In-memory cache for TTS audio with size and entry limits.
// Source: https://github.com/danny-avila/librechat | License: MIT | Copyright 2026 LibreChat // Provided via justclaw.it.com; keep this notice with the code. /** Cache Storage names earlier builds persisted read-aloud audio under (the first keyed by the * message text). Audio is now held only in this tab's memory, so these are just deleted. */ const LEGACY_TTS_CACHE_NAMES = ['tts-responses', 'tts-responses-v1', 'tts-responses-v1:pending']; const MAX_ENTRIES = 50; /** Long replies make large recordings, so the cache is bounded by size as well as count. */ const MAX_BYTES = 32 * 1024 * 1024; /** Spoken audio by hashed key, oldest first; it never leaves this tab and dies with it. */ const audioByKey = new Map<string, Blob>(); let totalBytes = 0; /** Bumped on every clear, so a request tha ... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key