v1.5.0

Namespace deprecation (docs link)

• The "namespace" terminology is now deprecated (still backwards compatible) everywhere
  • There is a new "keyspace" alternative for each parameter name or method with "namespace" in it
  • e.g. db.useNamespace(...) still works, but now prefer the db.useKeyspace(...) alternative
• All usages of "namespace" will be removed in an upcoming major release

Other deprecations (to be removed in v2.0)

• collection.bulkWrite() is now deprecated
  • Prefer to just manually call the functions yourself
• db.collections() is also now deprecated
  • Equivalent to (await db.listCollections()).map(c => new Collection(c.name))

Misc improvements/fixes

• collection.drop() will now work properly if the collection is in a non-default keyspace
  • Before, it would attempt to drop the collection from 'default_keyspace'
• returnDocument on findOneAnd* functions are finally optional
  • Defaults to 'before' on the Data API
• updateDbNamespace was changed to immediately set the new keyspace
  • Before, it would wait for the keyspace to be created first
    • This would lead to easier race conditions, and they keyspace wouldn't be set at all if creation errored
• [Admin]CommandSucceededEvents now have a warnings field for any warnings the Data API may return
  • These may point out deprecated/incorrect practices, or any other issues that aren't strictly an error.
  • (e.g. using the now-deprecated (create|find|drop)Namespace admin commands on DSE/HCD)
• Added missing DB statuses to the DatabaseStatus enum (e.g. 'HIBERNATED' and 'HIBERNATING')
  • You can check the diff here
• Better error for if the Admin classes don't have a token set
  • (for the astra variants)
• Various error message improvements throughout the client
  • Fixed some errors being obfuscated, improved the clarity of a few, added ' (+ more <num_errors> errors)' to the end of aggregated errors
• Exposed FetchH2 and FetchNative fetchers to make it easier to wrap them with your own for customization purposes
  • For use with setting a 'custom' client
• Exposed DEFAULT_KEYSPACE constant

Internal stuff

(Too many changes to list in detail, but here are the highlights)

• MAJOR overhaul of test suite
  • Massively parallelized tests (full test suite used to take over 50 min to run.. now it barely takes 10 minutes!)
  • Majorly improved test filtering through custom filtering functionality
  • Custom test report file generated to store full errors thrown during tests
  • Test suite sets up and cleans up after itself much better now (tests much more reproducible now)
  • Custom test script CLI to make running tests super easy
  • Vectorize tests wait for providers to warm up now instead of erroring out after the first try
  • Much more info here
• Updated ESLint + config
  • Including adding rules to add all missing semicolons/commas
• shell.nix + nix-direnv
  • Setups up shell env w/ nodejs_20, jq, and the astra CLI
  • Also adds the scripts dir to PATH
• repl.sh script
  • Much easier way to quickly test out astra-db-ts functionality
  • Added -light option to build script to set up REPL quicker
    • Doesn't typecheck or do any extraneous code transformations (such as stripping comments)
• returnDocumentResponses implementation
  • Fully implemented, but not used until Data API issue is resolved

v1.4.1...v1.5.0

v1.4.1

Minor findEmbeddingProviders typing update

• Split EmbeddingProviderParameterInfo into two separate types:
  • EmbeddingProviderProviderParameterInfo
  • EmbeddingProviderModelParameterInfo
• Added .displayName and .hint to EmbeddingProviderProviderParameterInfo

v1.4.0...v1.4.1

v1.4.0

Db namespaces update

• Db namespace is now undefined by default for non-Astra backends if unset
  • Still defaults to 'default_keyspace' for Astra usage
  • Will error out if attempting to perform an operation that requires a namespace
• db.useNamespace('<namespace>')
  • Switches the current working namespace
  • Does not affect collections previously spawned from the Db using the old namespace
• dbAdmin.createNamespace('<namespace>', { updateDbNamespace: true })
  • New idiom for working with namespaces right after creating them
  • Sets the DbAdmin's parent Db's namespace to the newly created one
  • See examples/non-astra-backends for an example

findEmbeddingProviders

• dbAdmin.findEmbeddingProviders()
  • Provides in-depth, structured information about all of the available embedding providers supported on the working database
  • Output may vary slightly depending on the cloud-provider/region

EmbeddingHeadersProvider

• All embeddingApiKey parameters now take in string | EmbeddingHeadersProvider | undefined
  • Two default token providers
    • AWSEmbeddingHeadersProvider for bedrock vectorization
    • EmbeddingAPIKeyHeaderProvider for symmetry (unit implementation)
  • New EmbeddingHeadersProvider subclasses may be created, additionally providing "token refreshing logic" as necessary
  • Fully backwards compatible—string API keys can still be passed in like normal

Misc improvements

• Better private field encapsulation for the main classes (Collection, Db, etc.)
  • private get _httpClient() escape hatch present if necessary
• Authentication failure errors are no longer intercepted/wrapped by the client
  • Now shows the raw error straight from the Data API
  • Root cause of error should be a bit less obfuscated now
• TokenProvider doesn't have to return a Promise anymore
  • Like the new EmbeddingHeadersProvider, it can now return string | nullish | Promise<string | nullish>
  • Avoids the slight performance penalty of awaiting a Promise if one isn't returned anymore
• Further documentation fixes/tweaks (especially around vectorize)

v1.3.0...v1.4.0

v1.3.0

Non-Astra backend support

• New environment parameter on DataAPIClient and db.admin()
  • Accepts 'astra' | 'dse' | 'hcd' | 'cassandra' | 'other'
  • Defaults to 'astra'
• All token parameters now take in string | TokenProvider | undefined
  • Two default token providers
    • UsernamePasswordTokenProvider for DSE, HCD
    • StaticTokenProvider for symmetry (unit implementation)
  • New TokenProvider subclasses may be created, additionally providing "token refreshing logic" as necessary
  • Tokens are now "optional" (i.e. you won't see a "missing token error" until the server throws one on an API request)
• New DataAPIDbAdmin for namespace CRUD on non-Astra instances
• New example in examples/ for running on non-astra

Cursor sort vectors

• New includeSortVector parameter on cursor
  • False by default
  • If set to true, fetches the $vector used in the cursor's vector sort (if applicable)
• New await cursor.getSortVector() to get the sort vector if the above is set

Deprecations

• vector/vectorize keyword parameters deprecated on document operations
  • Use raw $vector[ize] in documents for inserts
  • Use sort: { $vector[ize]: ... } for sorts
• deleteAll() deprecated in favor of deleteMany({})
  • deleteMany({}) actually works now

Misc improvements

• Package size made 33% smaller
• collection.distinct() made much faster when "distincting" on objects
• Some documentation fixes/tweaks (especially around vectorize)
• Few improved error messages

v1.2.1...v1.3.0

v1.2.1

Extended "non-standard" environment support

• Client will now try using fetch-h2 (by default) regardless of runtime/environment
  • Before, it only did this if the runtime were standard node
• Client will now default to fetch if fetch-h2 doesn't work by default
  • This will only happen if httpOptions is not set at all in DataAPIClientOptions
• Client accepts a new client: 'custom' type in httpOptions, allowing you to create your own component to make API calls for the client
• You can now manually pass in the fetch-h2 module to use yourself if astra-db-ts is having issues importing the default one when your code is minified (see the README)
• New examples/ folder which has some common examples of running astra-db-ts in a "non-standard" environment
  • Also updated the README to cover more cases as well

Faster insertMany

• Default batch size of documents was updated to 50
  • Data API now supports up to 100, >50 generally gives diminishing returns, especially depending on document size
  • You can still do your own tweaking with the chunkSize parameter to find what's fastest for you if necessary

v1.2.0

Added support for vectorize

• createCollection was updated w/ vectorize options

Non-standard runtime support

• Non-node runtimes will now trigger default use of the native fetch API instead of fetch-h2 which gives access to extended HTTP/2 and HTTP/1.1 support
• Extended HTTP/1.1 options

v1.1.0

• Updated SomeId to allow null (#35) f0da2f9
• added support for AstraAdmin.dbInfo (#34) 9b32edb
• Added extended http1 options (in DataAPIClient) (#33) c73f0da
• roughly added estimatedDocumentCount (#32) f6a37e9

v1.0.2...v1.1.0

v1.0.2

• Client using support + minor typing bug fix (#31) 8ea415c
• Minor ts-doc cleanup (#30) 6b0ead3

v1.0.1...v1.0.2

v1.0.1

1.0.1

• Updated createCollection documentation to mention checkExists
• Fixed ragstack detection formatting slightly to keep in-line w/ codebase
• Added couple replaceOne tests
• Marked some things as internal that weren't already
• '*' support for projections

v1.0.0...v1.0.1

v1.0.0

• Merge pull request #25 from datastax/signature-cleanup fe82a3e
• updated readme examples slightly a88b5e8
• some last-minute doc updates + error management fixes 2a3ac88
• more documentation & last minute config changes e389fa5

v1.0.0-alpha.2...v1.0.0