Does the new effects API cache apply only if the data is written to TSV?

[PaulRBerg]

What I mean is this - suppose you start with two rows in your TSV file called ETH-USD.tsv:

id  output
2025-08-20	4074.495786100515
2025-08-21	4330.491598768078

Then, the indexer continues to run and as time passes, there are more dates to obtain price data for.

Suppose you have two events on 2025-08-22 that both use the same effect to fetch the price data for that day. Will the 2nd run of the effect load the data from the cache?

If not, this means two external queries will be made for the same data. Right? And the only way to cache the data is to update the TSV file with the new data, and restart the indexer. Right?

@DZakh

[DZakh]

Suppose you have two events on 2025-08-22 that both use the same effect to fetch the price data for that day. Will the 2nd run of the effect load the data from the cache?

From cache. If they are in the same processing batch, the call will be deduplicated using an in-memory cache. If they are in different batches, we'll try to load the cache from the DB first and if there's no data, execute the effect function.

The TSV file is only for populating initial cache, the actual source of truth is the database.

[PaulRBerg]

Nice thank you for the quick answer!

I also saw the Persistence section in the docs:

To persist effect results, you can set the cache option to true when creating the effect. This will save the effect results in the database and reuse them in future indexer runs.

Every effect cache creates a new table in the database envio_effect_${effectName}. You can see it and query in Hasura console with admin secret.