow-up to the compact-window move. The number was quoted in three more places
that all still said 350k: the runtime comment describing the fallback, the
CLAUDE_CODE_AUTO_COMPACT_WINDOW default across every locale of the Claude Code
guide, and the auto-context toggle description.
The toggle description also described the wrong rule. It said ON marks "every model
above 200k tokens", which was only true while the compact window sat near that
floor — the predicate has always required a window that can host the compaction
threshold, and at 829,800 those two readings diverge.
|`ANTHROPIC_AUTH_TOKEN`| Uniquement lorsque le proxy exige une clé API ; sinon, elle n'est PAS définie, afin que votre connexion claude.ai (abonnement et connecteurs) reste active |
49
49
|`CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY`|`1` (découverte du sélecteur `/model` natif) |
50
-
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Seuil de compactage du contexte automatique (par défaut `350000`) ; injecté uniquement lorsque le contexte automatique est activé |
50
+
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Seuil de compactage du contexte automatique (par défaut `829800`) ; injecté uniquement lorsque le contexte automatique est activé |
|`ANTHROPIC_AUTH_TOKEN`| Only when the proxy requires an API key — otherwise it is NOT set, so your claude.ai login (subscription + connectors) stays active |
|`ANTHROPIC_AUTH_TOKEN`| Только когда прокси требует API-ключ — иначе переменная НЕ устанавливается, поэтому ваш вход в claude.ai (подписка + коннекторы) остаётся активным |
23
23
|`CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY`|`1` (обнаружение моделей в нативном селекторе `/model`) |
24
-
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Порог автосжатия контекста (по умолчанию `350000`); внедряется только при включённом автоконтексте |
24
+
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Порог автосжатия контекста (по умолчанию `829800`); внедряется только при включённом автоконтексте |
|`ANTHROPIC_AUTH_TOKEN`| Yalnızca proxy bir API anahtarı gerektirdiğinde — aksi takdirde AYARLANMAZ, böylece claude.ai girişiniz (abonelik + bağlayıcılar) aktif kalır |
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Otomatik bağlam sıkıştırma eşiği (varsayılan `350000`); yalnızca otomatik bağlam etkinleştirildiğinde enjekte edilir |
58
+
|`CLAUDE_CODE_AUTO_COMPACT_WINDOW`| Otomatik bağlam sıkıştırma eşiği (varsayılan `829800`); yalnızca otomatik bağlam etkinleştirildiğinde enjekte edilir |
|`ANTHROPIC_DEFAULT_HAIKU_MODEL`|`claudeCode.tierModels.haiku ?? claudeCode.smallFastModel` (isteğe bağlı; eski `ANTHROPIC_SMALL_FAST_MODEL` da geçerlidir) |
"claude.autoContextDesc": "Steuert, wie weit die 1M-Markierung geht. AN: jedes Modell über 200k Tokens (GPT-Modelle usw.) erhält eine Big-Context-Zeile. AUS: nur echte 1M-Modelle.",
1313
+
"claude.autoContextDesc": "Steuert, wie weit die 1M-Markierung geht. AN: jedes Modell, dessen Fenster den Komprimierungsschwellwert fasst, erhält eine Big-Context-Zeile. AUS: nur echte 1M-Modelle.",
1314
1314
"claude.autoContextInert": "Inaktiv, weil in der Konfigurationsdatei ein alter Kontextgrößen-Wert (maxContextTokens) steht. Dort entfernen, um es wieder zu aktivieren.",
1315
1315
"claude.autoCompactWindow": "Punkt für Auto-Zusammenfassung",
Copy file name to clipboardExpand all lines: gui/src/i18n/en.ts
+1-1Lines changed: 1 addition & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -1805,7 +1805,7 @@ export const en = {
1805
1805
"claude.fastOn": "ON",
1806
1806
"claude.fastOff": "OFF",
1807
1807
"claude.autoContext": "Use big context automatically",
1808
-
"claude.autoContextDesc": "Controls how far the 1M marking goes. ON: every model above 200k tokens (GPT models etc.) gets a big-context row. OFF: only true 1M models get one.",
1808
+
"claude.autoContextDesc": "Controls how far the 1M marking goes. ON: any model whose window can host the compaction threshold gets a big-context row. OFF: only true 1M models get one.",
1809
1809
"claude.autoContextInert": "Inactive because a legacy context-size value (maxContextTokens) exists in the config file. Remove it there to re-enable.",
0 commit comments