@axlsdk/axl 0.7.5 → 0.7.6

package/dist/index.d.cts

@@ -108,6 +108,8 @@ type StreamChunk = {
         reasoning_tokens?: number;
         cached_tokens?: number;
     };
+    /** Estimated cost in USD for this call, computed the same way as ProviderResponse.cost. */
+    cost?: number;
     /** Provider-specific opaque metadata (e.g. raw Gemini parts with thought signatures). */
     providerMetadata?: Record<string, unknown>;
 };

package/dist/index.d.ts

@@ -108,6 +108,8 @@ type StreamChunk = {
         reasoning_tokens?: number;
         cached_tokens?: number;
     };
+    /** Estimated cost in USD for this call, computed the same way as ProviderResponse.cost. */
+    cost?: number;
     /** Provider-specific opaque metadata (e.g. raw Gemini parts with thought signatures). */
     providerMetadata?: Record<string, unknown>;
 };

package/dist/index.js

@@ -165,29 +165,32 @@ async function fetchWithRetry(input, init, maxRetries = MAX_RETRIES) {
 
 // src/providers/openai.ts
 var OPENAI_PRICING = {
-  "gpt-4o": [25e-7, 1e-5],
-  "gpt-4o-mini": [15e-8, 6e-7],
-  "gpt-4-turbo": [1e-5, 3e-5],
-  "gpt-4": [3e-5, 6e-5],
-  "gpt-3.5-turbo": [5e-7, 15e-7],
-  "gpt-5": [125e-8, 1e-5],
-  "gpt-5-mini": [25e-8, 2e-6],
-  "gpt-5-nano": [5e-8, 4e-7],
-  "gpt-5.1": [125e-8, 1e-5],
-  "gpt-5.2": [175e-8, 14e-6],
-  "gpt-5.3": [175e-8, 14e-6],
-  "gpt-5.4": [25e-7, 15e-6],
-  "gpt-5.4-pro": [3e-5, 18e-5],
-  o1: [15e-6, 6e-5],
-  "o1-mini": [3e-6, 12e-6],
-  "o1-pro": [15e-5, 6e-4],
-  o3: [1e-5, 4e-5],
-  "o3-mini": [11e-7, 44e-7],
-  "o3-pro": [2e-5, 8e-5],
-  "o4-mini": [11e-7, 44e-7],
-  "gpt-4.1": [2e-6, 8e-6],
-  "gpt-4.1-mini": [4e-7, 16e-7],
-  "gpt-4.1-nano": [1e-7, 4e-7]
+  // gpt-4o era — cache reads at 50% of input rate
+  "gpt-4o": [25e-7, 1e-5, 0.5],
+  "gpt-4o-mini": [15e-8, 6e-7, 0.5],
+  "gpt-4-turbo": [1e-5, 3e-5, 0.5],
+  "gpt-4": [3e-5, 6e-5, 0.5],
+  "gpt-3.5-turbo": [5e-7, 15e-7, 0.5],
+  o1: [15e-6, 6e-5, 0.5],
+  "o1-mini": [3e-6, 12e-6, 0.5],
+  "o1-pro": [15e-5, 6e-4, 0.5],
+  // gpt-4.1 / o3 / o4 era — cache reads at 25% of input rate
+  "gpt-4.1": [2e-6, 8e-6, 0.25],
+  "gpt-4.1-mini": [4e-7, 16e-7, 0.25],
+  "gpt-4.1-nano": [1e-7, 4e-7, 0.25],
+  o3: [1e-5, 4e-5, 0.25],
+  "o3-mini": [11e-7, 44e-7, 0.25],
+  "o3-pro": [2e-5, 8e-5, 0.25],
+  "o4-mini": [11e-7, 44e-7, 0.25],
+  // gpt-5 era — cache reads at 10% of input rate
+  "gpt-5": [125e-8, 1e-5, 0.1],
+  "gpt-5-mini": [25e-8, 2e-6, 0.1],
+  "gpt-5-nano": [5e-8, 4e-7, 0.1],
+  "gpt-5.1": [125e-8, 1e-5, 0.1],
+  "gpt-5.2": [175e-8, 14e-6, 0.1],
+  "gpt-5.3": [175e-8, 14e-6, 0.1],
+  "gpt-5.4": [25e-7, 15e-6, 0.1],
+  "gpt-5.4-pro": [3e-5, 18e-5, 0.1]
 };
 var PRICING_KEYS_BY_LENGTH = Object.keys(OPENAI_PRICING).sort((a, b) => b.length - a.length);
 function estimateOpenAICost(model, promptTokens, completionTokens, cachedTokens) {
@@ -201,9 +204,9 @@ function estimateOpenAICost(model, promptTokens, completionTokens, cachedTokens)
     }
   }
   if (!pricing) return 0;
-  const [inputRate, outputRate] = pricing;
+  const [inputRate, outputRate, cacheMultiplier] = pricing;
   const cached = cachedTokens ?? 0;
-  const inputCost = (promptTokens - cached) * inputRate + cached * inputRate * 0.5;
+  const inputCost = (promptTokens - cached) * inputRate + cached * inputRate * cacheMultiplier;
   return inputCost + completionTokens * outputRate;
 }
 function isOSeriesModel(model) {
@@ -313,7 +316,7 @@ var OpenAIProvider = class {
     if (!res.body) {
       throw new Error("OpenAI stream response has no body");
     }
-    yield* this.parseSSEStream(res.body);
+    yield* this.parseSSEStream(res.body, options.model);
   }
   // ---------------------------------------------------------------------------
   // Internal helpers
@@ -387,7 +390,7 @@ var OpenAIProvider = class {
     if (msg.tool_call_id) out.tool_call_id = msg.tool_call_id;
     return out;
   }
-  async *parseSSEStream(body) {
+  async *parseSSEStream(body, model) {
     const reader = body.getReader();
     const decoder = new TextDecoder();
     let buffer = "";
@@ -404,7 +407,16 @@ var OpenAIProvider = class {
           const trimmed = line.trim();
           if (!trimmed || trimmed.startsWith(":")) continue;
           if (trimmed === "data: [DONE]") {
-            yield { type: "done", usage: usageData };
+            yield {
+              type: "done",
+              usage: usageData,
+              cost: usageData ? estimateOpenAICost(
+                model,
+                usageData.prompt_tokens,
+                usageData.completion_tokens,
+                usageData.cached_tokens
+              ) : void 0
+            };
             return;
           }
           if (trimmed.startsWith("data: ")) {
@@ -447,7 +459,16 @@ var OpenAIProvider = class {
           }
         }
       }
-      yield { type: "done", usage: usageData };
+      yield {
+        type: "done",
+        usage: usageData,
+        cost: usageData ? estimateOpenAICost(
+          model,
+          usageData.prompt_tokens,
+          usageData.completion_tokens,
+          usageData.cached_tokens
+        ) : void 0
+      };
     } finally {
       reader.releaseLock();
     }
@@ -705,6 +726,7 @@ var OpenAIResponsesProvider = class {
     const decoder = new TextDecoder();
     let buffer = "";
     const callIdMap = /* @__PURE__ */ new Map();
+    let eventType = "";
     try {
       while (true) {
         const { done, value } = await reader.read();
@@ -712,7 +734,6 @@ var OpenAIResponsesProvider = class {
         buffer += decoder.decode(value, { stream: true });
         const lines = buffer.split("\n");
         buffer = lines.pop() ?? "";
-        let eventType = "";
         for (const line of lines) {
           const trimmed = line.trim();
           if (!trimmed || trimmed.startsWith(":")) continue;
@@ -780,7 +801,17 @@ var OpenAIResponsesProvider = class {
         } : void 0;
         const reasoningItems = response?.output?.filter((item) => item.type === "reasoning") ?? [];
         const providerMetadata = reasoningItems.length > 0 ? { openaiReasoningItems: reasoningItems } : void 0;
-        return { type: "done", usage, providerMetadata };
+        return {
+          type: "done",
+          usage,
+          cost: usage ? estimateOpenAICost(
+            model,
+            usage.prompt_tokens,
+            usage.completion_tokens,
+            usage.cached_tokens
+          ) : void 0,
+          providerMetadata
+        };
       }
       case "response.failed": {
         const errorMsg = data.response?.error?.message ?? data.response?.status_details?.error?.message ?? "Unknown error";
@@ -913,7 +944,7 @@ var AnthropicProvider = class {
     if (!res.body) {
       throw new Error("Anthropic stream response has no body");
     }
-    yield* this.parseSSEStream(res.body);
+    yield* this.parseSSEStream(res.body, options.model);
   }
   // ---------------------------------------------------------------------------
   // Internal: request building
@@ -1156,13 +1187,14 @@ ${jsonInstruction}` : jsonInstruction;
   // ---------------------------------------------------------------------------
   // Internal: SSE stream parsing
   // ---------------------------------------------------------------------------
-  async *parseSSEStream(body) {
+  async *parseSSEStream(body, model) {
     const reader = body.getReader();
     const decoder = new TextDecoder();
     let buffer = "";
     let currentToolId = "";
     let currentToolName = "";
     let usage;
+    let cacheWrite = 0;
     try {
       while (true) {
         const { done, value } = await reader.read();
@@ -1219,7 +1251,7 @@ ${jsonInstruction}` : jsonInstruction;
             case "message_start": {
               if (event.message?.usage) {
                 const cacheRead = event.message.usage.cache_read_input_tokens ?? 0;
-                const cacheWrite = event.message.usage.cache_creation_input_tokens ?? 0;
+                cacheWrite = event.message.usage.cache_creation_input_tokens ?? 0;
                 const inputTokens = (event.message.usage.input_tokens ?? 0) + cacheRead + cacheWrite;
                 usage = {
                   prompt_tokens: inputTokens,
@@ -1250,13 +1282,33 @@ ${jsonInstruction}` : jsonInstruction;
               if (usage) {
                 usage.total_tokens = usage.prompt_tokens + usage.completion_tokens;
               }
-              yield { type: "done", usage };
+              yield {
+                type: "done",
+                usage,
+                cost: usage ? estimateAnthropicCost(
+                  model,
+                  usage.prompt_tokens,
+                  usage.completion_tokens,
+                  usage.cached_tokens,
+                  cacheWrite
+                ) : void 0
+              };
               return;
             }
           }
         }
       }
-      yield { type: "done", usage };
+      yield {
+        type: "done",
+        usage,
+        cost: usage ? estimateAnthropicCost(
+          model,
+          usage.prompt_tokens,
+          usage.completion_tokens,
+          usage.cached_tokens,
+          cacheWrite
+        ) : void 0
+      };
     } finally {
       reader.releaseLock();
     }
@@ -1383,7 +1435,7 @@ var GeminiProvider = class {
     if (!res.body) {
       throw new Error("Gemini stream response has no body");
     }
-    yield* this.parseSSEStream(res.body);
+    yield* this.parseSSEStream(res.body, options.model);
   }
   // ---------------------------------------------------------------------------
   // Internal: request building
@@ -1664,7 +1716,7 @@ var GeminiProvider = class {
   // ---------------------------------------------------------------------------
   // Internal: SSE stream parsing
   // ---------------------------------------------------------------------------
-  async *parseSSEStream(body) {
+  async *parseSSEStream(body, model) {
     const reader = body.getReader();
     const decoder = new TextDecoder();
     let buffer = "";
@@ -1720,7 +1772,17 @@ var GeminiProvider = class {
         }
       }
       const providerMetadata = accumulatedParts.length > 0 ? { geminiParts: accumulatedParts } : void 0;
-      yield { type: "done", usage, providerMetadata };
+      yield {
+        type: "done",
+        usage,
+        cost: usage ? estimateGeminiCost(
+          model,
+          usage.prompt_tokens,
+          usage.completion_tokens,
+          usage.cached_tokens
+        ) : void 0,
+        providerMetadata
+      };
     } finally {
       reader.releaseLock();
     }
@@ -2430,7 +2492,8 @@ Please fix and try again.`;
               response = {
                 content: content2,
                 tool_calls: void 0,
-                usage: chunk.usage
+                usage: chunk.usage,
+                cost: chunk.cost
               };
             }
           }