Caching data calls
Cache repeat web and data calls when freshness is less important than speed and cost.
Web search, scraping, enrichment, and verification calls are often metered. Marmot can cache provider responses on disk so repeated calls with the same input return quickly and skip the network.
Caching is disabled by default. Enable it per provider when repeated inputs are likely and the data does not need to be fresh every time.
When to use it
Use caching for:
- repeatable search or lookup scripts
- local demos
- enrichment or verification batches that may be rerun
- agent workflows where a search or scrape may be retried
Skip caching when every call must be fresh, or when the input is unlikely to repeat.
Enable it
marmot config set providers.tavily.cache.enabled true
marmot config set providers.tavily.cache.ttlDays 14Caching applies to sync web and data verbs: search, scrape, answer, map, enrich, lookup, and verify.
It does not cache AI verbs, and it does not cache async task results from research, crawl, or findall.
Force freshness
Use per-call flags when one call needs different cache behavior.
| Flag | What it does |
|---|---|
--no-cache | Skip cache read and write for this call. |
--refresh | Skip cache read, make a fresh provider call, and overwrite the cached entry. |
For storage paths, invalidation commands, and exact cache-key behavior, see Caching reference.