Can't add whisper STT

[saeedesmaili]

Groq provides openai compatible api for whisper:
https://console.groq.com/docs/openai

I have done the followings:

librechat.yaml:

speech:
  stt:
    openai:
      url: 'https://api.groq.com/openai/v1'
      apiKey: '${GROQ_API_KEY}'
      model: 'distil-whisper-large-v3-en'

deploy-compose.yml:

services:
  api:
    image: ghcr.io/danny-avila/librechat-dev-api:latest
    container_name: LibreChat-API
    ports:
      - 3080:3080
    depends_on:
      - mongodb
      - rag_api
    restart: always
    extra_hosts:
    - "host.docker.internal:host-gateway"
    env_file:
      - .env
    environment:
      - HOST=0.0.0.0
      - NODE_ENV=production
      - MONGO_URI=mongodb://mongodb:27017/LibreChat
      - MEILI_HOST=http://meilisearch:7700
      - RAG_PORT=${RAG_PORT:-8000}
      - RAG_API_URL=http://rag_api:${RAG_PORT:-8000}
      - GROQ_API_KEY=${GROQ_API_KEY} ## added this
    volumes:
      - type: bind
        source: ./librechat.yaml
        target: /app/librechat.yaml
      - ./images:/app/client/public/images
      - ./logs:/app/api/logs

and added GROQ_API_KEY environment variable.

But when I visit the speech settings after deploy, I still only have "browser" for the engine:

What should I do next to troubleshoot this?

[berry-13]

you used the key tts when the correct one should be stt

here's the correct YAML configuration:

speech:
  stt:
    openai:
      url: 'https://api.groq.com/openai/v1'
      apiKey: '${GROQ_API_KEY}'
      model: 'distil-whisper-large-v3-en'

[saeedesmaili]

Yeah that was a noob mistake, noticed and fixed it before your message, but the issue hasn't been resolved. I see the following the api container's logs, not sure if it's related:

2025-03-05 20:00:57 error: Config file YAML format is invalid: EISDIR: illegal operation on a directory, read

[berry-13]

try checking your config file and make sure it has permission for reading

[saeedesmaili]

It say's it's valid. Not sure how to check if it has permission (I'm using coolify to self-host librechat). I probably should wait until this settings comes to UI some day :/

[ghost]

When you check your environment file .env ... you will see that there are extra sections where you can add API keys for STT and TTS

#==================================================#
#          Speech to Text & Text to Speech         #
#==================================================#
 
STT_API_KEY=mega_secret_API_key
TTS_API_KEY=even_more_secret_key 

Hope this helps. Haven't setup STT and TTS yet though this seems like the most likely mistake. I use Groq as AI provider for their conversational models and this is why you would set the GROQ_API_KEY in the .env file

My librechat.yaml for Groq looks like this then:

custom:
    # Groq Example
    - name: 'groq'
      apiKey: '${GROQ_API_KEY}'
      baseURL: 'https://api.groq.com/openai/v1/'
      models:
        default:
          [
            'deepseek-r1-distill-llama-70b',
            'deepseek-r1-distill-llama-70b-specdec',
            'deepseek-r1-distill-qwen-32b',
            'gemma2-9b-it',
            'llama-3.1-8b-instant',
            'llama-3.2-11b-vision-preview',
            'llama-3.2-1b-preview',
            'llama-3.2-3b-preview',
            'llama-3.2-90b-vision-preview',
            'llama-3.3-70b-specdec',
            'llama-3.3-70b-versatile',
            'llama-guard-3-8b',
            'llama3-70b-8192',
            'llama3-8b-8192',
            'mixtral-8x7b-32768',
            'qwen-2.5-32b',
            'qwen-qwq-32b',  
          ]
        fetch: false
      titleConvo: true
      titleModel: 'mixtral-8x7b-32768'
      modelDisplayLabel: 'groq'

Hope this makes it clear what GROQ_API_KEY is referencing in the .env file