@codehz/ai 0.1.7 → 0.2.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@codehz/ai",
-  "version": "0.1.7",
+  "version": "0.2.0",
   "type": "module",
   "module": "dist/index.mjs",
   "exports": {

package/src/helpers/usage-mapping.ts

@@ -1,7 +1,7 @@
 /**
  * Provider usage → canonical Usage 映射
  *
- * best-effort 提取 reasoning / cache / billable 等扩展字段。
+ * best-effort 提取 reasoning / cache 等扩展字段。
  */
 
 import type { Usage } from "../types/index.js";
@@ -20,27 +20,6 @@ function record(obj: Record<string, number | undefined>): Partial<Usage> {
   return out;
 }
 
-function billableFromOpenAIStyle(
-  inputTokens: number | undefined,
-  outputTokens: number | undefined,
-  cachedInputTokens: number | undefined,
-  reasoningTokens: number | undefined,
-): Pick<Usage, "billableInputTokens" | "billableOutputTokens"> {
-  let billableInputTokens: number | undefined;
-  if (inputTokens !== undefined) {
-    billableInputTokens =
-      cachedInputTokens !== undefined ? Math.max(0, inputTokens - cachedInputTokens) : inputTokens;
-  }
-
-  let billableOutputTokens: number | undefined;
-  if (outputTokens !== undefined) {
-    billableOutputTokens =
-      reasoningTokens !== undefined ? Math.max(0, outputTokens - reasoningTokens) : outputTokens;
-  }
-
-  return record({ billableInputTokens, billableOutputTokens });
-}
-
 /** OpenAI Chat Completions `usage` */
 export function usageFromChatCompletions(raw: {
   prompt_tokens?: number;
@@ -63,7 +42,6 @@ export function usageFromChatCompletions(raw: {
     totalTokens,
     cachedInputTokens,
     reasoningTokens,
-    ...billableFromOpenAIStyle(inputTokens, outputTokens, cachedInputTokens, reasoningTokens),
   });
 }
 
@@ -90,7 +68,6 @@ export function usageFromOpenAIResponses(raw: {
     totalTokens,
     cachedInputTokens,
     reasoningTokens,
-    ...billableFromOpenAIStyle(inputTokens, outputTokens, cachedInputTokens, reasoningTokens),
   });
 }
 
@@ -102,22 +79,17 @@ export function usageFromAnthropicMessages(raw: {
   cache_read_input_tokens?: number;
   [key: string]: unknown;
 }): Partial<Usage> {
-  const inputTokens = num(raw.input_tokens);
+  const uncachedInputTokens = num(raw.input_tokens);
   const outputTokens = num(raw.output_tokens);
   const cacheWriteInputTokens = num(raw.cache_creation_input_tokens);
   const cachedInputTokens = num(raw.cache_read_input_tokens);
 
-  const inputParts = [inputTokens, cacheWriteInputTokens, cachedInputTokens].filter(
+  const inputParts = [uncachedInputTokens, cacheWriteInputTokens, cachedInputTokens].filter(
     (n): n is number => n !== undefined,
   );
-  const summedInput = inputParts.length > 0 ? inputParts.reduce((sum, n) => sum + n, 0) : undefined;
+  const inputTokens = inputParts.length > 0 ? inputParts.reduce((sum, n) => sum + n, 0) : undefined;
   const totalTokens =
-    summedInput !== undefined && outputTokens !== undefined ? summedInput + outputTokens : undefined;
-
-  let billableInputTokens: number | undefined;
-  if (inputTokens !== undefined || cacheWriteInputTokens !== undefined) {
-    billableInputTokens = (inputTokens ?? 0) + (cacheWriteInputTokens ?? 0);
-  }
+    inputTokens !== undefined && outputTokens !== undefined ? inputTokens + outputTokens : undefined;
 
   return record({
     inputTokens,
@@ -125,12 +97,10 @@ export function usageFromAnthropicMessages(raw: {
     totalTokens,
     cachedInputTokens,
     cacheWriteInputTokens,
-    billableInputTokens,
-    billableOutputTokens: outputTokens,
   });
 }
 
-/** Ollama 流式 chunk（无 cache / reasoning 细分时仅填基础与 billable 镜像） */
+/** Ollama 流式 chunk */
 export function usageFromOllama(raw: {
   prompt_eval_count?: number;
   eval_count?: number;
@@ -144,7 +114,5 @@ export function usageFromOllama(raw: {
     inputTokens,
     outputTokens,
     totalTokens,
-    billableInputTokens: inputTokens,
-    billableOutputTokens: outputTokens,
   });
 }

package/src/types/response.ts

@@ -24,10 +24,7 @@ export type Usage = {
   cachedInputTokens?: number;
   /** Tokens written to prompt cache (Anthropic cache_creation_input_tokens) */
   cacheWriteInputTokens?: number;
-  /** Best-effort billable input (full-rate input; excludes discounted cache reads where known) */
-  billableInputTokens?: number;
-  /** Best-effort billable output (non-reasoning slice when provider gives reasoning breakdown) */
-  billableOutputTokens?: number;
+
 };
 
 export type BillingInfo = {