Neynar SQL Playground

Retired September 14, 2026. Historical reference for the hosted SQL playground, and schema notes that still apply to our other data products.

Neynar SQL Playground has been retired

The hosted SQL playground was shut down on September 14, 2026 and is no longer available. For Farcaster data at scale, use one of our other data products — Parquet exports or the indexer service. Reach out on Slack if you need help migrating.

This page is kept for historical reference. The playground itself is gone, but the schema notes below still apply to Parquet exports and the indexer service, so they remain useful when porting old queries.

What it was

The playground was a hosted Redash instance pointed at a Postgres database of Farcaster data. It offered ad-hoc SQL queries in the browser, shareable dashboards, and SQL-over-API using a Redash API key separate from the read and write API keys.

The schema was read out of the database directly with:

SQL

SELECT table_name, column_name, data_type, is_nullable
FROM information_schema.columns
WHERE table_schema = 'public'
ORDER BY table_name, ordinal_position;

A snapshot of the schema as of Nov 21, 2024 is archived here.

Porting old queries

The table shapes below are not unique to the playground — the same casts, links and reactions tables ship in Parquet exports and through the indexer service. If you are moving queries off the playground, these notes still apply.

Data is more raw than our APIs, and you will likely need to join tables. If you are building a client rather than doing analysis, prefer the APIs over SQL.

1. Follows

links table has follower <> follow data:

• fid → target_fid row means fid follows target_fid

2. Reactions

• reaction_type 1 is “like” and 2 is “recast” in the reactions table
• hash in the reactions table is the “reaction hash” and target_hash is the hash of the cast that received the reaction

3. hex <> bytea

The Redash UI converted bytea data to hex for display, but queries had to do the conversion themselves — as they still do against Parquet and the indexer service:

• bytea to hex

SQL

select ENCODE(hash, 'hex') as hash from casts
limit 1

• hex to bytea

SQL

select * from casts where hash = DECODE('hex_hash_without_0x', 'hex')

(swap hex_hash_without_0x with the actual cast hash minus the `0x)