Numia Docs

What's Numia's Data Warehouse?

Migration in progress

BQ
LegacyBigQuery
OD
New homeObsessionDB

Numia’s Data Warehouse historically ran on Google BigQuery. Both stacks are live today, but new datasets, chains, and product integrations land on ObsessionDB first. If you currently consume Numia data via BigQuery, reach out to your contact so we can plan the cutover.

Numia’s Data Warehouse is a real-time streaming of on-chain data into ObsessionDB, our ClickHouse-based analytical database. ClickHouse gives us sub-second analytical queries over the full history of every chain we index, with predictable per-cluster pricing instead of per-byte-scanned billing.

We can index your L1 or L2, or build models on top of existing chains, and make them available to you and your community over standard SQL.

Blockchain Nodes

Cosmos HubCosmos Hub
CelestiaCelestia
SolanaSolana

Indexing Pipeline

Numia Ingestion

Streams blocks, decodes txs, derives events & messages

Data Warehouse

ObsessionDB

ClickHouse cluster — raw + derived tables per chain

Access Layer

Numia SQL · APIs · AI

Direct SQL, REST/GraphQL, Lenses, NumiaAI

Why ClickHouse?

  • Real-time ingestion at block speed. Materialized views derive events and messages from raw block-results as soon as they’re written — no batch latency.
  • Columnar storage + late-materialization. Wide, JSON-heavy block payloads compress aggressively, and only the columns referenced by your query are read from disk.
  • Predictable cost model. You pay for the cluster, not for every exploration. There’s no per-query bytes-scanned bill.
  • ReplacingMergeTree semantics. Re-ingestion of a block is idempotent — late corrections deduplicate automatically using ingestion_timestamp (raw) or processed_at (derived) as the version column.
  • Standard SQL with rich JSON functions. JSONExtractArrayRaw, JSON_VALUE, and ARRAY JOIN let us decode Cosmos SDK payloads directly inside the engine — no external parsing layer.
  • Native client ecosystem. Standard ClickHouse drivers exist for Python, Node.js, Go, Rust, JDBC, and the HTTP interface, so you can plug ObsessionDB into your existing stack.

Deliver to any destination

ObsessionDB is our analytical core, but it’s not where the data has to live for you. Our pipeline is decoupled from its storage target — we work with each customer to land Numia’s on-chain data wherever their stack already is:

  • Snowflake
  • Databricks
  • BigQuery
  • PostgreSQL
  • MongoDB
  • …or any other warehouse / database you operate.

This is how we already serve institutional customers such as Dune and Token Terminal. If you need Numia data piped into a specific destination, reach out and we’ll scope the integration with you.