@vibe-cafe/vibe-usage 0.5.0 → 0.5.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibe-cafe/vibe-usage",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "description": "Track your AI coding tool token usage and sync to vibecafe.ai",
   "type": "module",
   "bin": {

package/src/parsers/codex.js

@@ -115,15 +115,19 @@ export async function parse() {
 
         const model = info.model || payload.model || turnContextModel || sessionModel;
 
+        // OpenAI API: input_tokens INCLUDES cached, output_tokens INCLUDES reasoning.
+        // Normalize to Anthropic-style semantics where each field is non-overlapping.
+        const cachedInput = usage.cached_input_tokens || usage.cache_read_input_tokens || 0;
+        const reasoningOutput = usage.reasoning_output_tokens || 0;
         entries.push({
           source: 'codex',
           model,
           project: sessionProject,
           timestamp,
-          inputTokens: usage.input_tokens || 0,
-          outputTokens: usage.output_tokens || 0,
-          cachedInputTokens: usage.cached_input_tokens || usage.cache_read_input_tokens || 0,
-          reasoningOutputTokens: usage.reasoning_output_tokens || 0,
+          inputTokens: (usage.input_tokens || 0) - cachedInput,
+          outputTokens: (usage.output_tokens || 0) - reasoningOutput,
+          cachedInputTokens: cachedInput,
+          reasoningOutputTokens: reasoningOutput,
         });
       } catch {
         continue;

package/src/parsers/gemini-cli.js

@@ -59,28 +59,32 @@ export async function parse() {
       if (isNaN(ts.getTime())) continue;
 
       if (tokens) {
-        // New format: { input, output, cached, thoughts, tool, total }
+        // Gemini API: input INCLUDES cached, output INCLUDES thoughts. Normalize to non-overlapping.
+        const cached = tokens.cached || 0;
+        const thoughts = tokens.thoughts || 0;
         entries.push({
           source: 'gemini-cli',
           model: msg.model || data.model || 'unknown',
           project: 'unknown',
           timestamp: ts,
-          inputTokens: tokens.input || 0,
-          outputTokens: tokens.output || 0,
-          cachedInputTokens: tokens.cached || 0,
-          reasoningOutputTokens: tokens.thoughts || 0,
+          inputTokens: (tokens.input || 0) - cached,
+          outputTokens: (tokens.output || 0) - thoughts,
+          cachedInputTokens: cached,
+          reasoningOutputTokens: thoughts,
         });
       } else {
-        // Old format: { promptTokenCount, candidatesTokenCount, ... }
+        // Gemini API: promptTokenCount INCLUDES cachedContentTokenCount. Normalize to non-overlapping.
+        const cached = usage.cachedContentTokenCount || 0;
+        const thoughts = usage.thoughtsTokenCount || 0;
         entries.push({
           source: 'gemini-cli',
           model: msg.model || data.model || 'unknown',
           project: 'unknown',
           timestamp: ts,
-          inputTokens: usage.promptTokenCount || usage.input_tokens || 0,
-          outputTokens: usage.candidatesTokenCount || usage.output_tokens || 0,
-          cachedInputTokens: usage.cachedContentTokenCount || 0,
-          reasoningOutputTokens: usage.thoughtsTokenCount || 0,
+          inputTokens: (usage.promptTokenCount || usage.input_tokens || 0) - cached,
+          outputTokens: (usage.candidatesTokenCount || usage.output_tokens || 0) - thoughts,
+          cachedInputTokens: cached,
+          reasoningOutputTokens: thoughts,
         });
       }
     }

package/src/parsers/index.js

@@ -45,7 +45,7 @@ export function aggregateToBuckets(entries) {
     b.outputTokens += e.outputTokens || 0;
     b.cachedInputTokens += e.cachedInputTokens || 0;
     b.reasoningOutputTokens += e.reasoningOutputTokens || 0;
-    b.totalTokens += (e.inputTokens || 0) + (e.outputTokens || 0);
+    b.totalTokens += (e.inputTokens || 0) + (e.outputTokens || 0) + (e.reasoningOutputTokens || 0);
   }
 
   return Array.from(map.values());