v1.3.0

1.3.0

Opt-in connection pooling. Additive — upgrading needs no code changes, and
behavior is unchanged until you enable it.

Added

• Opt-in connection pooling. Every request previously opened a TCP
  connection, sent one request, read one response, and closed. Reuse removes the
  handshake from every call after the first:

  import { Engine, closeAllPools } from 'montycat';
  
  const engine = new Engine({
    host: '127.0.0.1', port: 21210, username: 'user', password: 'password',
    store: 'Company',
    pool: {},                   // the only new field; omit for today's behavior
  });
  
  Sales.connectEngine(engine);
  await Sales.insertValue({ value: newSale });   // unchanged
  
  closeAllPools();              // before exit

  Pools live in a module-level registry keyed by (host, port, useTls), so every
  keyspace class pointing at one server shares a single pool. connectEngine is
  Object.assign, which would have propagated a pool reference by accident; the
  config is now copied explicitly instead, and connectEngine's parameter is no
  longer typed any. useTls is part of the key because a plaintext and a TLS
  connection to one address are not interchangeable.

  Disabled by default: an idle pooled connection still holds one of the engine's
  connection permits, so the bound is conservative (maxIdle: 8,
  idleTimeoutMs: 30000). Subscriptions are never pooled. A connection is held
  exclusively for one request/response, so concurrent calls each get their own
  rather than interleaving writes on one socket.

  Exported closeAllPools and the PoolConfig type from the package root.

• Precomputed vectors. Vectors produced elsewhere — another model, a batch
  pipeline, an existing embedding store — can now be supplied directly, and the
  server skips embedding entirely. Requires a Montycat Semantic server 1.3.0 or
  newer.

  Writes take an optional vector, applied after the write succeeds:

  await Docs.insertValue({
    value: { text: 'a document' },
    vector: myEmbedding,          // number[], omit for server-side embedding
  });

  Available on insertValue, insertCustomKeyValue, and updateValue for both
  in-memory and persistent keyspaces. insertBulk takes vectors: number[][],
  paired with bulk by position; updateBulk takes vectors for numeric
  keys and customVectors for custom keys.

  Search takes an optional query vector, which bypasses text embedding. The
  query string may be empty when one is supplied:

  await Docs.semanticSearchGetValues({ query: '', vector: myQueryEmbedding });

  Available on semanticSearchGetKeys, semanticSearchGetKeysWhere,
  semanticSearchGetValues, and semanticSearchGetValuesWhere.

  Dimensions must match the keyspace's enrolled model; the server validates
  before anything reaches the index. A supplied vector is not overwritten by
  background embedding — a later ordinary write to the same item clears that
  protection and re-embeds from text.

Changed

• The response reader no longer re-splits an accumulating string. It
  previously did response += data.toString() followed by response.split("\n")
  on every data event, which is O(n²) on a large response. Pooled connections
  use a Buffer accumulator that scans only newly-arrived bytes for the
  delimiter.

  The retained-remainder behaviour that made the old reader pooling-safe is
  preserved, and now lives with the connection rather than in a per-request
  closure — a response's trailing bytes belong to the next response, and
  discarding them when the call returned would have corrupted it.

• The per-request 120s timeout no longer rides on socket.setTimeout, so it
  cannot fire against a pooled socket sitting idle between requests.

1.2.2

Adds a way to read the server's real semantic configuration, and a safe way to
change an enrolled keyspace's embedding model. Additive — upgrading from 1.2.1
requires no code changes.

Added

• Engine.getSemanticStatus({ store, keyspace }) returns the server's actual
  semantic settings rather than what the caller assumed: the DB-wide switch and
  default model, plus each enrolled keyspace's model, dimensions, field,
  storage type, and whether a backfill is still pending.
• Engine.reembedSemanticSearch({ model, field, store, keyspace }) atomically
  drops one keyspace's vectors, records the new configuration, and starts a
  complete backfill. It reports the previous model alongside the new one, so a
  caller can confirm what it replaced.

Changed

• Documented that enableSemanticSearch leaves an already-enrolled keyspace
  alone. It was never a way to switch models; reembedSemanticSearch is.
  Behavior is unchanged — only the documentation was misleading.
• Corrected the disableSemanticSearch docs: dropVectors is not "required
  before switching to a different embedding model". Use
  reembedSemanticSearch, which does not leave the keyspace unsearchable in
  between.

1.2.1

Documentation and CI only — no library code changed, so upgrading from 1.2.0 is
optional.

Added

• README sections for behavior that was previously undocumented: response shape
  ({status, payload, error} and u128 keys arriving as strings), real-time
  subscriptions with the stop() handle and the port + 1 subscription port,
  TLS via useTls, and owner/access management with createOwner, grantTo,
  revokeFrom, and ValidPermissions.
• ci.yml workflow building and running the test suite on Linux and macOS
  against Node 18, 20, 22, and 24.
• Changelog link in the README.

Changed

• The publish workflow now runs npm test before publishing to npm.