[Bug]: AttributeError - ListenClient Object Missing Prerecorded Attributes (SDK v5.3.0)

[harrymagick]

Summary

The necessary client object for pre-recorded transcription is not exposed via the DeepgramClient.

What happened?

System

Operating Environment: Linux Fedora 42 KDE Plasma 6.5 (Clean Install)
Python Version: 3.13 (running in a dedicated VENV)
Deepgram SDK Version: 5.3.0

Description

I am attempting to migrate a transcription script to use the latest Deepgram Python SDK (v5.3.0). Despite ensuring a clean install, explicit version control, and trying all documented import paths, the final API call consistently fails with an AttributeError.

Constant Error and Tried Syntax

The issue is that the necessary client object for pre-recorded transcription is not exposed via the DeepgramClient.

The script consistently fails at the API call, indicating missing attributes, regardless of the syntax used:

Syntax Attempted	Resulting Error
self.deepgram.listen.prerecorded.transcribe_file(...)	AttributeError: 'ListenClient' object has no attribute 'prerecorded'
self.deepgram.prerecorded.transcribe_file(...)	AttributeError: 'DeepgramClient' object has no attribute 'prerecorded'
self.deepgram.listen.transcribe_prerecorded(...)	AttributeError: 'ListenClient' object has no attribute 'transcribe_prerecorded'

Summary of Failed Imports (Context for Internal SDK Issue)

We also confirm that the following canonical imports failed, indicating a likely package structure issue in the v5.3.0 build for this environment:

• from deepgram.models import PrerecordedOptions (Failed with ModuleNotFoundError)
• from deepgram.source import FileSource (Failed with ModuleNotFoundError)

Request: Please advise on the correct syntax for the transcribe_file method using raw bytes (request=audio_bytes) for SDK v5.3.0, or confirm if this is a known packaging bug requiring a fix or a downgrade.

Steps to reproduce

1. Installation: pip install deepgram-sdk==5.3.0 (Confirmed via pip show deepgram-sdk).
2. Audio Load: Loading audio as raw bytes (audio_bytes = audio_file.read()) to bypass FileSource import issues.
3. Client Initialization: self.deepgram = DeepgramClient(api_key=DG_KEY).

Minimal code sample

import os
import sys

from deepgram import DeepgramClient 

DG_KEY = os.getenv("DEEPGRAM_API_KEY")

try:
    # 1. Initialize client
    client = DeepgramClient(api_key=DG_KEY)
    
    # 2. Prepare audio bytes (Mocking a small file read)
    # The actual audio file (75MB) is loaded as bytes via file.read()
    audio_bytes = b'dummy_audio_data' 

    # 3. Call the API (This is the line that fails with AttributeError)
    print("Attempting API call...")
    response = client.listen.transcribe_prerecorded(
        request=audio_bytes,
        options={
            "model": "nova-2",
            "diarize": True,
            "language": "en"
        }
    )
    print("Success:", response)

except AttributeError as e:
    print(f"FAILED with AttributeError: {str(e)}")
    # Expected output: 'ListenClient' object has no attribute 'transcribe_prerecorded'
except Exception as e:
    print(f"An unexpected error occurred: {str(e)}")

Logs / traceback

Transport

HTTP

API endpoint / path

/v1/listen/prerecorded

Model(s) used

nova-2

How often?

Always

Is this a regression?

• Yes, it worked in an earlier version

Last working SDK version (if known)

No response

SDK version

5.3.0

Python version

3.13

Install method

pip

OS

Linux (x86_64)

Environment details

Link to minimal repro (optional)

No response

Session ID (optional)

No response

Project ID (optional)

No response

Request ID (optional)

No response

Code of Conduct

• I agree to follow this project’s Code of Conduct