// the find
splitgraph/seafowl
Analytical database for data-driven Web applications 🪶
Seafowl is an analytical database built on DataFusion and Delta Lake, designed specifically for querying data directly from the browser over HTTP. The core idea is that SELECT queries are GET requests with deterministic URLs, so you can put a CDN in front and have query results cached at the edge — which is a genuinely different architectural bet than most databases make. It's for developers building read-heavy data visualizations or notebooks where the query layer needs to be publicly cacheable.
The HTTP cache semantics for queries is the actual clever part: GET /q?query=SELECT... means Cloudflare or Varnish caches the result, and you get sub-millisecond p99 on repeated queries without touching the database. WebAssembly UDFs are a real differentiator — you can extend the query engine from the browser without touching server code. DataFusion underneath means the query engine is not a toy; it handles real columnar workloads and benefits from upstream improvements. The single-binary deploy story with optional S3-backed storage and scale-to-zero is well thought out for the serverless/edge deployment model it targets.
514 stars and 18 forks after years of existence signals that the market hasn't materialized — the use case (browser-direct SQL queries) is narrow enough that most teams reaching for an analytics database want something that connects to their BI tool, not a CDN-friendly HTTP endpoint. The long-term roadmap items (PostgreSQL-compatible wire protocol, JSON functions) are still listed as future work, and the last push was February 2025, which is not a good sign for active maintenance. Write path is awkward — you're either uploading files or using DML over HTTP with an API key, which doesn't compose well with existing data pipelines. The split between SQLite (single-node) and Postgres (metadata) backends adds operational complexity that the documentation doesn't fully acknowledge.