Prompt Caching
Reusing the computed state of an unchanged prompt prefix.
Aperçu
Prompt caching stores the attention state for a long, unchanging prefix so subsequent requests skip recomputing it. It solves the cost problem that a system prompt or document set repeated across thousands of calls is processed from scratch every time. It exists because the prefix is usually the largest part of the input and the part that never changes.
Graphe des technologies
Comment Prompt Caching se rattache au reste de la pile.
Utilise et dépend de
- KV Cache technique Infrastructure Prompt Caching based on KV Cache High ✓ source
Appliqué à
- AI Agent application AI Agents Prompt Caching used for AI Agent High ✓ source
- Inference concept Infrastructure Prompt Caching used for Inference High ✓ source
- RAG technique Techniques Prompt Caching used for RAG High ✓ source
Outils
Logiciels qui implémentent Prompt Caching ou s'y intègrent.