Advanced Kafka Client

Atrahasis Pro now speaks Kafka. A Kafka tab sits next to HTTP, GraphQL, WebSocket and SSE, and it talks to a real cluster through librdkafka, built into the application: no Java, no sidecar, nothing to install. Pro carries the whole client that ships in SE, and adds the parts a production platform needs: a Schema Registry, Avro on both the reading and the producing side, and credentials that come from your secret store instead of the tab.

The client

Everything a Kafka tab does, in Pro and SE alike.

• Every security shape. PLAINTEXT, SSL, SASL_PLAINTEXT and SASL_SSL; PLAIN, SCRAM-SHA-256, SCRAM-SHA-512 and OAUTHBEARER; TLS trust from the system store, a CA file or a pasted CA; mutual TLS with a certificate and key or a PKCS#12 keystore.
• Start anywhere. Consume from latest, earliest, a specific offset, a point in time, or the last N messages.
• A live window. Key, value, partition, offset, timestamp and headers as they arrive; filter by text and partition while the stream runs; pause, resume and clear.
• Export. Write what you read as JSON, JSON Lines or CSV.
• Full producer control. Key, value, headers, explicit or automatic partition, tombstones, acks, retries, timeout and compression (gzip, snappy, lz4, zstd).
• Topic management. Create, describe, edit and delete a topic, partitions and configuration included, without a CLI.

What makes it advanced

The parts that are Pro only.

• Schema Registry, both directions. Connect either side of a tab to a Confluent-compatible registry over HTTP. The consumer connection decodes what you read; its twin on the producer encodes what you send. Each connection has its own credentials (none, basic, bearer), TLS trust and verification level, and a PEM client identity for registries behind mutual TLS.
• Avro that reads as data. A message carrying a registry schema id is decoded with the schema the registry holds for that id, so an Avro topic reads as JSON instead of bytes. Exports carry the decoded value, not the wire bytes.
• Bring your own schema. No registry, or a topic the registry does not know? Paste up to ten Avro schemas and messages are decoded against them.
• Produce against a schema. Pick a subject and the value is encoded against the registry's schema before it leaves. A check runs first and names the field that is wrong; what was checked is what is sent, pinned to the exact schema id the check validated, so a version registered in between cannot change what your bytes claim to be.
• Registry health you can see. Each registry connection reports checking, ready or failed, with the real fault (credentials, TLS, address) instead of a bare timeout.
• Credentials from your secret store. Every security field takes {{variables}}, and in Pro a variable can resolve from the keychain or Vault. The resolved value is built for one connection attempt and dropped; nothing sensitive is written into the tab or the project.