Skip to content

Fix transcription endpoint requiring openarc_asr parameter when using Qwen3 ASR models #117

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
SearchSavior merged 2 commits into from
May 24, 2026
Merged

Fix transcription endpoint requiring openarc_asr parameter when using Qwen3 ASR models #117

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b353260
Fix transcription endpoint requiring `openarc_asr` parameter when usi…
solidDoWant May 22, 2026
b51c88b
update docs
solidDoWant May 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions docs/commands.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -279,8 +279,7 @@ This page contains example commands to help you choose models and configure Open
--device CPU
```

Use the `/v1/audio/transcriptions` endpoint with `openarc_asr` in the request body:
I have not tested our implementation with any community tooling yet; however, all tests using the openai python library are passing, and usually that's enough.
Chunking can be configured on a per-request basis via the `openarc_asr` in the request body for the `/v1/audio/transcriptions`. If not set, the below defaults will be used. Defaults values cannot currently be configured.

For the options in `extra_body`, they will likely not have support in any third party tool you don't build from scratch. I'm working on improving how these can be configured. Currently, the behavior is modified per request, so you can tinker with performance on CPU and GPU. At this time NPU device is unsupported.

Expand All @@ -300,6 +299,7 @@ This page contains example commands to help you choose models and configure Open
model="<model-name>",
file=f,
response_format="verbose_json",
# Optional. The below values will be used as defaults if `openarc_asr` is not provided.
extra_body={
"openarc_asr": json.dumps({
"qwen3_asr": {
Expand Down
11 changes: 6 additions & 5 deletions src/server/routes/openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -510,11 +510,12 @@ async def openai_audio_transcriptions(
normalized_model_type = ModelType(selected_model_type)

if normalized_model_type == ModelType.QWEN3_ASR:
if not openarc_asr:
raise ValueError("openarc_asr required for Qwen3 ASR models")
cfg = OpenArcASRConfig.model_validate(json.loads(openarc_asr))
if not cfg.qwen3_asr:
raise ValueError("openarc_asr.qwen3_asr required for Qwen3 ASR models")
payload = json.loads(openarc_asr) if openarc_asr else {}
if not payload.get("qwen3_asr"):
# Fall back to defaults if qwen3_asr config is not provided
payload["qwen3_asr"] = {}

cfg = OpenArcASRConfig.model_validate(payload)
gen_config = cfg.qwen3_asr.model_copy(update={"audio_base64": audio_base64})
result = await _workers.transcribe_qwen3_asr(model, gen_config)
else:
Expand Down