v0.4.0

Highlight: Schema Registry headers survive jq

kite consume --json now always prints valid UTF-8 JSON. A key, value, or header value that isn't valid UTF-8 now goes into a sibling _b64 field (standard padded base64). Before, kite wrote those raw bytes and jq and other JSON tools silently replaced them with U+FFFD. This covers Confluent Schema Registry's binary __value_schema_id header (0x01 + 16-byte GUID) as well as Avro/Protobuf payloads.

{"key":"o-1","headers":[{"key":"__value_schema_id","value_b64":"AW2jNtjx0w+YTEfQPuihShI="}],"value":"{\"order_id\":1}"}

• kite produce --json accepts key_b64, value_b64, and array-header value_b64. So kite consume --json | jq ... | kite produce --json copies schema-ID headers and binary payloads byte for byte.
• Text output is unchanged: valid UTF-8 stays in key, value, and header value.
• The bytes on Kafka don't change, so Confluent serializers and deserializers still see the same binary schema IDs.
• Invalid base64, or a record with both value and value_b64, is rejected with the line number.

Behavior change: for records whose bytes aren't UTF-8, scripts now find .value_b64 (or .key_b64) instead of .value. Decode with base64 -d, e.g. jq -r .value_b64 | base64 -d. The header object form on produce ({"h":"v"}) is still string-only; use the array form for binary headers.

Also

• kite update finds the latest release through the GitHub releases/latest redirect instead of api.github.com, so it no longer runs into the 60 requests/hour API limit.

What's Changed

• kite update: resolve the latest tag via the releases/latest redirect, not api.github.com by @devin-ai-integration[bot] in #88
• --json: base64 fields that aren't valid UTF-8 (key_b64/value_b64/header value_b64) so output is jq-safe by @devin-ai-integration[bot] in #89
• release prep: bump version to 0.4.0, point install examples at v0.4.0, add release notes by @devin-ai-integration[bot] in #90

Full Changelog: v0.3.1...v0.4.0