---
title: "Semantic Caching · agent orchestration · Agent Fieldbook"
url: https://agentfieldbook.org/design/orchestration/semantic-cache/
description: "Embed each incoming query; if a semantically similar past query is in the cache, return its stored answer instead of calling the model. Cache-hit queries cost nothing."
section: "Design · Orchestration"
source: Agent Fieldbook — generated from the published page
---

# Semantic Caching

Embed each incoming query; if a semantically similar past query is in the cache, return its stored answer instead of calling the model. Cache-hit queries cost nothing.

Verified against [github.com](https://github.com/zilliztech/GPTCache)

- [github.com](https://github.com/zilliztech/GPTCache)

## How it works

## At a glance

| Field | Value |
| --- | --- |
| Technique | Caching |
| Cost lever | Lower, cache hits skip the model entirely |
| Latency lever | Much lower on cache hits (ms vs seconds) |
| Quality lever | Depends on similarity threshold, too loose returns stale answers |
| Example | GPTCache (Zilliz)· Portkey semantic cache |

## When to use

Workloads with repetitive or near-duplicate queries (FAQ bots, support, docs search).

## Basis

## Verified against

- [https://github.com/zilliztech/GPTCache](https://github.com/zilliztech/GPTCache)

The corpus does not rewrite vendor documentation. Where copy is quoted verbatim it is marked as such and attributed to its source.