$ s2b --about
Event and OpenTelemetry ingestion into your own bucket
POST JSON or point an OTLP exporter at api.s2b.tznc.dev. Send2Bucket answers once the batch is stored durably, then writes gzip NDJSON to your S3 or GCS bucket, or appends to your Iceberg or DuckLake table.
- [ok] 200 only after the batch is stored durably
- [ok] JSON, OTLP/HTTP and OTLP/gRPC, one API key
- [ok] S3, GCS, any S3 endpoint, Iceberg, DuckLake
- [ok] free: 100 MB a month · paid: $20 with 50 GB
curl https://api.s2b.tznc.dev/v1/events \
-H "Authorization: Bearer $S2B_API_KEY" \
-H "Content-Type: application/json" \
-d '[
{"event":"order.paid","order_id":"ord_20931","cents":4200},
{"event":"email.sent","order_id":"ord_20931"}
]'{"accepted":2,"bytes":106}$ aws s3 cp s3://analytics-raw/raw/s2b/stream=3f6c2a9e-8b1d-4e7a-9c55-0d2b7e41a8f3/date=2026-11-19/hour=03/dlv_01jd3kb7m4.ndjson.gz - | gunzip | head -1
{"v":1,"delivery_batch_id":"dlv_01jd3kb7m4","batch_id":"bat_01jd3k9qx2","received_at":"2026-11-19T03:04:05.123456Z","signal":"event","data":{"event":"order.paid","order_id":"ord_20931","cents":4200}}how it works
One request in, one gzip NDJSON object out.
client ──POST /v1/events──▶ s2b ──60 s or 16 MiB──▶ your bucket / table
│
└─▶ 200 OK- send JSON to /v1/events or OTLP to /v1/logs, /v1/traces, /v1/metrics or gRPC on :443.
- store batches events durably before answering 200.
- deliver writes the batch to your destination as one object or table commit.
{
"accepted": 2,
"bytes": 106
}api
Authorization: Bearer s2b_ingest_… on every request. Items up to 8 KiB, requests up to 16 MiB.
| method | path | what |
|---|---|---|
| POST | /v1/events | one JSON object or an array; gzip or identity |
| POST | /v1/logs /v1/traces /v1/metrics | OTLP/HTTP, application/x-protobuf |
| gRPC | api.s2b.tznc.dev:443 | OTLP Export RPCs for logs, traces, metrics |
| status | error |
|---|---|
| 400 | invalid_json · invalid_item · item_too_large · invalid_encoding |
| 401 | unauthorized |
| 409 | destination_not_ready |
| 413 | body_too_large (16 MiB decompressed) |
| 415 | unsupported_encoding |
| 429 | quota_exceeded {plan, remaining_bytes, resets_at} |
| 503 | buffer_full · destination_backlog, Retry-After: 30 |
buckets and tables you own
One destination per ingestion stream, in your account. Buckets get gzip NDJSON; tables get rows by catalog commit.
| kind | service | you provide |
|---|---|---|
s3 | Amazon S3 | IAM role with your external ID |
gcs | Google Cloud Storage | grant our service account |
s3_compatible | R2, MinIO, B2, Wasabi… | endpoint, region, access key |
ducklake | DuckLake | Postgres catalog DSN, data path |
iceberg_rest | Apache Iceberg REST | catalog URI, warehouse, token |
s3_tables | Amazon S3 Tables | table bucket ARN, IAM role |
supabase_analytics | Supabase Analytics Buckets | project ref, catalog token, S3 keys |
# buckets
<prefix>/s2b/stream=<stream_id>/
date=YYYY-MM-DD/hour=HH/
<delivery_batch_id>.ndjson.gz
# tables
signal_type VARCHAR
delivery_batch_id VARCHAR
ingested_at TIMESTAMPTZ
payload VARCHAR -- envelope JSONquery with what you already run
No query API and no lock-in: point your engine at the bucket or catalog. Each guide answers the same questions on the same data.
| engine | reads |
|---|---|
| DuckDB | objects, Iceberg, DuckLake in place |
| Trino | objects and Iceberg; also Starburst |
| BigQuery | GCS objects, external or loaded |
| Snowflake | external table, Snowpipe, Iceberg catalog |
| Athena | objects with partition projection, S3 Tables |
| ClickHouse | objects in place, or loaded into MergeTree |
| PostgreSQL | loaded into jsonb, or pg_duckdb |
| MySQL | loaded into JSON |
| Python | Polars, PyIceberg, pandas |
| Hex | notebook over any of the above |
| MotherDuck | shared DuckDB views |
SELECT data->>'$.event' AS event, count(*) AS events
FROM read_json('s3://<bucket>/<prefix>/s2b/stream=*/date=*/hour=*/*.ndjson.gz',
hive_partitioning = true)
WHERE signal = 'event' AND date = current_date
GROUP BY ALL ORDER BY events DESC;delivery semantics
- [ok] at least once: every acknowledged record is delivered
- [ok] same object key and bytes on every retry
- [ok] our retries to your destination never add usage
- [!!] no global ordering: order by received_at or your own timestamps
- [!!] no request deduplication: a client retry is stored and billed again
prices
Metered on ingested, uncompressed bytes. 1 GB = 1,000,000,000 bytes.
| plan | price | allowance |
|---|---|---|
| free | $0 | 100 MB a month per organization, hard cap. A request past it gets a 429 with the reset time. |
| trial | $0 for 7 days | 10 GB over 7 days, once per organization, no card. Then back to Free. |
| paid | $20 a month | 50 GB included each billing period, then the per-GB rates in the table. |
| paid usage in a billing period | per GB |
|---|---|
| First 50 GB | included |
| Above 50 GB through 1 TB | $0.500 |
| Next 9 TB | $0.450 |
| Next 90 TB | $0.400 |
| Above 100 TB | $0.350 |
faq
Where is my data stored?
In a bucket or table in your own account: Amazon S3, Google Cloud Storage, an S3-compatible bucket, an Iceberg REST catalog, S3 Tables, a Supabase Analytics Bucket or DuckLake. Send2Bucket holds payloads only until they are delivered.
What does a 200 response mean?
The batch is stored durably and will be delivered at least once. Batches are delivered to your destination every 60 seconds or at 16 MiB, whichever comes first.
Do I need an OpenTelemetry Collector?
No. OpenTelemetry SDKs and exporters send OTLP over HTTP or gRPC straight to Send2Bucket with an API key. An existing Collector can forward to it with the otlphttp exporter.
How much does Send2Bucket cost?
Free covers 100 MB a month. Paid is $20 a month with 50 GB included, then $0.500 per GB, falling with volume to $0.350 per GB. Storage and query fees are billed by your cloud provider.
How do I query the delivered data?
With the engine you already use. The docs show DuckDB, Athena, BigQuery, Snowflake, Trino, ClickHouse, Spark, PostgreSQL and others reading the same objects and tables.
What happens to my data if I stop using Send2Bucket?
Delivered data stays in your bucket or table. Disconnecting a stream stops new deliveries and removes nothing you own.