When using OpenAITextEmbeddingGenerationService in c# getting authorization error with openai key.

[giannik]

When running the following code to use text embedding model (text-embedding-ada-002) I am getting the followig error : Incorrect API key provided: text-emb**-002. The error occurs when calling kernel.Memory.SaveInformationAsync()

The key and password are valid because when i try it without AITextEmbeddingGenerationService but with OpenAIChatCompletionService then it works . See detailed error message below.

       var kernel = Kernel.Builder
                .WithLogger(_logger)
                 .WithOpenAITextCompletionService(llmProvider.Model, decryptedPassword)
                .WithOpenAITextEmbeddingGenerationService("ada", "text-embedding-ada-002", llmProvider.OpenAIKey )
                .WithMemoryStorage(new VolatileMemoryStore())
                .Build();

            //add memories
              var memoryCollectionName = "mycollectionid"; 

           foreach (var mapping in mappings)
            {
                await kernel.Memory.SaveInformationAsync(memoryCollectionName, id: mapping.Key, text:  mapping.Value);
            }

2023-07-21 14:00:10.1029|Default|00-e8aa119eabce9cdbf4f27206afa40010-e6b83ca39c5f4a39-00||Intelli.Embeddings.Search.VolatileSearch.Controllers.AdminController|ERROR|Error while executing query Microsoft.SemanticKernel.AI.AIException: Access denied: The request is not authorized, HTTP status: 401
 ---> Azure.RequestFailedException: Incorrect API key provided: text-emb**********-002. You can find your API key at https://platform.openai.com/account/api-keys.
Status: 401 (Unauthorized)
ErrorCode: invalid_api_key

Content:
{
    "error": {
        "message": "Incorrect API key provided: text-emb**********-002. You can find your API key at https://platform.openai.com/account/api-keys.",
        "type": "invalid_request_error",
        "param": null,
        "code": "invalid_api_key"
    }
}

Headers:
Date: Fri, 21 Jul 2023 11:00:09 GMT
Connection: keep-alive
Vary: REDACTED
X-Request-ID: REDACTED
Strict-Transport-Security: REDACTED
CF-Cache-Status: REDACTED
Server: cloudflare
CF-RAY: REDACTED
Alt-Svc: REDACTED
Content-Type: application/json; charset=utf-8
Content-Length: 272

   at Azure.Core.HttpPipelineExtensions.ProcessMessageAsync(HttpPipeline pipeline, HttpMessage message, RequestContext requestContext, CancellationToken cancellationToken)
   at Azure.AI.OpenAI.OpenAIClient.GetEmbeddingsAsync(String deploymentOrModelName, EmbeddingsOptions embeddingsOptions, CancellationToken cancellationToken)
   at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.RunRequestAsync[T](Func`1 request) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 374
   --- End of inner exception stack trace ---
   at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.RunRequestAsync[T](Func`1 request) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 411
   at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.InternalGetEmbeddingsAsync(IList`1 data, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 127
   at Microsoft.SemanticKernel.AI.Embeddings.EmbeddingGenerationExtensions.GenerateEmbeddingAsync[TValue,TEmbedding](IEmbeddingGeneration`2 generator, TValue value, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel.Abstractions\AI\Embeddings\EmbeddingGenerationServiceExtensions.cs:line 29
   at Microsoft.SemanticKernel.Memory.SemanticTextMemory.SaveInformationAsync(String collection, String text, String id, String description, String additionalMetadata, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel\Memory\SemanticTextMemory.cs:line 38
   at Intelli.Embeddings.Search.VolatileSearch.Core.Services.VolatileQueryService.SearchAsync(String indexName, String query, LLMProvider provider) in C:\IntelliDev\src\IntelliCore\Intelli.Embeddings.Search.VolatileSearch.Core\Services\VolatileQueryService.cs:line 98
   at Intelli.Embeddings.Search.VolatileSearch.Controllers.AdminController.QueryPost(AdminQueryViewModel model) in C:\IntelliDev\src\IntelliCore.Modules\Intelli.Embeddings.Search.VolatileSearch\Controllers\AdminController.cs:line 637    at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.RunRequestAsync[T](Func`1 request) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 411
   at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.InternalGetEmbeddingsAsync(IList`1 data, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 127
   at Microsoft.SemanticKernel.AI.Embeddings.EmbeddingGenerationExtensions.GenerateEmbeddingAsync[TValue,TEmbedding](IEmbeddingGeneration`2 generator, TValue value, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel.Abstractions\AI\Embeddings\EmbeddingGenerationServiceExtensions.cs:line 29
   at Microsoft.SemanticKernel.Memory.SemanticTextMemory.SaveInformationAsync(String collection, String text, String id, String description, String additionalMetadata, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel\Memory\SemanticTextMemory.cs:line 38
   at Intelli.Embeddings.Search.VolatileSearch.Core.Services.VolatileQueryService.SearchAsync(String indexName, String query, LLMProvider provider) in C:\IntelliDev\src\IntelliCore\Intelli.Embeddings.Search.VolatileSearch.Core\Services\VolatileQueryService.cs:line 98
   at Intelli.Embeddings.Search.VolatileSearch.Controllers.AdminController.QueryPost(AdminQueryViewModel model) in C:\IntelliDev\src\IntelliCore.Modules\Intelli.Embeddings.Search.VolatileSearch\Controllers\AdminController.cs:line 637

[giannik]

its weird in the exception trace to see :

Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.InternalGetEmbeddingsAsync

since im not using AzureSdk

[anthonypuppo]

The method has the following signature:

public static KernelBuilder WithOpenAITextEmbeddingGenerationService(this KernelBuilder builder, string modelId, string apiKey, string? orgId = null, string? serviceId = null, bool setAsDefault = false, HttpClient? httpClient = null)

So, you're passing "text-embedding-ada-002" as the second argument for apiKey which is why it's not working. Change to the below and it should work:

.WithOpenAITextEmbeddingGenerationService("text-embedding-ada-002", llmProvider.OpenAIKey)

If the intention was to use ada as the service ID you can adjust the call like so:

.WithOpenAITextEmbeddingGenerationService("text-embedding-ada-002", llmProvider.OpenAIKey, serviceId: "ada")

Hopefully that helps. Also, SK uses the Azure.AI.OpenAI package internally which is why you're seeing that namespace in the stack trace.

[giannik]

@anthonypuppo Thank you for response.

I tried your suggestion :

    var kernel = Kernel.Builder
               .WithLogger(_logger)
               .WithOpenAITextCompletionService(llmProvider.Model, decryptedPassword)
               .WithOpenAITextEmbeddingGenerationService("text-embedding-ada-002", llmProvider.OpenAIKey )
               .WithMemoryStorage(new VolatileMemoryStore())
              .Build();

but still got similar error message. I verified the openapi key is correct by running it against the chatcompletion service and it worked.

ErrorCode: invalid_api_key

Content: { "error": { "message": "Incorrect API key provided: CfDJ8PmP**EPBA. You can find your API key at https://platform.openai.com/account/api-keys.", "type": "invalid_request_error", "param": null, "code": "invalid_api_key" } }

Headers: Date: Fri, 21 Jul 2023 17:00:28 GMT Connection: keep-alive Vary: REDACTED X-Request-ID: REDACTED Strict-Transport-Security: REDACTED CF-Cache-Status: REDACTED Server: cloudflare CF-RAY: REDACTED Alt-Svc: REDACTED Content-Type: application/json; charset=utf-8 Content-Length: 448

at Azure.Core.HttpPipelineExtensions.ProcessMessageAsync(HttpPipeline pipeline, HttpMessage message, RequestContext requestContext, CancellationToken cancellationToken) at Azure.AI.OpenAI.OpenAIClient.GetEmbeddingsAsync(String deploymentOrModelName, EmbeddingsOptions embeddingsOptions, CancellationToken cancellationToken) at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.RunRequestAsync[T](Func1 request) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 374 --- End of inner exception stack trace --- at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.RunRequestAsync[T](Func1 request) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 411 at Microsoft.SemanticKernel.Connectors.AI.OpenAI.AzureSdk.ClientBase.InternalGetEmbeddingsAsync(IList1 data, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\Connectors\Connectors.AI.OpenAI\AzureSdk\ClientBase.cs:line 127 at Microsoft.SemanticKernel.AI.Embeddings.EmbeddingGenerationExtensions.GenerateEmbeddingAsync[TValue,TEmbedding](IEmbeddingGeneration2 generator, TValue value, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel.Abstractions\AI\Embeddings\EmbeddingGenerationServiceExtensions.cs:line 29 at Microsoft.SemanticKernel.Memory.SemanticTextMemory.SaveInformationAsync(String collection, String text, String id, String description, String additionalMetadata, CancellationToken cancellationToken) in C:\webapps\semantic-kernel\repo\dotnet\src\SemanticKernel\Memory\SemanticTextMemory.cs:line 38 at Intelli.Embeddings.Search.VolatileSearch.Core.Services.VolatileQueryService.SearchAsync(String indexName, String query, LLMProvider provider) in C:\IntelliDev\src\IntelliCore\Intelli.Embeddings.Search.VolatileSearch.Core\Services\VolatileQueryService.cs:line 110

[anthonypuppo]

var kernel = Kernel.Builder
    .WithLogger(_logger)
     .WithOpenAITextCompletionService(llmProvider.Model, decryptedPassword)
    .WithOpenAITextEmbeddingGenerationService("text-embedding-ada-002", llmProvider.OpenAIKey )
    .WithMemoryStorage(new VolatileMemoryStore())
    .Build();

You're providing two separate values for the API key in each of those calls (decryptedPassword and llmProvider.OpenAIKey). Since it looks like you're decrypting the key before use, I assume both should be decryptedPassword?

[giannik]

Im embarrassed! That was it. I was using the unencrypted openai key. Thank you for opening my eyes.