Add cached read and writes to stats and cost calculation for LiteLLM provider (#4206)

* Add cached read and writes to cost calculation for LiteLLM

* Fixed property issue

Author: mollux

src/api/providers/fetchers/litellm.ts

@@ -58,6 +58,8 @@ export async function getLiteLLMModels(apiKey: string, baseUrl: string): Promise
 					outputPrice: modelInfo.output_cost_per_token
 						? modelInfo.output_cost_per_token * 1000000
 						: undefined,
+					cacheWritesPrice: modelInfo.cache_creation_input_token_cost ? modelInfo.cache_creation_input_token_cost * 1000000 : undefined,
+					cacheReadsPrice: modelInfo.cache_read_input_token_cost ? modelInfo.cache_read_input_token_cost * 1000000 : undefined,
 					description: `${modelName} via LiteLLM proxy`,
 				}
 			}

src/api/providers/lite-llm.ts

@@ -3,6 +3,8 @@ import { Anthropic } from "@anthropic-ai/sdk" // Keep for type usage only
 
 import { litellmDefaultModelId, litellmDefaultModelInfo } from "@roo-code/types"
 
+import { calculateApiCostOpenAI } from "../../shared/cost"
+
 import { ApiHandlerOptions } from "../../shared/api"
 
 import { ApiStream, ApiStreamUsageChunk } from "../transform/stream"
@@ -66,7 +68,7 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 
 			for await (const chunk of completion) {
 				const delta = chunk.choices[0]?.delta
-				const usage = chunk.usage as OpenAI.CompletionUsage
+				const usage = chunk.usage as LiteLLMUsage
 
 				if (delta?.content) {
 					yield { type: "text", text: delta.content }
@@ -82,8 +84,12 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 					type: "usage",
 					inputTokens: lastUsage.prompt_tokens || 0,
 					outputTokens: lastUsage.completion_tokens || 0,
+					cacheWriteTokens: lastUsage.cache_creation_input_tokens || 0,
+					cacheReadTokens: lastUsage.prompt_tokens_details?.cached_tokens || 0,
 				}
 
+				usageData.totalCost = calculateApiCostOpenAI(info, usageData.inputTokens, usageData.outputTokens, usageData.cacheWriteTokens, usageData.cacheReadTokens)
+
 				yield usageData
 			}
 		} catch (error) {
@@ -119,3 +125,8 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 		}
 	}
 }
+
+// LiteLLM usage may include an extra field for Anthropic use cases.
+interface LiteLLMUsage extends OpenAI.CompletionUsage {
+	cache_creation_input_tokens?: number
+}