@posthog/ai 8.6.7 → 8.7.0

package/dist/openai/index.mjs

@@ -544,7 +544,7 @@ function formatOpenAIResponsesInput(input, instructions) {
   return messages;
 }
 
-var version = "8.6.7";
+var version = "8.7.0";
 
 const DEFAULT_MAX_DEPTH = 3;
 const MAX_STACK_LINES = 20;
@@ -842,6 +842,18 @@ function extractRequestId(result) {
   return result?._request_id ?? undefined;
 }
 
+/**
+ * Reads `cache_write_tokens` from a usage details object — Chat Completions'
+ * `prompt_tokens_details` or the Responses API's `input_tokens_details`, both of
+ * which carry the field — and returns 0 when it is absent. A defensive reader
+ * (mirroring `extractRequestId`) that tolerates the loosely-typed usage shapes
+ * OpenAI-compatible providers return, used to populate
+ * `$ai_cache_creation_input_tokens`.
+ */
+function extractCacheWriteTokens(details) {
+  return details?.cache_write_tokens ?? 0;
+}
+
 /**
  * Assembles the `$ai_provider_metadata` blob for OpenAI / Azure OpenAI events.
  * Provider-specific fields (system fingerprint, request id) live here rather
@@ -1389,7 +1401,8 @@ let WrappedCompletions$1 = class WrappedCompletions extends AzureOpenAI.Chat.Com
                     inputTokens: chunk.usage.prompt_tokens ?? 0,
                     outputTokens: chunk.usage.completion_tokens ?? 0,
                     reasoningTokens: chunk.usage.completion_tokens_details?.reasoning_tokens ?? 0,
-                    cacheReadInputTokens: chunk.usage.prompt_tokens_details?.cached_tokens ?? 0
+                    cacheReadInputTokens: chunk.usage.prompt_tokens_details?.cached_tokens ?? 0,
+                    cacheCreationInputTokens: extractCacheWriteTokens(chunk.usage.prompt_tokens_details)
                   };
                 }
               }
@@ -1500,7 +1513,8 @@ let WrappedCompletions$1 = class WrappedCompletions extends AzureOpenAI.Chat.Com
               inputTokens: result.usage?.prompt_tokens ?? 0,
               outputTokens: result.usage?.completion_tokens ?? 0,
               reasoningTokens: result.usage?.completion_tokens_details?.reasoning_tokens ?? 0,
-              cacheReadInputTokens: result.usage?.prompt_tokens_details?.cached_tokens ?? 0
+              cacheReadInputTokens: result.usage?.prompt_tokens_details?.cached_tokens ?? 0,
+              cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.prompt_tokens_details)
             },
             completionId: result.id,
             providerMetadata: buildProviderMetadata({
@@ -1561,6 +1575,7 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
         outputTokens: result.usage?.output_tokens ?? 0,
         reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
         cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+        cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
         rawUsage: result.usage
       },
       stopReason: result.status ?? undefined,
@@ -1637,7 +1652,8 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
                     inputTokens: chunk.response.usage.input_tokens ?? 0,
                     outputTokens: chunk.response.usage.output_tokens ?? 0,
                     reasoningTokens: chunk.response.usage.output_tokens_details?.reasoning_tokens ?? 0,
-                    cacheReadInputTokens: chunk.response.usage.input_tokens_details?.cached_tokens ?? 0
+                    cacheReadInputTokens: chunk.response.usage.input_tokens_details?.cached_tokens ?? 0,
+                    cacheCreationInputTokens: extractCacheWriteTokens(chunk.response.usage.input_tokens_details)
                   };
                 }
               }
@@ -1730,6 +1746,7 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
               outputTokens: result.usage?.output_tokens ?? 0,
               reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
               cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+              cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
               rawUsage: result.usage
             },
             stopReason: result.status ?? undefined,
@@ -1845,6 +1862,7 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
           outputTokens: result.usage?.output_tokens ?? 0,
           reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
           cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+          cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
           rawUsage: result.usage
         },
         stopReason: result.status ?? undefined,
@@ -2091,7 +2109,8 @@ class WrappedCompletions extends Completions {
                     inputTokens: chunk.usage.prompt_tokens ?? 0,
                     outputTokens: chunk.usage.completion_tokens ?? 0,
                     reasoningTokens: chunk.usage.completion_tokens_details?.reasoning_tokens ?? 0,
-                    cacheReadInputTokens: chunk.usage.prompt_tokens_details?.cached_tokens ?? 0
+                    cacheReadInputTokens: chunk.usage.prompt_tokens_details?.cached_tokens ?? 0,
+                    cacheCreationInputTokens: extractCacheWriteTokens(chunk.usage.prompt_tokens_details)
                   };
                 }
               }
@@ -2148,6 +2167,7 @@ class WrappedCompletions extends Completions {
                   outputTokens: usage.outputTokens,
                   reasoningTokens: usage.reasoningTokens,
                   cacheReadInputTokens: usage.cacheReadInputTokens,
+                  cacheCreationInputTokens: usage.cacheCreationInputTokens,
                   webSearchCount: usage.webSearchCount,
                   rawUsage: rawUsageData
                 },
@@ -2215,6 +2235,7 @@ class WrappedCompletions extends Completions {
               outputTokens: result.usage?.completion_tokens ?? 0,
               reasoningTokens: result.usage?.completion_tokens_details?.reasoning_tokens ?? 0,
               cacheReadInputTokens: result.usage?.prompt_tokens_details?.cached_tokens ?? 0,
+              cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.prompt_tokens_details),
               webSearchCount: calculateWebSearchCount(result),
               rawUsage: result.usage
             },
@@ -2281,6 +2302,7 @@ class WrappedResponses extends Responses {
         outputTokens: result.usage?.output_tokens ?? 0,
         reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
         cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+        cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
         webSearchCount: calculateWebSearchCount(result),
         rawUsage: result.usage
       },
@@ -2369,7 +2391,8 @@ class WrappedResponses extends Responses {
                     inputTokens: chunk.response.usage.input_tokens ?? 0,
                     outputTokens: chunk.response.usage.output_tokens ?? 0,
                     reasoningTokens: chunk.response.usage.output_tokens_details?.reasoning_tokens ?? 0,
-                    cacheReadInputTokens: chunk.response.usage.input_tokens_details?.cached_tokens ?? 0
+                    cacheReadInputTokens: chunk.response.usage.input_tokens_details?.cached_tokens ?? 0,
+                    cacheCreationInputTokens: extractCacheWriteTokens(chunk.response.usage.input_tokens_details)
                   };
                 }
               }
@@ -2401,6 +2424,7 @@ class WrappedResponses extends Responses {
                   outputTokens: usage.outputTokens,
                   reasoningTokens: usage.reasoningTokens,
                   cacheReadInputTokens: usage.cacheReadInputTokens,
+                  cacheCreationInputTokens: usage.cacheCreationInputTokens,
                   webSearchCount: usage.webSearchCount,
                   rawUsage: rawUsageData
                 },
@@ -2475,6 +2499,7 @@ class WrappedResponses extends Responses {
               outputTokens: result.usage?.output_tokens ?? 0,
               reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
               cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+              cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
               webSearchCount: calculateWebSearchCount(result),
               rawUsage: result.usage
             },
@@ -2592,6 +2617,7 @@ class WrappedResponses extends Responses {
           outputTokens: result.usage?.output_tokens ?? 0,
           reasoningTokens: result.usage?.output_tokens_details?.reasoning_tokens ?? 0,
           cacheReadInputTokens: result.usage?.input_tokens_details?.cached_tokens ?? 0,
+          cacheCreationInputTokens: extractCacheWriteTokens(result.usage?.input_tokens_details),
           rawUsage: result.usage
         },
         stopReason: result.status ?? undefined,