Merge pull request #963 from iceljc/master

add additional llm cost

Author: iceljc

src/Infrastructure/BotSharp.Abstraction/Conversations/Models/TokenStatsModel.cs

@@ -6,6 +6,7 @@ public class TokenStatsModel
     public string Model { get; set; }
     public string Prompt { get; set; }
     public int PromptCount { get; set; }
+    public int CachedPromptCount { get; set; }
     public int CompletionCount { get; set; }
     public AgentLlmConfig LlmConfig { get; set; }
 }

src/Infrastructure/BotSharp.Abstraction/MLTasks/Settings/LlmModelSetting.cs

@@ -62,12 +62,22 @@ public class LlmModelSetting
     /// </summary>
     public int Dimension { get; set; }
 
+    public LlmCost AdditionalCost { get; set; } = new();
+
     public override string ToString()
     {
         return $"[{Type}] {Name} {Endpoint}";
     }
 }
 
+public class LlmCost
+{
+    public float CachedPromptCost { get; set; } = 0f;
+    public float AudioPromptCost { get; set; } = 0f;
+    public float ReasoningCompletionCost { get; } = 0f;
+    public float AudioCompletionCost { get; } = 0f;
+}
+
 public enum LlmModelType
 {
     Text = 1,

src/Infrastructure/BotSharp.Core/Conversations/Services/TokenStatistics.cs

@@ -41,9 +41,11 @@ public void AddToken(TokenStatsModel stats, RoleDialogModel message)
         var settingsService = _services.GetRequiredService<ILlmProviderService>();
         var settings = settingsService.GetSetting(stats.Provider, _model);
 
-        var deltaPromptCost = stats.PromptCount / 1000f * settings.PromptCost;
+        var deltaPromptCost = (stats.PromptCount - stats.CachedPromptCount) / 1000f * settings.PromptCost;
+        var deltaCachedPromptCost = stats.CachedPromptCount / 1000f * (settings.AdditionalCost?.CachedPromptCost ?? 0f);
         var deltaCompletionCost = stats.CompletionCount / 1000f * settings.CompletionCost;
-        var deltaTotal = deltaPromptCost + deltaCompletionCost;
+
+        var deltaTotal = deltaPromptCost + deltaCachedPromptCost + deltaCompletionCost;
         _promptCost += deltaPromptCost;
         _completionCost += deltaCompletionCost;
 
@@ -53,6 +55,8 @@ public void AddToken(TokenStatsModel stats, RoleDialogModel message)
         stat.SetState("prompt_total", stats.PromptCount + inputCount, isNeedVersion: false, source: StateSource.Application);
         var outputCount = int.Parse(stat.GetState("completion_total", "0"));
         stat.SetState("completion_total", stats.CompletionCount + outputCount, isNeedVersion: false, source: StateSource.Application);
+        var cachedCount = int.Parse(stat.GetState("cached_prompt_total", "0"));
+        stat.SetState("cached_prompt_total", stats.CachedPromptCount + cachedCount, isNeedVersion: false, source: StateSource.Application);
 
         // Total cost
         var total_cost = float.Parse(stat.GetState("llm_total_cost", "0"));

src/Plugins/BotSharp.Plugin.OpenAI/Providers/Chat/ChatCompletionProvider.cs

@@ -84,6 +84,7 @@ public async Task<RoleDialogModel> GetChatCompletions(Agent agent, List<RoleDial
                 Provider = Provider,
                 Model = _model,
                 PromptCount = response.Value?.Usage?.InputTokenCount ?? 0,
+                CachedPromptCount = response.Value?.Usage?.InputTokenDetails?.CachedTokenCount ?? 0,
                 CompletionCount = response.Value?.Usage?.OutputTokenCount ?? 0
             });
         }