{% extends "base.html" %} {% block title %}Translation cache · mneme{% endblock %} {% block content %}

Translation cache

Translation APIs charge per character. Multiple phrasings of the same idea reuse one paid translation. Each (source_lang, target_lang) pair gets its own namespace so a French hit can't leak into a German request.

LLM endpoint: {{ spark_url }} · model: {{ model }}

Try the same sentence twice — the second run is an exact-match (~ms). Try a paraphrase — that's a semantic-match.

Result

no translation yet

History (this session)

#LayerLatencySource → TargetTranslation
{% endblock %} {% block scripts %} {% endblock %}