Record counts per time bucket (a histogram(_timestamp, …) query via _search).
Elasticsearch-compatible bulk write: POST /api/{org}/_bulk. Accepts structured actions or a
raw NDJSON body. delete and update actions are rejected client-side with
ValidationError because the server silently ignores them.
Writes records with POST /api/{org}/{stream}/_json. Give _timestamp in epoch µs (ms, ns
and RFC 3339 are auto-detected by the server); without it the server uses its own time.
Writes newline-delimited JSON with POST /api/{org}/{stream}/_multi. One bad line rejects
the whole request (ValidationError).
Writes SDK log records as OTLP/JSON (POST /api/{org}/v1/logs, header stream-name). The
server stores body, severity, severity_number, trace_id, span_id and flattened
attributes; trace flags are not stored.
Runs a SQL query: POST /api/{org}/_search?type=logs. Quote stream names with double quotes
(SELECT * FROM "default"); without ORDER BY rows come newest first. Times are converted to
microseconds (the server silently returns nothing for millisecond values).
Rows around one timestamp: GET /api/{org}/{stream}/_around. The anchor is a Date or the
exact _timestamp (µs) of a row; the anchor row is included.
Follows a stream by polling _search with a _timestamp cursor (OpenObserve has no live
tail). Every poll re-reads overlapMs before the cursor (never before start) and drops rows
it already delivered (by full-row equality), so rows that are exact duplicates within that
window are delivered once.
Distinct values with counts: GET /api/{org}/{stream}/_values?fields=a,b.
The OpenObserve logs module:
client.logs.