Streamhouse – all-in-one event streaming for startups

Name: Streamhouse – all-in-one event streaming for startups
Availability: InStock
Author: gbram

by gbram·Mar 24, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidBig BrainBold Bet

Stateless agents with S3 storage cuts Kafka+Flink costs by 92%, but confluent-kafka wire protocol still needs fixes.

Strengths

•S3-native storage at $0.023/GB instead of 3x-replicated EBS at $0.10+/GB
•Single Rust binary replaces Kafka, Flink, schema registry, and ZooKeeper
•Built-in DataFusion SQL engine handles windows, JOINs, and continuous queries

Weaknesses

•Confluent-kafka clients need wire protocol fixes before production use
•Author admits more soak testing needed for massive sustained loads

Post Description

Hey HN, I'm building StreamHouse — an event streaming platform where you get Kafka-style streaming, a schema registry, SQL queries, and pipeline sinks without managing five separate services.

Storage is S3 instead of broker disks. Agents are stateless — if one dies, another picks up. Metadata lives in Postgres. You can use the managed platform at streamhouse.app, or self-host the whole thing with Docker Compose. Either way it's one service instead of five

What works today:

- CLI + Python/TypeScript SDKs - REST and gRPC APIs (~2,600 msg/s on one EC2 instance) - Pipelines that sink to Postgres and ClickHouse - Schema registry (Avro, Protobuf, JSON Schema) - SQL over streaming data - Multi-tenant with org isolation and API key auth

What doesn't work yet: confluent-kafka clients need wire protocol fixes, and I need more soak testing before I'd trust it for massive sustained loads.

https://github.com/gbram1/streamhouse