@posthog/ai 8.9.1 → 8.9.2

package/dist/openai/index.mjs

@@ -577,7 +577,7 @@ function formatOpenAIResponsesInput(input, instructions) {
   return messages;
 }
 
-var version = "8.9.1";
+var version = "8.9.2";
 
 const DEFAULT_MAX_DEPTH = 3;
 const MAX_STACK_LINES = 20;
@@ -734,15 +734,22 @@ const captureAiGeneration$1 = async (client, options) => {
       };
     }
     httpStatus = httpStatus ?? 200;
-    let costOverrideData = {};
+
+    // A configured price applies only to a count the provider reported, so a call with no
+    // reported usage sends no cost instead of asserting $0. $ai_total_cost_usd sums the sides
+    // that were priced, which makes it the cost of the known side alone when the other side
+    // went unreported: a lower bound on the true total, not an assertion of it.
+    const costOverrideData = {};
     if (options.costOverride) {
-      const inputCostUSD = (options.costOverride.inputCost ?? 0) * (usage.inputTokens ?? 0);
-      const outputCostUSD = (options.costOverride.outputCost ?? 0) * (usage.outputTokens ?? 0);
-      costOverrideData = {
-        $ai_input_cost_usd: inputCostUSD,
-        $ai_output_cost_usd: outputCostUSD,
-        $ai_total_cost_usd: inputCostUSD + outputCostUSD
-      };
+      if (usage.inputTokens !== undefined) {
+        costOverrideData.$ai_input_cost_usd = (options.costOverride.inputCost ?? 0) * usage.inputTokens;
+      }
+      if (usage.outputTokens !== undefined) {
+        costOverrideData.$ai_output_cost_usd = (options.costOverride.outputCost ?? 0) * usage.outputTokens;
+      }
+      if (Object.keys(costOverrideData).length > 0) {
+        costOverrideData.$ai_total_cost_usd = (costOverrideData.$ai_input_cost_usd ?? 0) + (costOverrideData.$ai_output_cost_usd ?? 0);
+      }
     }
 
     // The caller's own token counts override the SDK-derived ones further down, via the
@@ -786,7 +793,9 @@ const captureAiGeneration$1 = async (client, options) => {
       $ai_input: safeInput,
       $ai_output_choices: safeOutput,
       $ai_http_status: httpStatus,
-      $ai_input_tokens: usage.inputTokens ?? 0,
+      ...(usage.inputTokens !== undefined ? {
+        $ai_input_tokens: usage.inputTokens
+      } : {}),
       ...(usage.outputTokens !== undefined ? {
         $ai_output_tokens: usage.outputTokens
       } : {}),
@@ -1339,8 +1348,6 @@ function monitoredStreamTee(source, createStream) {
 class OpenAIChatStreamAccumulator {
   accumulatedContent = '';
   usage = {
-    inputTokens: 0,
-    outputTokens: 0,
     webSearchCount: 0
   };
   toolCalls = new Map();
@@ -1442,8 +1449,6 @@ class OpenAIChatStreamAccumulator {
 class OpenAIResponsesStreamAccumulator {
   output = [];
   usage = {
-    inputTokens: 0,
-    outputTokens: 0,
     webSearchCount: 0
   };
   consume(event, receivedAt = Date.now()) {
@@ -1552,20 +1557,17 @@ function buildChatSuccessOptions(context, result) {
     })
   };
 }
-function buildChatErrorOptions(context, error, metadata = {}) {
+function buildChatErrorOptions(context, error, metadata) {
   return {
     ...context.monitoring,
     model: context.params.model,
     provider: context.provider,
     input: sanitizeOpenAI(context.params.messages, context.client),
     output: [],
-    latency: 0,
+    latency: metadata.latency,
     baseURL: context.baseURL,
     modelParameters: getModelParams(context.modelParametersSource),
-    usage: {
-      inputTokens: 0,
-      outputTokens: 0
-    },
+    usage: metadata.usage ?? {},
     completionId: metadata.completionId,
     providerMetadata: buildProviderMetadata({
       systemFingerprint: metadata.systemFingerprint
@@ -1615,21 +1617,18 @@ function buildBackgroundResponseOptions(context, response) {
     includeRequestId: true
   });
 }
-function buildResponsesErrorOptions(context, error, completionId) {
+function buildResponsesErrorOptions(context, error, metadata) {
   return {
     ...context.monitoring,
     model: context.params.model,
     provider: context.provider,
     input: buildSanitizedResponsesInput(context),
     output: [],
-    latency: 0,
+    latency: metadata.latency,
     baseURL: context.baseURL,
     modelParameters: getModelParams(context.modelParametersSource),
-    usage: {
-      inputTokens: 0,
-      outputTokens: 0
-    },
-    completionId,
+    usage: metadata.usage ?? {},
+    completionId: metadata.completionId,
     error
   };
 }
@@ -1651,7 +1650,7 @@ function buildEmbeddingSuccessOptions(context, usage, latency) {
     }
   };
 }
-function buildEmbeddingErrorOptions(context, error) {
+function buildEmbeddingErrorOptions(context, error, latency) {
   return {
     eventType: AIEvent.Embedding,
     ...context.monitoring,
@@ -1659,12 +1658,10 @@ function buildEmbeddingErrorOptions(context, error) {
     provider: context.provider,
     input: withPrivacyMode(context.client, context.monitoring.privacyMode, context.params.input),
     output: null,
-    latency: 0,
+    latency,
     baseURL: context.baseURL,
     modelParameters: getModelParams(context.modelParametersSource),
-    usage: {
-      inputTokens: 0
-    },
+    usage: {},
     error
   };
 }
@@ -1743,7 +1740,9 @@ let WrappedCompletions$1 = class WrappedCompletions extends AzureOpenAI.Chat.Com
                 modelParametersSource: body
               }, error, {
                 completionId: accumulated.completionId,
-                systemFingerprint: accumulated.systemFingerprint
+                systemFingerprint: accumulated.systemFingerprint,
+                usage: accumulated.usage,
+                latency: (Date.now() - startTime) / 1000
               }));
               throw error;
             }
@@ -1789,7 +1788,9 @@ let WrappedCompletions$1 = class WrappedCompletions extends AzureOpenAI.Chat.Com
           params: openAIParams,
           monitoring: posthogParams,
           modelParametersSource: body
-        }, error));
+        }, error, {
+          latency: (Date.now() - startTime) / 1000
+        }));
         throw error;
       });
       return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -1891,7 +1892,11 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
                 params: openAIParams,
                 monitoring: posthogParams,
                 modelParametersSource: body
-              }, error, accumulated.completionId));
+              }, error, {
+                completionId: accumulated.completionId,
+                usage: accumulated.usage,
+                latency: (Date.now() - startTime) / 1000
+              }));
               throw error;
             }
           })().catch(() => {
@@ -1939,7 +1944,9 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
           params: openAIParams,
           monitoring: posthogParams,
           modelParametersSource: body
-        }, error));
+        }, error, {
+          latency: (Date.now() - startTime) / 1000
+        }));
         throw error;
       });
       return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2031,7 +2038,9 @@ let WrappedResponses$1 = class WrappedResponses extends AzureOpenAI.Responses {
         params: openAIParams,
         monitoring: posthogParams,
         modelParametersSource: body
-      }, error));
+      }, error, {
+        latency: (Date.now() - startTime) / 1000
+      }));
       throw error;
     });
     return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2068,7 +2077,7 @@ let WrappedEmbeddings$1 = class WrappedEmbeddings extends AzureOpenAI.Embeddings
         params: openAIParams,
         monitoring: posthogParams,
         modelParametersSource: body
-      }, error));
+      }, error, (Date.now() - startTime) / 1000));
       throw error;
     });
     return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2156,7 +2165,9 @@ class WrappedCompletions extends Completions {
                 modelParametersSource: body
               }, error, {
                 completionId: accumulated.completionId,
-                systemFingerprint: accumulated.systemFingerprint
+                systemFingerprint: accumulated.systemFingerprint,
+                usage: accumulated.usage,
+                latency: (Date.now() - startTime) / 1000
               }));
               throw error;
             }
@@ -2202,7 +2213,9 @@ class WrappedCompletions extends Completions {
           params: openAIParams,
           monitoring: posthogParams,
           modelParametersSource: body
-        }, error));
+        }, error, {
+          latency: (Date.now() - startTime) / 1000
+        }));
         throw error;
       });
       return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2304,7 +2317,11 @@ class WrappedResponses extends Responses {
                 params: openAIParams,
                 monitoring: posthogParams,
                 modelParametersSource: body
-              }, error, accumulated.completionId));
+              }, error, {
+                completionId: accumulated.completionId,
+                usage: accumulated.usage,
+                latency: (Date.now() - startTime) / 1000
+              }));
               throw error;
             }
           })().catch(() => {
@@ -2352,7 +2369,9 @@ class WrappedResponses extends Responses {
           params: openAIParams,
           monitoring: posthogParams,
           modelParametersSource: body
-        }, error));
+        }, error, {
+          latency: (Date.now() - startTime) / 1000
+        }));
         throw error;
       });
       return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2444,7 +2463,9 @@ class WrappedResponses extends Responses {
         params: openAIParams,
         monitoring: posthogParams,
         modelParametersSource: body
-      }, error));
+      }, error, {
+        latency: (Date.now() - startTime) / 1000
+      }));
       throw error;
     });
     return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2481,7 +2502,7 @@ class WrappedEmbeddings extends Embeddings {
         params: openAIParams,
         monitoring: posthogParams,
         modelParametersSource: body
-      }, error));
+      }, error, (Date.now() - startTime) / 1000));
       throw error;
     });
     return preserveProviderPromise(parentPromise, wrappedPromise);
@@ -2527,13 +2548,10 @@ class WrappedTranscriptions extends Transcriptions {
         if (Symbol.asyncIterator in value) {
           const [stream1, stream2] = monitoredStreamTee(value, (iterator, controller) => new Stream(iterator, controller));
           (async () => {
+            let usage = {};
             try {
               let finalContent = '';
               let firstTokenTime;
-              let usage = {
-                inputTokens: 0,
-                outputTokens: 0
-              };
               const doneEvent = 'transcript.text.done';
               for await (const chunk of stream1) {
                 // Track first token on text delta events
@@ -2575,13 +2593,10 @@ class WrappedTranscriptions extends Transcriptions {
                 provider: 'openai',
                 input: openAIParams.prompt,
                 output: [],
-                latency: 0,
+                latency: (Date.now() - startTime) / 1000,
                 baseURL: this.baseURL,
                 modelParameters: getModelParams(body),
-                usage: {
-                  inputTokens: 0,
-                  outputTokens: 0
-                },
+                usage,
                 error
               });
               throw error;
@@ -2624,13 +2639,10 @@ class WrappedTranscriptions extends Transcriptions {
           provider: 'openai',
           input: openAIParams.prompt,
           output: [],
-          latency: 0,
+          latency: (Date.now() - startTime) / 1000,
           baseURL: this.baseURL,
           modelParameters: getModelParams(body),
-          usage: {
-            inputTokens: 0,
-            outputTokens: 0
-          },
+          usage: {},
           error
         });
         throw error;