After passing in the baseURL for the OPENAI object, Api will report an error when generating the component

[OrzMiku]

I have modified the code in these areas of the program

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: 'my api url',
});


/openv0/server/modules/multipass/passes/build-component-generation-context/rag_icons.js
  7,16: const openai = new OpenAI({
/openv0/server/modules/multipass/passes/design-component-iteration-from-description/index.js
  10,16: const openai = new OpenAI({
/openv0/server/modules/multipass/passes/design-component-iteration-from-json/index.js
  10,16: const openai = new OpenAI({
/openv0/server/modules/multipass/passes/design-component-new-from-description/index.js
  10,16: const openai = new OpenAI({
/server/modules/multipass/passes/design-component-new-from-json/index.js
  10,16: const openai = new OpenAI({
/server/modules/multipass/passes/generate-component-iteration/index.js
  6,16: const openai = new OpenAI({
/server/modules/multipass/passes/generate-component-new/index.js
  6,16: const openai = new OpenAI({
/server/modules/multipass/passes/validate-fix-generated-component/index.js
  9,16: const openai = new OpenAI({

I am confident that my interface address can be accessed and used normally. It supports the GPT4 model, and I have tested it using the OpenAI package in my own program. But in this program, I encountered an error while generating the component using my own Api address. The error is as follows：

gpt-4 model

Server is running on http://localhost:3000
Connected to the database
{
  module: 'multipass/preset',
  preset: {
    name: 'componentNew_description',
    description: 'generate a new component from a text description'
  }
}
{
  module: 'multipass/run',
  status: 'starting',
  query: {
    description: 'google search ui',
    framework: 'react',
    components: 'nextui',
    icons: 'lucide'
  },
  preset: {
    name: 'componentNew_description',
    description: 'generate a new component from a text description'
  },
  passes: [
    'design-component-new-from-description',
    'build-component-generation-context',
    'generate-component-new',
    'validate-check-generated-component',
    'validate-fix-generated-component',
    'postprocess-generated-component',
    'store-component',
    'export-component'
  ]
}
> pass 0/7
> init : passes/design-component-new-from-description
undefined:1
()
 ^

SyntaxError: Unexpected token ')'
    at Object.run (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/passes/design-component-new-from-description/index.js:154:17)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async run (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/index.js:31:22)
    at async Object.preset (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/index.js:80:10)
    at async /home/orzmiku/DevDrive/aicode/openv0/server/api.js:138:21

Node.js v18.18.0

gpt-3.5-turbo model

(base) orzmiku@DevMiku:~/DevDrive/aicode/openv0/server$ node api.js 
Server is running on http://localhost:3000
Connected to the database
{
  module: 'multipass/preset',
  preset: {
    name: 'componentNew_description',
    description: 'generate a new component from a text description'
  }
}
{
  module: 'multipass/run',
  status: 'starting',
  query: {
    description: 'google search ui',
    framework: 'react',
    components: 'nextui',
    icons: 'lucide'
  },
  preset: {
    name: 'componentNew_description',
    description: 'generate a new component from a text description'
  },
  passes: [
    'design-component-new-from-description',
    'build-component-generation-context',
    'generate-component-new',
    'validate-check-generated-component',
    'validate-fix-generated-component',
    'postprocess-generated-component',
    'store-component',
    'export-component'
  ]
}
> pass 0/7
> init : passes/design-component-new-from-description
{
{
  "new_component_name": "GoogleSearchUI",
  "new_component_description": "The GoogleSearchUI component is designed to replicate the user interface of the Google search engine. It allows users to input search queries and displays search results in a visually appealing and user-friendly manner.",
  "use_library_components": [
    {
      "library_component_name": "",
Input",
      "library_component_usage_reason": "To provide a text input field where users can enter their search query."
    },
    {
      "library_component_name": "Button",
      "library_component_usage_reason": "To display a search button that users can click to initiate the search."
    },
    {
      "library_component_name": "Card",
      "library_component_usage_reason": "To display the search results in a card-like format, with relevant information and a title."
    },
    {
      "library_component_name": "Avatar",
      "library_component_usage_reason": "To display the profile picture or logo of the source of the search result."
    },
    {
      "library_component_name": "Link",
      "library_component_usage_reason": "To create clickable links that direct users to the source of the search result."
    }
  ]
}undefined:2
{
^

SyntaxError: Unexpected token '{'
    at Object.run (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/passes/design-component-new-from-description/index.js:154:17)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async run (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/index.js:31:22)
    at async Object.preset (/home/orzmiku/DevDrive/aicode/openv0/server/modules/multipass/index.js:80:10)
    at async /home/orzmiku/DevDrive/aicode/openv0/server/api.js:138:21

Node.js v18.18.0

[just-a-stone]

mark. my env also has to use baseURL.

[mooerccx]

I encountered this problem too

[raidendotai]

• the chat completion functions in multipass plugins are independent from everything else, and can be (re)implemented in any way you want
• seems like the error here happens in parsing the JSON string returned from completion (in my implementation, i used eval() because gpt function_call does not return a strictly parseable JSON)
• i suggest you log the returned object right before this line and use it to troubleshoot
• maybe we need an interface to integrate multiple LLMs straight from .env config, but honestly it's not top priority now