优化 gpt说话人流程 计算token数量流程

VideoAnalysis/Program.cs

@@ -8,12 +8,15 @@ using VideoAnalysisCore.AICore.ChatGPT.KIMI;
 using VideoAnalysisCore.AICore.SherpaOnnx;
 using SqlSugar;
 
+
+
 namespace Learn.VideoAnalysis
 {
     public class Program
     {
         public static void Main(string[] args)
         {
+
             var builder = WebApplication.CreateBuilder(args);
 
             // Add services to the container.

VideoAnalysisCore/AICore/ChatGPT/Dto/CallGPTRes.cs

@@ -0,0 +1,40 @@
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+using VideoAnalysisCore.Enum;
+using VideoAnalysisCore.Model.Dto;
+
+namespace VideoAnalysisCore.AICore.ChatGPT.Dto
+{
+    public class Assessment 
+    {
+    
+    }
+    public class CallGPTRes
+    {
+        /// <summary>
+        /// 教师发言时间
+        /// <para>秒</para>
+        /// </summary>
+        public decimal TeacherSpeaking { get; set; } = 0;
+        /// <summary>
+        /// 学生发言时间
+        /// <para>秒</para>
+        /// </summary>
+        public decimal StudentSpeaking { get; set; } = 0;
+        /// <summary>
+        /// 视频时间轴
+        /// </summary>
+        public IEnumerable<TimeBase>? TimeBase { get; set; }
+        /// <summary>
+        /// GPT模型id
+        /// </summary>
+        public string GPTModel { get; set; } = string.Empty;
+        /// <summary>
+        /// AI综合评估
+        /// </summary>
+        public Assessment Assessment { get; set; }
+    }
+}

VideoAnalysisCore/AICore/ChatGPT/KIMI/KIMI_GPT.cs

@@ -9,6 +9,9 @@ using System.Text;
 using FFmpeg.NET.Services;
 using Microsoft.Extensions.Primitives;
 using VideoAnalysisCore.AICore.ChatGPT.Dto;
+using System.ComponentModel.DataAnnotations;
+using VideoAnalysisCore.Enum;
+using System.Reflection;
 
 namespace VideoAnalysisCore.AICore.ChatGPT.KIMI
 {
@@ -50,15 +53,21 @@ namespace VideoAnalysisCore.AICore.ChatGPT.KIMI
                 criteriaBuilder.Append(item.NamePrompt);
                 criteriaBuilder.Append("返回bool值|");
             }
-            criteriaBuilder.Append("100:分析授课中使用的高频词10个|");
+            //拼接枚举提问
-            criteriaBuilder.Append("101:分析授课中引发学生思考问题的时间段数组|");
+            foreach (var value in System.Enum.GetValues(typeof(QuestionTypeEnum)))
-            criteriaBuilder.Append("102:分析授课中让学生产生讨论欲望的时间段|");
+            {
-            criteriaBuilder.Append("104:分析授课中巧妙地引入学习主题的时间段|");
+                var enumValue = (QuestionTypeEnum)value;
-            criteriaBuilder.Append("105:分析授课中教学语言简明清晰准确生动的时间段|");
+                var displayAttribute = enumValue.GetType()
-            criteriaBuilder.Append("106:分析授课中教师激发学生思考的时间段|");
+                    .GetField(enumValue.ToString())?
-            criteriaBuilder.Append("107:统计授课中教师提问类型的次数 类型重复回答,老师追问,简单性表演,老师补充答案,表扬并补充答案|");
+                    .GetCustomAttribute<DisplayAttribute>();
+                if (displayAttribute == null) continue;
+                criteriaBuilder.Append(enumValue.GetHashCode());
+                criteriaBuilder.Append(":");
+                criteriaBuilder.Append(displayAttribute.Prompt);
+                criteriaBuilder.Append("|");
+            }
 
-            var resFormat = "问题编号:int,结果:array,问题解释:string";
+            var resFormat = "问题编号:int,结果:array|bool,问题解释:string";
             var postMessages = 
                 $"以下是一段音频的字幕,分析这段字幕(格式 说话人:开始秒:结束秒:内容|下一段字幕)." +
                 $"来简明的回答提出的问题 问题列表 {criteriaBuilder} " +
@@ -67,24 +76,32 @@ namespace VideoAnalysisCore.AICore.ChatGPT.KIMI
             var modelsResp = await moonshotClient.ListModels();
             if (modelsResp is null || modelsResp.data.Count == 0)
                 throw new Exception("未获取KIMI模型类型");
+            var reqTokenCount =await moonshotClient.GetAsTiMateTokenCount(postMessages);
+            var modelId = reqTokenCount > 32 * 1000 ? "moonshot-v1-128k" : "moonshot-v1-32k";
             var chatRep = new ChatReq
             {
-                max_tokens = postMessages.Length * 2,
+                max_tokens =1000 * 31,
                 temperature = 0.3,
                 frequency_penalty = 0,
                 presence_penalty = 0,
-                model = modelsResp.data.First(s=>s.id.Contains("v1-32k")).id,
+                model = modelId,
                 messages = new List<MessagesItem>(){
                     new MessagesItem(postMessages,"system"),
+                    //todo 规定返回json格式
+                    //new MessagesItem(postMessages,"assistant"),
                 }
             };
             var chatResp = await moonshotClient.Chat(chatRep);
-            var chatResContent = chatResp?.choices.FirstOrDefault()?.message.content;
+           if(chatResp is null || chatResp.error != null)
-            if(chatResp is null || string.IsNullOrEmpty(chatResContent))
                 throw new Exception($"KIMI模型返回异常 Chat 请求参数: {JsonSerializer.Serialize(chatRep)} " +
-                    $" chatResp {chatResp} chatResContent {chatResContent}");
+                    $" chatResp {JsonSerializer.Serialize(chatResp)}");
+            var chatResContent = chatResp?.choices.FirstOrDefault()?.message.content;
             var questionRes = JsonSerializer.Deserialize<QuestionRes[]>(chatResContent);
-            //分析gpt返回结果
+            //todo 分析gpt返回结果
+            //todo 分析上课时间段情况 分析 独立学习 小组合作 随堂练习等情况
+
+
+
         }
     }
 }

VideoAnalysisCore/AICore/ChatGPT/KIMI/MoonshotClient.cs

@@ -56,12 +56,16 @@ namespace VideoAnalysisCore.AICore.ChatGPT.KIMI
         {
             var requestBody = Newtonsoft.Json.JsonConvert.SerializeObject(chatReq);
             var chatResp = await PostJsonStreamAsync("/v1/chat/completions", requestBody);
+            var resStr = chatResp.Content.ReadAsStringAsync();
             return await chatResp.Content.ReadFromJsonAsync<ChatRes>();
         }
 
         /// <summary>
-        ///  Get as timate token count
+        /// 计算token长度
         /// </summary>
+        /// <param name="chatReqText">文本</param>
+        /// <returns></returns>
+        /// <exception cref="Exception"></exception>
         public async Task<int?> GetAsTiMateTokenCount(string chatReqText)
         {
             var response = await PostJsonAsync("/v1/tokenizers/estimate-token-count", chatReqText);

VideoAnalysisCore/AICore/ChatGPT/KIMI/MoonshotModel.cs

@@ -16,6 +16,7 @@
         /// 
         /// </summary>
         public string role { get; set; }
+        public bool partial { get; set; } = false;
 
         /// <summary>
         /// 
@@ -134,8 +135,21 @@
     }
 
 
+    public class ChatResError
+    {
+        /// <summary>
+        /// 错误信息
+        /// </summary>
+        public string message { get; set; } = string.Empty;
+        /// <summary>
+        /// 错误类型
+        /// </summary>
+        public string type { get; set; } = string.Empty;
+
+    }
     public class ChatRes
     {
+        public ChatResError? error { get; set; }
         public string id { get; set; }
         public int created { get; set; }
         /// <summary>

VideoAnalysisCore/Common/AppCommon.cs

@@ -164,10 +164,11 @@ namespace VideoAnalysisCore.Common
                         else
                             studentSpeaking += speakerRes.Total;
                         spList.Add(speakerRes.SpeakerIndex);
-                        break;
                     }
                 }
-                results.Add(segment, spList);
+                var sp = spList.Distinct().ToList();
+                if(sp.Count>0)
+                    results.Add(segment, sp);
             }
             //拼接 提示词字幕源
             var stringBuilder = new StringBuilder();
@@ -182,8 +183,6 @@ namespace VideoAnalysisCore.Common
                 stringBuilder.Append(item.Key.Text);
                 stringBuilder.Append("|");
             }
-            //todo 分析上课时间段情况 分析 独立学习 小组合作 随堂练习等情况
-
             return new TotalCaptionsDto
             {
                 StudentSpeaking = studentSpeaking,
@@ -193,7 +192,8 @@ namespace VideoAnalysisCore.Common
                 {
                     Start = s.Key.Start.TotalSeconds,
                     End = s.Key.End.TotalSeconds,
-                    Type = s.Value.Count == 1 && s.Value.First() == techerId 
+                    Content = s.Key.Text,
+                    TimeBaseType = s.Value.Count == 1 && s.Value.First() == techerId 
                         ? TimeBaseTypeEnum.教师讲授
                         : TimeBaseTypeEnum.互动交流
                 })

VideoAnalysisCore/Enum/QuestionTypeEnum.cs

@@ -0,0 +1,20 @@
+using System;
+using System.Collections.Generic;
+using System.ComponentModel.DataAnnotations;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace VideoAnalysisCore.Enum
+{
+    enum QuestionTypeEnum
+    {
+
+        [Display(Prompt = "分析授课中使用的高频词10个")]
+        高频词 =100,
+        [Display(Prompt = "分析字幕中每5分钟的一个概览 返回结构 array[0]=概览 时间段")]
+        时间段概览 = 101,
+        [Display(Prompt = "统计授课中教师提问类型的次数 类型重复回答,老师追问,简单性表演,老师补充答案,表扬并补充答案")]
+        提问类型 = 102,
+    }
+}

VideoAnalysisCore/Model/Dto/SpeakerCaptionsDto.cs

@@ -20,9 +20,13 @@ namespace VideoAnalysisCore.Model.Dto
         /// </summary>
         public double End { get; set; }
         /// <summary>
+        /// 字幕内容
+        /// </summary>
+        public string Content { get; set; }
+        /// <summary>
         /// 时间段 类型
         /// </summary>
-        public TimeBaseTypeEnum Type { get; set; }
+        public TimeBaseTypeEnum TimeBaseType { get; set; }
 
     }
     public class TotalCaptionsDto