warpstreamlabs/bento v1.22.0 on GitHub

1.22.0 - 2026-10-09

statsd metric new fields: send_queue_capacity, buf_pool_capacity & send_loop_count @triddell
credentials object to GCP components with fields impersonate_service_account & impersonate_delegates @NguyenThanhDat741862
kafka_franz components now include GSSAPI (kerberos) in available sasl.mechanism options @kmetaxas
redis_scan input can now handle hash values with additional Bento config fields: data_type, value_format, scan_count, hash_scan_count @maxtheaxe
mqtt_v5 experimental input & output components enabling Bento to send/receive MQTT 5 messages @ifnesi
prefix field added to aws_s3 and gcp_cloud_storage caches enabling multiple caches to share a bucket, scoping all operations including key listings @ecordell
aws_sqs input field visibility_timeout (default 30s); set to 0 to use the queue's own timeout @ReguiguiMohamed
DATE & TIMESTAMP_MILLIS type options added to parquet_encode processor @triddell
parse_bytes bloblang method for parsing humanised byte sizes (such as 10MB or 20MiB) into an integer number of bytes @SJ1397
prefix_metadata_keys added to http_server input enables prefixing metadata with the source i.e. header_, path_ @kshivam4781
azure_queue_storage input field delete_message, defaulting to true, which can be set to false to leave acknowledged messages on the queue @istairbn

YAML merge keys <<: *ref_field are now resolved when running streams, aligning behaviour with the lint subcommand @youdie006
mongodb cache operator exists correctly handling mongo.ErrNoDocuments & non-bytes return type @gregfurman
gcp_bigquery_write_api output's batching.byte_size lint rule triggered at 1 MB while claiming a 10 MB limit, incorrectly flagging valid batch configurations well under BigQuery's actual AppendRows request size limit @SJ1397
aws_kinesis input stop duplicate EFO shard consumers accumulating under backpressure @matus-tomlein
fsevent reconnect on watcher error instead of stalling @henrikschristensen
azure_queue_storage input deleted messages on nack, losing them; nacked messages are now left on the queue and redelivered once their visibility timeout expires @istairbn
kafka_franz sends sasl extensions when oauth2 is enabled @bgajdzis-pentaleap
Inputs that automatically retry nacked messages could stay blocked on shutdown when a read was cancelled during the retry backoff @ReguiguiMohamed
aws_sqs input looked for the queue's VisibilityTimeout on each message, which ReceiveMessage never returns, so in-flight messages could not be refreshed with anything but 30s @ReguiguiMohamed

nats_kv output & cache migrated to new jetstream package & connection retry logic (increased frequency of retries & indefinite number of retries) @henrikschristensen

Full Changelog: v1.21.2...v1.22.0