GeminiTextRequest.cs

using System;
using System.IO;
using System.Collections.Generic;
using System.Threading.Tasks;


namespace Gemini
{

    public class GeminiTextRequest: AIHttpClient
    {
        /** 
        <summary>
        Send query to Google Cloud Vertex AI via network and save the reply.
        </summary>
        <param name="strings"> 
        The first string must always be the text prompt.
        All subsequent strings are string representations of any type of data where currently are .png images.
        If you want to change image type, change "image/png" in InlineData.mimeType to the appropriate type.
        If you want to send video's, files stored on Google CLoud, or any other kind of data, please write your own data type as a property of Part according to the documentation:
        "https://cloud.google.com/dotnet/docs/reference/Google.Cloud.AIPlatform.V1/latest/Google.Cloud.AIPlatform.V1.Part"
        </param>
        */
        public async Task<GeminiTextResponse> SendMsg(params string[] strings)
        {
            FullRequest fullRequest = new FullRequest();
            List<Part> partsList = new List<Part>(); // Create a list to hold Part objects
            Content content = new Content { role = "user"};

            for (int i = 0; i < strings.Length; i++)
            {
                Part part = new Part();
                if (i == 0){
                    part.text = strings[0];

                    // Add the created Part object to the parts list
                    partsList.Add(part);

                }
                else 
                {
                   part.inlineData = new InlineData
                        {
                            mimeType = "image/png",
                            data = strings[i]
                        };
                    // Add the created Part object to the parts list
                    partsList.Add(part);
                }
            }
            Part[] partsArray = partsList.ToArray();
            content.parts = partsArray;
            fullRequest.contents = content;

            // Change the LLM parameters by uncomment below

            // GenerationConfig gc = new GenerationConfig();
            // gc.temperature = 0.1f;
            // gc.topP = 0.9f;
            // gc.topK = 20;
            // gc.candidateCount = 1;
            // gc.maxOutputTokens = 5;
            // fullRequest.generationConfig = gc;

            GeminiTextResponse geminiTextResponse = await base.PostAsync<FullRequest, GeminiTextResponse>(fullRequest, $"https://us-central1-aiplatform.googleapis.com/v1/projects/{PROJECT_ID}/locations/us-central1/publishers/google/models/{MODEL_ID}:{QUERY}");

            //if(geminiTextResponse!=null && geminiTextResponse.candidates.Length>0)
            //    contents.Add(geminiTextResponse.candidates[0].content);

            return geminiTextResponse;
        }
    }

    public class FullRequest
    {
        public Content contents;
        // public GenerationConfig generationConfig;
    }

    public partial class Content
    {
       public Part[] parts { get; set; }
    }

    public class Part
    {
       public string text { get; set; }
       public InlineData inlineData { get; set; }
    }

    public class InlineData
    {
        public string mimeType { get; set; }
        public string data { get; set; } // Base64-encoded data
    }

    public class GenerationConfig
    {
        public float temperature { get; set; }
        public float topP { get; set; }
        public float topK { get; set; }
        public int candidateCount { get; set; }
        public int maxOutputTokens { get; set; }
    }
}