Streams

Datasets

• blocks, logs, traces, the raw chain surfaces.
• usdc.transfers, the de-duplicated feed with Memo and CCTP attached.
• certificates, one commit certificate per block, for teams that archive settlement evidence.

Sinks

Postgres, Snowflake, BigQuery, ClickHouse, S3, Kafka and plain webhook. Filtering happens upstream, so a stream watching twelve addresses does not ship the whole chain into your warehouse and bill you for the storage.

POST /v1/$KEY/streams{  "dataset": "usdc.transfers",  "range": { "from": 5000000, "to": "latest" },  "filter": { "addresses": ["0x2c…"] },  "sink": { "type": "postgres", "dsn": "…", "table": "arc_usdc" },  "delivery": "exactly-once"}

The gap report

Every stream can produce a signed statement that no block in a range was skipped. This is the artefact to attach to a reconciliation, and it is the reason a stream can be trusted as a ledger source rather than as a cache.

GET /v1/$KEY/streams/str_8841/gap-report {  "range": { "from": 5000000, "to": 5185952 },  "blocksExpected": 185953,  "blocksDelivered": 185953,  "missing": [],  "signature": "0x…"}

Throughput

At ~0.5s blocks a full-chain stream is a sustained write, not a trickle. Sinks are backpressured and a slow consumer lags rather than losing data. Lag is visible per stream in the dashboard and available over the API for alerting.