admin/ai-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Make Voxtral the only transcription provider
Some checks failed
CI / test (push) Failing after 8s
Details
Build and Deploy / build-and-deploy (push) Successful in 27s
Details

Browse Source
This commit is contained in:
Grendgi
2026-06-09 16:54:54 +03:00
parent 5c965be8c9
commit 9bd6d726f0
15 changed files with 128 additions and 900 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
k8s/configmap.yaml
View File

@@ -11,20 +11,11 @@ data:
LLM_BASE_URL: "http://10.2.3.5:8002"
LLM_MODEL: "qwen2.5-14b"
LLM_TIMEOUT: "5m"
TRANSCRIPTION_PROVIDERS: "voxtral-small"
WHISPERX_URL: ""
WHISPERX_TIMEOUT: "10m"
WHISPERX_LEAD_SILENCE: "800ms"
# Fill these after Qwen2-Audio and Voxtral are exposed as OpenAI-compatible
# chat-completions endpoints on the AI server.
QWEN_AUDIO_BASE_URL: ""
QWEN_AUDIO_MODEL: "Qwen/Qwen2-Audio-7B-Instruct"
QWEN_AUDIO_TIMEOUT: "10m"
# Voxtral Small is the only transcription provider. It is exposed on the AI
# server through an OpenAI-compatible /v1/audio/transcriptions endpoint.
VOXTRAL_BASE_URL: "http://10.2.3.5:8004"
VOXTRAL_MODEL: "mistralai/Voxtral-Small-24B-2507"
VOXTRAL_TIMEOUT: "10m"
AUDIO_LLM_MAX_TOKENS: "4096"
FFMPEG_PATH: "/usr/bin/ffmpeg"
AI_STATS_SIDECAR_URL: "http://10.2.3.5:9090"
AI_STATS_TIMEOUT: "8s"
WORKER_POLL_INTERVAL: "2s"