BadRequestError occurred when using AzureAISearchVectorStore and AzureOpenAIEmbeddings

[Columboom]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

import dotenv from 'dotenv';
import { ClientSecretCredential } from "@azure/identity";
import { MarkdownTextSplitter } from "langchain/text_splitter";
import { TextLoader } from "langchain/document_loaders/fs/text";
import { AzureOpenAIEmbeddings } from "@langchain/openai";
import {
    AzureAISearchVectorStore,
    AzureAISearchQueryType,
  } from "@langchain/community/vectorstores/azure_aisearch";
import { SearchClient } from "@azure/search-documents";

dotenv.config();
const clientId = process.env.client_id;
const clientSecret = process.env.client_secret;
const tenantId = process.env.tenant_id;
const aisearchEndpoint = process.env.aisearch_endpoint;
const indexName = "test_js";

class MyToken:...
const myToken = new MyToken();
console.log(await myToken.token());

const MY_PROXY = "xxx"
const HEADERS = {
 ...
};

const azureEmbeddingDeploymentName = 'TEXT-EMBEDDING-ADA-002';
const azureOpenAIApiVersion = "2024-02-01";

const embeddings = new AzureOpenAIEmbeddings({
  openAIApiVersion: azureOpenAIApiVersion,
  azureOpenAIBasePath: MY_PROXY+"/openai/deployments", 
  azureOpenAIApiEmbeddingsDeploymentName: azureEmbeddingDeploymentName,
  azureADTokenProvider: async () => await myToken.token(),
  defaultHeaders: HEADERS,
  verbose: true,
});

const credentials = new ClientSecretCredential(
  tenantId,
  clientId,
  clientSecret
);
const USER_AGENT_PREFIX = "langchainjs-azure-aisearch";
const vectorStore = new AzureAISearchVectorStore(
    embeddings,
    {
        client: new SearchClient(aisearchEndpoint, indexName, credentials, {
                        userAgentOptions: { userAgentPrefix: USER_AGENT_PREFIX },
                    }),
        search: { type: AzureAISearchQueryType.SimilarityHybrid },
    },
  );

export async function main() {
  const loader = new TextLoader("./xxx.md", "utf-8");
  const documents = await loader.load();
  const textSplitter = new MarkdownTextSplitter({
    chunkSize: 4096,
    chunkOverlap: 1024,
    keepSeparator: true,
  });
  const docs = await textSplitter.splitDocuments(documents);
  await vectorStore.addDocuments(docs);
  console.log(`Added ${docs.length} documents.`);

}

main().catch(console.error);

Description

I tried to use AzureAISearchVectorStore to embedding and store my docs in an index. I used azureADTokenProvider instead of apikey to authenticate. And I modified azureOpenAIBasePath to make sure the request url was correct.
However, BadRequestError: 400 status code (no body) occurred when I ran the code.
I wonder where the problem is and how I can solve it.
Appreciate it if you can help me out.

System Info

BadRequestError: 400 status code (no body)
at APIError.generate (file:///c:/xxx/node_modules/openai/error.mjs:41:20)
at AzureOpenAI.makeStatusError (file:///c:/xxx/node_modules/openai/core.mjs:295:25)
at AzureOpenAI.makeRequest (file:///c:/xxx/node_modules/openai/core.mjs:339:30)
at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
at async file:///c:/work/code/cloudrag/node_modules/@langchain/openai/dist/azure/embeddings.js:116:29
at async RetryOperation._fn (c:\xxx\node_modules\p-retry\index.js:50:12) {
status: 400,
headers: {
'content-length': '73',
'content-type': 'application/json',
date: 'Wed, 07 May 2025 08:26:40 GMT',
'request-id': 'f27f6674-be38-4071-9e03-71f4f225d4c7',
server: 'istio-envoy',
'x-envoy-upstream-service-time': '3'
},
request_id: undefined,
error: undefined,
code: undefined,
param: undefined,
type: undefined,
attemptNumber: 1,
retriesLeft: 6
}

[dosubot[bot]]

Hello @Columboom! 👋 I'm here to help you solve bugs, answer questions, and become a contributor while you wait for a human maintainer. Let's tackle this issue together!

To resolve the BadRequestError with a 400 status code when using AzureAISearchVectorStore and AzureOpenAIEmbeddings, ensure that all required fields and environment variables are correctly set. Specifically, check the following:

1. API Version: Verify that the azureOpenAIApiVersion is specified correctly. Ensure it matches the version supported by your Azure OpenAI deployment.

2. Base Path: Confirm that the azureOpenAIBasePath is correctly set. It should point to the correct endpoint for your Azure OpenAI deployment. Ensure that the proxy URL and path are correct.

3. Deployment Name: Check that the azureOpenAIApiEmbeddingsDeploymentName is correctly set. It should match the deployment name configured in your Azure portal.

4. Token Provider: Ensure that the azureADTokenProvider is correctly implemented and returns a valid token. You can log the token to verify its validity.

5. Search Client Configuration: Verify that the SearchClient is correctly configured with the right endpoint and credentials. Ensure that the aisearchEndpoint and indexName are correct.

6. Headers and Proxy: Double-check the headers and proxy settings to ensure they are correctly configured. Incorrect headers or proxy settings can lead to request failures.

If all configurations are correct and the error persists, consider checking the Azure portal for any additional settings or restrictions that might affect the API requests. Additionally, ensure that your Azure subscription is active and has the necessary permissions for the operations you're attempting.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}