ndb
by ClowkThe API of Elasticsearch. The filters of MongoDB. One Go binary.
ndb runs the routes you know from Elasticsearch and the query language you know from MongoDB on plain NDJSON. One daemon, one data directory. No JVM, no cluster, no schema.
curl --unix-socket /run/ndb/ndb.sock \
http://ndb/calls/_search -d '{
"query": {
"status": { "$in": ["missed", "busy"] },
"ts": { "$gte": "now-1d/d" }
},
"sort": { "ts": -1 },
"limit": 50
}'Install
curl -fsSL https://ndb.clowk.in/install.sh | bashWhat you get
Elasticsearch routes
_search, _count, _bulk, _aggregate, _update_by_query and _delete_by_query, with index patterns like logs-2026-* and date math like now-7d/d.
MongoDB filters
$in, $gte, $regex, $elemMatch, $and and $or, updates with $set, $inc and $push, and an aggregate pipeline.
Outbox and streams
A change feed with consumer positions, and Redis-style streams with groups, retries and a dead letter.
Two write modes
SDKs write through a write-ahead log, or your app appends to its own NDJSON files and ndb follows them.
NDJSON is the truth
Indexes and columnar segments come from your data. Delete them and ndb builds them again.
Same answer everywhere
Go, Ruby, Rails and TypeScript SDKs run the same shared test cases as the daemon.
Questions
Is ndb a cluster?
No. ndb is one node with one data directory. There is no replication and no sharding.
Do I need a JVM or a schema?
No. ndb is one Go binary, and it reads plain NDJSON.
Will the HTTP API change?
The HTTP API is stable in 1.x. Routes and keys change only in a major release.