Implementation:Microsoft Semantic kernel MistralAI Embeddings TestData

Knowledge Sources	Microsoft_Semantic_kernel
Domains	MistralAI, Embeddings, Unit_Testing, Test_Data
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete mock MistralAI embeddings API response JSON file used in unit tests, provided by the Connectors.MistralAI.UnitTests project.

Description

embeddings_response.json is a test fixture file containing a mock response from the MistralAI embeddings API. At 2072 lines, this is the largest of the connector test data embedding files. The JSON structure follows the OpenAI-compatible embeddings API format that MistralAI uses:

• id -- A unique request identifier string (e.g., "994dfff08057489aa745f50f9ce07f22")
• object -- The response type, set to "list"
• data -- An array of embedding objects, each containing:
  • object -- Set to "embedding" indicating the item type
  • embedding -- An array of floating-point numbers representing the embedding vector

The embedding values are typical float32 numbers (e.g., -0.0249176025390625, 0.042816162109375) representing dense vector representations from MistralAI's embedding models. The OpenAI-compatible format used by MistralAI distinguishes this from Google's format (which uses embeddings[].values) and HuggingFace's format (which uses bare nested arrays).

This file is consumed by multiple test classes: MistralClientTests, MistralAIEmbeddingGeneratorTests, and MistralAITextEmbeddingGenerationServiceTests.

Usage

This file is loaded at test time through the MistralTestBase.GetTestResponseAsString() method, which reads from the ./TestData/ directory using File.ReadAllText(). It is also loaded directly in MistralClientTests via a helper that constructs a mock HTTP handler. Developers working on the MistralAI connector would use this file when writing embedding-related unit tests or when the MistralAI API response format changes.

Code Reference

Source Location

• Repository: Microsoft_Semantic_kernel
• File: dotnet/src/Connectors/Connectors.MistralAI.UnitTests/TestData/embeddings_response.json
• Lines: 1-2072

Signature

{
    "id": "994dfff08057489aa745f50f9ce07f22",
    "object": "list",
    "data": [
        {
            "object": "embedding",
            "embedding": [
                -0.0249176025390625,
                -0.00296783447265625,
                0.042816162109375,
                0.0162811279296875,
                0.0435791015625,
                0.03594970703125,
                0.048065185546875,
                0.01406097412109375,
                -0.039581298828125,
                -0.01355743408203125,
                -0.054718017578125,
                0.03143310546875,
                -0.0259857177734375,
                -0.021820068359375,
                -0.0282745361328125,
                0.0032672882080078125,
                -0.007137298583984375,
                0.04217529296875,
                0.029449462890625,
                0.035858154296875
            ]
        }
    ]
}

Import

// Via the MistralTestBase base class:
using SemanticKernel.Connectors.MistralAI.UnitTests;

public class MistralAIEmbeddingGeneratorTests : MistralTestBase
{
    [Fact]
    public async Task ShouldReturnEmbeddings()
    {
        // Load the test response
        var content = this.GetTestResponseAsString("embeddings_response.json");
        // ...
    }
}

// MistralTestBase implementation:
protected string GetTestResponseAsString(string fileName)
{
    return File.ReadAllText($"./TestData/{fileName}");
}

I/O Contract

Inputs

Name	Type	Required	Description
N/A (static file)	JSON file	N/A	This is a static test fixture; it has no runtime inputs

Outputs

Name	Type	Description
id	string	Unique request identifier for the embeddings API call
object	string	Response type identifier, always "list" for embedding responses
data	array	Array of embedding result objects
data[].object	string	Item type identifier, always "embedding"
data[].embedding	float[]	Array of floating-point embedding vector components (hundreds of dimensions)

Usage Examples

Loading in MistralAI Embedding Tests

using System.Net.Http;
using Xunit;

public class MistralAITextEmbeddingGenerationServiceTests : MistralTestBase
{
    [Fact]
    public async Task ItReturnsEmbeddingsCorrectly()
    {
        // Arrange
        var content = this.GetTestResponseAsString("embeddings_response.json");
        this.DelegatingHandler = new AssertingDelegatingHandler(
            "https://api.mistral.ai/v1/embeddings", content);
        this.HttpClient = new HttpClient(this.DelegatingHandler);

        var service = new MistralAITextEmbeddingGenerationService(
            "mistral-embed", "key", httpClient: this.HttpClient);

        // Act
        var embeddings = await service
            .GenerateEmbeddingsAsync(new List<string> { "test input" });

        // Assert
        Assert.NotNull(embeddings);
        Assert.Single(embeddings);
    }
}

Using in MistralClient Integration Tests

// In MistralClientTests.cs:
var client = this.CreateMistralClient(
    "mistral-tiny",
    "https://api.mistral.ai/v1/embeddings",
    "embeddings_response.json"
);

Related Pages

• Environment:Microsoft_Semantic_kernel_DotNet_SDK_Environment
• Microsoft_Semantic_kernel_Google_Embeddings_TestData -- Mock Google AI embeddings data (Google-specific JSON schema)
• Microsoft_Semantic_kernel_Vertex_Embeddings_TestData -- Mock Vertex AI embeddings data (includes statistics metadata)
• Microsoft_Semantic_kernel_HuggingFace_Embeddings_TestData -- Mock HuggingFace embeddings data (bare nested array format)
• Microsoft_Semantic_kernel_HuggingFace_TextGeneration_TestData -- Mock HuggingFace text generation data (different task type)