codex-backend-sdk 0.3.8__tar.gz → 0.3.9__tar.gz

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/CHANGELOG.md

@@ -2,6 +2,15 @@
 
 All notable changes to this project will be documented in this file.
 
+## [0.3.9] - 2026-07-17
+
+### Fixed
+- Detect anonymous audio buffers by their file signature before ChatGPT transcription uploads, correcting filenames and MIME types when callers provide misleading generic names such as `audio.mp3` for WAV data.
+- Restored reliable `AIClient.audio_to_text(bytes)` integration with `/backend-api/transcribe` without requiring callers to understand multipart backend constraints.
+
+### Tests
+- Added regression coverage for WAV buffers carrying an incorrect `.mp3` name and verified the complete Codex Agent transcription path against the real ChatGPT backend.
+
 ## [0.3.8] - 2026-07-17
 
 ### Changed

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codex-backend-sdk
-Version: 0.3.8
+Version: 0.3.9
 Summary: Unofficial Python SDK for the ChatGPT Codex backend API
 License: MIT
 License-File: LICENSE

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/codex_backend_sdk/__init__.py

@@ -17,7 +17,7 @@ Quickstart:
     print(response.output_text)
 """
 
-__version__ = "0.3.8"
+__version__ = "0.3.9"
 
 from .oauth import run_oauth_flow, refresh_access_token
 from .storage import load_tokens, save_tokens, TokenStore

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/codex_backend_sdk/resources/openai_oauth.py

@@ -126,7 +126,7 @@ class AudioTranscriptions:
 
         response = self._client._post_chatgpt_raw(
             "/transcribe",
-            files={"file": _coerce_file(file)},
+            files={"file": _coerce_audio_file(file)},
             data={key: _form_value(value) for key, value in data.items()},
             headers=extra_headers,
             params=extra_query,
@@ -136,3 +136,38 @@ class AudioTranscriptions:
         if _is_given(response_format) and response_format == "text":
             return transcription.text
         return transcription
+
+
+def _coerce_audio_file(file: Any) -> Any:
+    coerced = _coerce_file(file)
+    if isinstance(coerced, tuple):
+        return coerced
+    if not hasattr(coerced, "read"):
+        return coerced
+
+    position = coerced.tell() if hasattr(coerced, "tell") else None
+    header = coerced.read(12)
+    if position is not None and hasattr(coerced, "seek"):
+        coerced.seek(position)
+
+    detected = _audio_type(header)
+    if detected is None:
+        return coerced
+    extension, mime_type = detected
+    name = getattr(coerced, "name", "audio") or "audio"
+    stem = str(name).rsplit("/", 1)[-1].rsplit(".", 1)[0] or "audio"
+    return (f"{stem}.{extension}", coerced, mime_type)
+
+
+def _audio_type(header: bytes) -> tuple[str, str] | None:
+    if header.startswith(b"RIFF") and header[8:12] == b"WAVE":
+        return "wav", "audio/wav"
+    if header.startswith(b"ID3") or header[:2] in {b"\xff\xfb", b"\xff\xf3", b"\xff\xf2"}:
+        return "mp3", "audio/mpeg"
+    if header.startswith(b"OggS"):
+        return "ogg", "audio/ogg"
+    if header.startswith(b"fLaC"):
+        return "flac", "audio/flac"
+    if header[4:8] == b"ftyp":
+        return "m4a", "audio/mp4"
+    return None

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "codex-backend-sdk"
-version = "0.3.8"
+version = "0.3.9"
 description = "Unofficial Python SDK for the ChatGPT Codex backend API"
 readme = "README.md"
 license = { text = "MIT" }

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/tests/test_openai_oauth_resources.py

@@ -1,3 +1,5 @@
+from io import BytesIO
+
 import pytest
 
 from codex_backend_sdk import CreateEmbeddingResponse, OpenAI, Transcription
@@ -92,6 +94,24 @@ def test_audio_transcriptions_create_posts_to_chatgpt_backend():
     assert headers["ChatGPT-Account-ID"] == "account-id"
 
 
+def test_audio_transcriptions_detects_anonymous_audio_format():
+    client = FakeOpenAIClient()
+    audio = BytesIO(b"RIFF\x00\x00\x00\x00WAVEfmt ")
+    audio.name = "audio.mp3"
+
+    client.audio.transcriptions.create(
+        model="gpt-4o-mini-transcribe",
+        file=audio,
+    )
+
+    _, kwargs, _ = client.chatgpt_raw_posts[0]
+    filename, uploaded, mime_type = kwargs["files"]["file"]
+    assert filename == "audio.wav"
+    assert uploaded is audio
+    assert mime_type == "audio/wav"
+    assert audio.tell() == 0
+
+
 def test_audio_transcriptions_text_format_returns_string():
     client = FakeOpenAIClient()
 

{codex_backend_sdk-0.3.8 → codex_backend_sdk-0.3.9}/uv.lock

@@ -160,7 +160,7 @@ wheels = [
 
 [[package]]
 name = "codex-backend-sdk"
-version = "0.3.8"
+version = "0.3.9"
 source = { editable = "." }
 dependencies = [
     { name = "pydantic" },