TheBlueSky commented 3 months ago

Bug Report

IMPORTANT: We primarily use GitHub as an issue tracker. Just open an issue here if you have encountered a bug in MiroTalk. If you have ideas, suggestions, questions, doubts about MiroTalk or need support, please use the MiroTalk Forum instead:

👉 https://discord.gg/rgGYfeYW3N

Describe the bug

I am installed MiroTalk using the latest Docker image, mirotalk/p2p:latest, and configure the ChatGPT section with my API Key, while leaving the rest of the settings as is:

CHATGPT_ENABLED=true
CHATGPT_BASE_PATH=https://api.openai.com/v1/
CHATGTP_APIKEY=MyApiKey
CHATGPT_MODEL=gpt-3.5-turbo-instruct
CHATGPT_MAX_TOKENS=1000
CHATGPT_TEMPERATURE=0

When I enable ChatGPT in that chat window and send a message, the server crashes with the following error message:

mirotalk  | /src/app/src/server.js:704
mirotalk  |                     if (error instanceof OpenAI.APIError) {
mirotalk  |                                          ^
mirotalk  |
mirotalk  | ReferenceError: OpenAI is not defined
mirotalk  |     at Socket.<anonymous> (/src/app/src/server.js:704:42)
mirotalk  |     at process.processTicksAndRejections (node:internal/process/task_queues:95:5)

By looking at the source code, it seems that OpenAI being undefined masks the original error. So, it isn't clear what is actually failing.

To reproduce

Steps to reproduce the behavior:

Create a Docker Compose file like the one below these steps
Download .env.template file from this repo into the same directory as the Docker Compose file, and rename it to .env
Update ChatGPT section in the .env file as shown in the bug description section above (you may also need to update other configurations, such as host)
Run the container with docker compose up -d && docker compose logs -f, so that you see the logs
Navigate to the application, enter a room, open the chat window, enable ChatGPT, and post a message
See error in the logs

services:
  mirotalk:
    image: mirotalk/p2p:latest
    container_name: mirotalk
    hostname: mirotalk-docker
    volumes:
        - ./.env:/src/.env:ro
    ports:
      - 3000:3000/tcp
    privileged: false
    security_opt:
      - no-new-privileges:true
    restart: unless-stopped

Expected behavior

If there is an error in the configuration or an error from OpenAI, I expect to see it in the logs
OpenAI for some reason is undefined, while it should be at this point based on the fact that it is enabled and has the needed configuration. The code here should have taken care of this.
The expected behaviour is that I do not get an error and get ChatGPT response instead :)

Screenshots

Desktop - Mobile

Please complete the following information:

Device or OS: Windows
Browser: Brave (1.61.120) and Vivaldi (6.5.3206.55)
App Version: 1.2.75

Additional context

-

miroslavpejic85 commented 3 months ago

Hello @TheBlueSky, Thank you for reporting it. I've made a small adjustment. Please check your server console for any potential errors that can help you to identifying the issue.