Comment by dakolli

When the prefix matches a request sent to the same Providor. The thing is the TTL is different for each provider, some cache for 5 minutes some cache for 1hr. Its ideal to only use one provider per agent session / and per model with the best cache hit % if you care about costs.