airbyte-cdk

Change Error message if Stream is not found

Vector DB CDK: Add text splitting options to document processing

Ensuring invalid user-provided urls does not generate sentry issues

Vector DB CDK adjustments: Prevent failures with big records and OpenAI embedder

[ISSUE #30353] File-Based CDK: remove file_type from stream config

Connector Builder: fix datetime format inference for str parsable as int but not isdecimal

Vector DB CDK: Add Azure OpenAI embedder

File-based CDK: improve error message for CSV parsing error

File-based CDK: migrated parsing error to config error to avoid sentry alerts

Add from-field embedder to vector db CDK

FIle-based CDK: Update spec and fix autogenerated headers with skip after

Vector DB CDK adjustments: Fix id generation, improve config spec, add base test case

[Issue #29660] Support empty keys with record selection

Add vector db CDK helpers

File-based CDK: allow user to provided column names for CSV files

File-based CDK: allow for extension mismatch

File-based CDK: Remove CSV noisy log

Source-S3 V4: feature parity rollout

File-based CDK: Do not stop processing files in slice on error

Check config against spec in embedded sources and remove list endpoint from connector builder module

low-code: allow formatting datetime as milliseconds since unix epoch

File-based CDK: handle legacy options

Fix title and description of datetime_format fields

File-based CDK cursor and entrypoint updates

Low code CDK: Decouple SimpleRetriever and HttpStream

Add utils for embedding sources in other Python applications

Relax pydantic version requirement and update to protocol models version 0.4.0

Support many format for cursor datetime

File-based CDK updates

Connector Builder: Ensure we return when there are no slices

low-code: deduplicate query params if they are already encoded in the URL

Fix RemoveFields transformation issue

Breaking change: Rename existing SessionTokenAuthenticator to LegacySessionTokenAuthenticator and make SessionTokenAuthenticator more generic

Connector builder: warn if the max number of records was reached

Remove pyarrow from main dependency and add it to extras

Fix pyyaml and cython incompatibility

Connector builder: Show all request/responses as part of the testing panel

[ISSUE #27494] allow for state to rely on transformed field

Ensuring the state value format matches the cursor value from the record

Fix issue with incremental sync following data feed release

Support data feed like incremental syncs

Fix return type of RecordFilter: changed from generator to list

Connector builder module: serialize request body as string

Fix availability check to handle HttpErrors which happen during slice extraction

Refactoring declarative state management

Error message on state per partition state discrepancy

Supporting state per partition given incremental sync and partition router

Use x-www-urlencoded for access token refresh requests

Replace with when making oauth calls

Emit messages using message repository

Add utils for inferring datetime formats

Add a metadata field to the declarative component schema

make DatetimeBasedCursor.end_datetime optional

Remove SingleUseRefreshTokenOAuthAuthenticator from low code CDK and add generic injection capabilities to ApiKeyAuthenticator

Connector builder: add latest connector config control message to read calls

Add refresh token update capabilities to OAuthAuthenticator

Make step and cursor_granularity optional