[BUG] speechmatics-batch: Transcript Result dataclass missing is_eos and attaches_to

[seanl-kai]

Which SDK are you using?

• speechmatics-rt (Real-Time SDK)
• speechmatics-batch (Batch SDK)
• speechmatics-voice (Voice SDK)
• All

Package Information

• package: speechmatics-batch
• package version: 0.4.7
• python version: 3.13
• os version: macOS 26.3.1 (a)

Describe the bug
Completed an upgrade from the legacy API to the new one, which caused a regression in a later processing step. This is due to the move to a dataclass driven approach, however a number of fields are missing from the RecognitionResult class. These include is_eos and attaches_to.

The solution we have opted for is to manually read the transcript from the API after job completion, bypassing the dataclass entirely:

transcript = await client._transport.get(f"/jobs/{job.id}/transcript", params={"format": FormatType.JSON.value})

To Reproduce
Steps to reproduce the behavior:

1. Wait for a job completion.
2. Attempt to iterate over results in the transcript, reading missing fields.

Expected behavior
A clear and concise description of what you expected to happen.

The ability to access fields that are available from the API endpoint.

Actual behavior
A clear and concise description of what actually happened.

Fields are missing after the transformation into the Result dataclass.

Code snippet
If applicable, add a minimal code snippet that reproduces the issue:

job = await client.submit_job(
    audio_file=path_to_file,
    transcription_config=transcription_config,
)

test = await client.wait_for_completion(job.id)
for result in test.results or []:
    result.is_eos