Configuration

The backend is configured with environment variables.

Variable

Required

Description

DASHSCOPE_API_KEY

Default provider

Alibaba Cloud DashScope key for Qwen3-ASR-Flash, Qwen Omni Realtime and Qwen-Audio ASR. The default provider is qwen, so it is needed unless ASR_PROVIDER names another one (mock needs no key)

DASHSCOPE_WORKSPACE_ID

No

Model Studio workspace id. Needed only by qwen-omni/qwen3.8-omni-flash-realtime and the qwen-audio/* models, which are served from <id>.ap-southeast-1.maas.aliyuncs.com

API_KEYS

No

Comma-separated sub:secret pairs for API key exchange

GROQ_API_KEY

No

Groq API key for gpt-oss-120b post-processing; enables the +groq configurations

ASR_CONFIGURATION

No

Default configuration id (default: derived from ASR_PROVIDER/ASR_MODEL/GROQ_API_KEY)

ASR_CONFIGURATIONS

No

Comma-separated configurations clients may select (default: every configuration with credentials)

ASR_PROVIDER

No

Provider for the derived default: qwen (default), qwen-omni, qwen-audio, byteplus, openai, xai, elevenlabs, iapp, openrouter, meta, paxa, gemini, ai-gateway, or mock

ASR_MODEL

No

Model for the derived default (default: the provider’s own default model)

BYTEPLUS_API_KEY

No

BytePlus key; enables the byteplus configurations

BYTEPLUS_LANGUAGE

No

BytePlus language hint, e.g. th-TH (default: unset — Mandarin/English only)

BYTEPLUS_RESOURCE_ID

No

BytePlus resource id (default: volc.seedasr.sauc.duration)

BYTEPLUS_BASE_URL

No

BytePlus endpoint base, without the mode path segment

OPENAI_API_KEY

No

OpenAI key; enables the openai (gpt-live-transcribe) configuration

XAI_API_KEY

No

xAI key; enables the xai (grok-voice-transcribe-2.0, streaming) configuration

ELEVENLABS_API_KEY

No

ElevenLabs key; enables the elevenlabs (scribe_v2_realtime) configuration

IAPP_API_KEY

No

iApp key; enables the iapp (chinda-asr-realtime-v1, streaming, Thai only) configuration

OPENROUTER_API_KEY

No

OpenRouter key; enables the openrouter configurations (microsoft/mai-transcribe-2, meta/muse-voice-transcribe-1.0, microsoft/mai-transcribe-1.5, google/gemini-3.5-transcribe)

META_API_KEY

No

Meta key; enables the meta (muse-voice-transcribe-1.0, direct realtime API) configuration

PAXA_API_KEY

No

Paxa Labs key; enables the paxa configurations (batch and realtime)

PAXA_CONVENTION

No

written to render numbers as digits and use the Thai repetition mark (default: the vendor’s spoken)

GEMINI_API_KEY

No

Google Gemini API key; enables the gemini (gemini-3.5-transcribe-live) configuration

AI_GATEWAY_API_KEY

No

Vercel AI Gateway key; enables the ai-gateway (microsoft/mai-transcribe-2-streaming) configuration

OIDC_DISCOVERY_URL

No

OIDC provider discovery URL (alternative to API keys)

OIDC_CLIENT_ID

No

OIDC client ID (default: vxbeamer-mobile)

OIDC_AUDIENCE

No

Expected token audience (default: same as client ID)

OIDC_SECRET

No

HMAC secret for session tokens (default: local-dev-secret)

PUBLIC_URL

No

Address clients should use for pairing links (default: the address the request came in on)

WEBHOOK_URL

No

Endpoint to POST completed transcriptions to

PORT

No

HTTP port (default: 8787)

WS_IDLE_TIMEOUT_MS

No

How long a recording socket may go without audio before it is closed and its vendor connection released (default: 60000; 0 disables)

QWEN_OMNI_STICKY_LINGER_MS

No

How long an unused Qwen Omni connection is kept open after a recording, for the next one from the same client (default: 30000)

QWEN_OMNI_STICKY_MAX_AUDIO_SECONDS

No

Audio a kept Qwen Omni connection may carry across recordings before it is retired (default: 180)

Eval storage

Where eval votes and eval-sets are saved. The backend only signs upload URLs; the browser uploads directly to the bucket, so recordings never pass through the server and no bucket credential ever reaches a browser. Setting EVAL_STORAGE_BUCKET turns the feature on; without it, evals still run and winners still apply, they are just not recorded.

Variable

Required

Description

EVAL_STORAGE_BUCKET

No

S3-compatible bucket for votes and eval-sets. Enables eval storage

EVAL_STORAGE_ACCESS_KEY_ID

With bucket

Access key id used to sign upload URLs

EVAL_STORAGE_SECRET_ACCESS_KEY

With bucket

Secret access key used to sign upload URLs

EVAL_STORAGE_REGION

No

Bucket region (default: us-east-1)

EVAL_STORAGE_ENDPOINT

No

Custom S3-compatible endpoint, e.g. MinIO or R2 (default: AWS)

EVAL_STORAGE_FORCE_PATH_STYLE

No

Set to true for endpoints needing path-style addressing, e.g. MinIO

EVAL_STORAGE_PREFIX

No

Key prefix placed before votes/ and eval-sets/

Objects are keyed by kind, then UTC date, then time and message id:

votes/2026/07/16/2026-07-16T09-30-00-123Z-<messageId>.json
eval-sets/2026/07/16/2026-07-16T09-30-00-123Z-<messageId>.json

A vote ({"type":"vote",…}) is written on every winner pick and carries the winning configuration id, the full candidate list, per-candidate cost/latency/errors, and the audio’s duration — but no audio and no transcripts, so it is safe to store unconditionally. An eval-set ({"type":"eval-set",…}) is written only when the user ticks save-for-eval, and adds the recording as a base64 WAV plus every candidate’s transcript. A pick’s vote and eval-set share a key suffix, so one finds the other.