NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #4999 most downloaded on PyPI
A framework for writing Airbyte Connectors.
Last release 2 days ago
02 Oct 2026
Ships on a steady schedule
a new release about every 9 days
Nearly every release is documented
notes for 60 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
5 years old
1285 releases · first in 2021
One column per quarter.
fix connector builder output serialization
fix connector builder output serialization
add transform_record() to class DefaultFileBasedStream
add transform_record() to class DefaultFileBasedStream
add python-snappy to file-based dependencies
add python-snappy to file-based dependencies
concurrent-cdk: add cursor definition based on sync mode to ConcurrentSourceAdapter
concurrent-cdk: add cursor definition based on sync mode to ConcurrentSourceAdapter
Decouple low-code request_parameter_provider from cursor, add optional cursor_granularity to ConcurrentCursor
Decouple low-code request_parameter_provider from cursor, add optional cursor_granularity to ConcurrentCursor
Fix pandas missing dependency
Fix pandas missing dependency
Bug fix: Return a connection status failure on an expected check failure
Bug fix: Return a connection status failure on an expected check failure
Declarative async job components
Declarative async job components
add migration of global stream_state to per_partition format
add migration of global stream_state to per_partition format
Connector builder: add flag to disable cache
Connector builder: add flag to disable cache
Fix error in incremental sync docs
Fix error in incremental sync docs
Add Global Parent State Cursor
Add Global Parent State Cursor
Add limitation for number of partitions to PerPartitionCursor
Add limitation for number of partitions to PerPartitionCursor
Fix source-declarative-manifest
Fix source-declarative-manifest
Replace pydantic BaseModel with dataclass
Replace pydantic BaseModel with dataclass
use orjson instead of json to speed up JSON parsing
use orjson instead of json to speed up JSON parsing
Update json error message parser to identify additional error message fields in response bodies
Update json error message parser to identify additional error message fields in response bodies
Raise exceptions in file-based check, improve UI errors
Raise exceptions in file-based check, improve UI errors
add codeflash to dev environment
add codeflash to dev environment
Cache the result of interpolated strings when the evaluated value is equal to its raw representation
Cache the result of interpolated strings when the evaluated value is equal to its raw representation
CDK: refactor error handling in abstract source
CDK: refactor error handling in abstract source
Added support for RFR for Full-Refresh Substreams
Added support for RFR for Full-Refresh Substreams
Stop support for incoming legacy state message format
Stop support for incoming legacy state message format
Move the @deprecated decorator to the class level.
Move the @deprecated decorator to the class level.
Added test utils for integration tests
Added test utils for integration tests
file-based cdk: add excel file type support
file-based cdk: add excel file type support
Have better fallback error message on HTTP error
Have better fallback error message on HTTP error
Ensure at least one element returned by decoder
Ensure at least one element returned by decoder
resumable full refresh: fix issue when live traffic regression tests pass state to connector
resumable full refresh: fix issue when live traffic regression tests pass state to connector
Add PrintBuffer to emit records in batches
Add PrintBuffer to emit records in batches
Resumable full refresh: Add SubstreamResumableFullRefreshCursor to Python CDK to allow connectors to allow checkpointing on parent records
Resumable full refresh: Add SubstreamResumableFullRefreshCursor to Python CDK to allow connectors to allow checkpointing on parent records
Align BackoffStrategy interfaces to take attempt_count as a full-fledge parameter
Align BackoffStrategy interfaces to take attempt_count as a full-fledge parameter
Add ability to stop stream when retry-after is greater than a duration
Add ability to stop stream when retry-after is greater than a duration
Fix case where stream wont have a state attribute and needs to resolve get_updated_state
Fix case where stream wont have a state attribute and needs to resolve get_updated_state
General performance enhancement
fix declarative schema refs for Decoder
fix declarative schema refs for Decoder
Fixed: Resolved an issue in HttpClient that prevented correct error messages from being presented.
Fixed: Resolved an issue in HttpClient that prevented correct error messages from being presented.
Adding text field to declarative manifest schema for general connector description.
Adding text field to declarative manifest schema for general connector description.
add name property to http_client for convenience
add name property to http_client for convenience
low-code: fix record selector factory when using custom components
low-code: fix record selector factory when using custom components
fix OOM on predicate for streamable responses
fix OOM on predicate for streamable responses
low code: add new Decoders: JsonlDecoder, IterableDecoder
low code: add new Decoders: JsonlDecoder, IterableDecoder
low-code: fix overwrite for default backoff strategy
low-code: fix overwrite for default backoff strategy
CDK: fix handling for rate limit errors when checking connection
CDK: fix handling for rate limit errors when checking connection
resumable full refresh: Automatically apply RFR to streams (not including substreams) that are not incremental and implement next_page_token
resumable full refresh: Automatically apply RFR to streams (not including substreams) that are not incremental and implement next_page_token
Deprecate AvailabilityStrategy
Deprecate AvailabilityStrategy
CDK: add not exiting when rate limited
CDK: add not exiting when rate limited
Add failure_type to HttpResponseFilter(retry after pypi read error)
Add failure_type to HttpResponseFilter(retry after pypi read error)
Add failure_type to HttpResponseFilter
Add failure_type to HttpResponseFilter
Remove 3.11-style union
Remove 3.11-style union
Clean invalid fields from configured catalog
Clean invalid fields from configured catalog
resumable full refresh: Fix bug where checkpoint reader stops syncing too early if first partition is complete
resumable full refresh: Fix bug where checkpoint reader stops syncing too early if first partition is complete
file-based cdk: add config option to limit number of files for schema discover resumable full refresh: Fix bug for substreams depending on RFR parent
file-based cdk: add config option to limit number of files for schema discover resumable full refresh: Fix bug for substreams depending on RFR parent stream would not paginate over parent
CDK: add incomplete status to availability check during read
CDK: add incomplete status to availability check during read
CDK: flush buffer for each RATE_LIMITED message print
CDK: flush buffer for each RATE_LIMITED message print
CDK: add running stream status with rate limit reason to backoff approach
CDK: add running stream status with rate limit reason to backoff approach
CDK: add incomplete stream status to nonexistent stream handling
CDK: add incomplete stream status to nonexistent stream handling
Integrate HttpClient into HttpStream class. See migration guide for more details.
Integrate HttpClient into HttpStream class. See migration guide for more details.
CDK: Add support for input format parsing at jinja macro format_datetime
CDK: Add support for input format parsing at jinja macro format_datetime
Nothing published for this version
Your coding agent can read these notes before it upgrades. Set up the MCP server →