// the find
elastic/elasticsearch
Free and Open Source, Distributed, RESTful Search Engine
Elasticsearch is the dominant distributed search and analytics engine, built on Lucene, handling full-text search, vector search, aggregations, and time-series data at scale. It's production-grade infrastructure used by companies with billions of documents — not a weekend project or a replacement for a simple Postgres FTS setup. If you're evaluating it for a real workload, the question isn't 'is it good?' but 'do you actually need it?'
The BM25 + dense vector hybrid search story has matured significantly — you can do kNN retrieval alongside traditional full-text in a single query without bolting two systems together. The aggregation engine handles cardinality estimation, percentiles, and time-bucketing across massive datasets without pulling data to the client. ES|QL (the new query language) is a genuine improvement over the JSON query DSL — actual pipelines with typed columns, not nested JSON objects that take 20 minutes to debug. Operational tooling around cluster health, index lifecycle management, and shard allocation is battle-tested; you can automate hot-warm-cold tiering without writing custom scripts.
The licensing situation is messy and has been for years — the switch from Apache 2.0 to SSPL in 2021 created a fork (OpenSearch) and ongoing confusion about what 'free and open' actually means; verify your use case against the current Elastic license before committing. Operational complexity is high for self-hosted deployments: you're managing JVM heap tuning, shard count mistakes compound over time, and a misconfigured cluster can fall into split-brain or out-of-memory states that are painful to recover from. Elasticsearch is not a database and using it as one (consistency guarantees, transactions, relational integrity) will burn you — it's eventually consistent by design and that surprises teams who treat it like Postgres. The managed cloud offering (Elastic Cloud) is expensive relative to alternatives like OpenSearch Service or a self-hosted cluster on commodity hardware.