Using JinaAIEmbedding Still Requires OPENAI_API_KEY in LlamaIndex 0.9.3

[sujeet-agrahari]

Describe the bug
When using JinaAIEmbedding as the embedding model in LlamaIndex 0.9.3, the script still throws an error requiring OPENAI_API_KEY. However, since OpenAI’s embeddings are not used, this should not be required.

To Reproduce
Code to reproduce the behavior:

import {
  JinaAIEmbedding,
  VectorStoreIndex,
  Document,
  Settings
} from 'llamaindex';

// Configure JinaAIEmbedding
Settings.embedModel = new JinaAIEmbedding({
  apiKey: 'jina_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx', // Dummy API Key
  model: 'jina-embeddings-v3',
});

async function testEmbedding() {
  try {
    const doc = new Document({ text: "This is a test document." });
    const index = await VectorStoreIndex.fromDocuments([doc]);
    console.log("Embedding created successfully.");
  } catch (error) {
    console.error("Error:", error);
  }
}

testEmbedding();

Error:

The OPENAI_API_KEY environment variable is missing or empty; either provide it, or instantiate the OpenAI client with an apiKey option, like new OpenAI({ apiKey: 'My API Key' }).

Expected behavior
The script should work without requiring OPENAI_API_KEY since it only relies on JinaAIEmbedding, and I am not using any LLM.

Desktop (please complete the following information):

• OS: macOS
• JS Runtime / Framework / Bundler (select all applicable)
• Node.js
• Deno
• Bun
• Next.js
• ESBuild
• Rollup
• Webpack
• Turbopack
• Vite
• Waku
• Edge Runtime
• AWS Lambda
• Cloudflare Worker
• Others (please elaborate on this)
• Version [e.g. 22]

Additional context
I am trying to use the package in a VSCode extension development environment.

[marcusschiesser]

Before fixing this, we should move the jina embedding class to its own npm package

[sujeet-agrahari]

Before fixing this, we should move the jina embedding class to its own npm package

+1.

Currently the llamaindex package size is huge.

Update on the issue:

This behavior occurs only when using indexStore.asQueryEngine(), but it does not happen with asRetriever().

If this is the expected behavior, we should enhance the documentation and close the issue.