@yhong91/vibetime 0.1.27 → 0.1.28

package/bin/vibetime.mjs

@@ -1928,7 +1928,7 @@ function countTextLines(text) {
 }
 
 // src/lib/constants.ts
-var PACKAGE_VERSION = true ? "0.1.27" : "0.1.1";
+var PACKAGE_VERSION = true ? "0.1.28" : "0.1.1";
 var DEFAULT_API_URL = "http://121.196.224.82:3001";
 var DEFAULT_BACKFILL_BATCH_SIZE = 50;
 var DEFAULT_BACKFILL_BATCH_BYTES = 800 * 1024;
@@ -4894,7 +4894,8 @@ async function parseCodexSessionFile(filePath, options) {
   let lastTurnIdForComplete;
   let turnIdAtLastUserMessage;
   let sessionMetaLocked = false;
-  let lastTokenUsageKey;
+  let lastTotalTokenUsage;
+  let lastPerCallUsageKey;
   const pendingToolCalls = /* @__PURE__ */ new Map();
   for (const [index, line] of lines.entries()) {
     const lineNumber = index + 1;
@@ -5042,32 +5043,79 @@ async function parseCodexSessionFile(filePath, options) {
         }
         const usage = tokenUsageFromPayload(payload);
         if (usage) {
-          const metrics = { ...usage, reasoningEffort };
-          const usageKey = [
-            usage.tokensInput,
-            usage.tokensCachedInput,
-            usage.tokensOutput,
-            usage.tokensReasoningOutput,
-            usage.tokensTotal
-          ].join(":");
-          if (usageKey === lastTokenUsageKey) {
+          const input = usage.tokensInput ?? 0;
+          const cached = usage.tokensCachedInput ?? 0;
+          const output = usage.tokensOutput ?? 0;
+          const reasoning = usage.tokensReasoningOutput ?? 0;
+          const total = usage.tokensTotal ?? 0;
+          const hasCumulativeUsage = Object.keys(objectField(info, "total_token_usage")).length > 0;
+          const usageKey = [input, cached, output, reasoning, total].join(":");
+          if (!hasCumulativeUsage) {
+            if (usageKey !== lastPerCallUsageKey) {
+              events.push(withBackfillRefs(baseCodexEvent({
+                ts,
+                type: "model.usage",
+                sessionId,
+                turnId: currentTurnId,
+                cwd,
+                project,
+                model: rewriteCodexModelForTier(model, serviceTier),
+                confidence: "partial",
+                metrics: usage
+              }), { filePath, sourcePathHash, lineNumber, topType, payloadType, options }));
+              lastPerCallUsageKey = usageKey;
+            }
             break;
           }
-          lastTokenUsageKey = usageKey;
-          events.push(withBackfillRefs(baseCodexEvent({
-            ts,
-            type: "model.usage",
-            sessionId,
-            turnId: currentTurnId,
-            cwd,
-            project,
-            // Only the model.usage event carries the -fast suffix; tool and
-            // turn events keep the bare model so other queries (e.g. "what
-            // model was the user on") don't show tier-specific names.
-            model: rewriteCodexModelForTier(model, serviceTier),
-            confidence: "partial",
-            metrics
-          }), { filePath, sourcePathHash, lineNumber, topType, payloadType, options }));
+          if (lastTotalTokenUsage && total >= lastTotalTokenUsage.total) {
+            const deltaInput = input - lastTotalTokenUsage.input;
+            const deltaCached = cached - lastTotalTokenUsage.cached;
+            const deltaOutput = output - lastTotalTokenUsage.output;
+            const deltaReasoning = reasoning - lastTotalTokenUsage.reasoning;
+            const deltaTotal = total - lastTotalTokenUsage.total;
+            if (deltaInput > 0 || deltaOutput > 0 || deltaTotal > 0) {
+              events.push(withBackfillRefs(baseCodexEvent({
+                ts,
+                type: "model.usage",
+                sessionId,
+                turnId: currentTurnId,
+                cwd,
+                project,
+                model: rewriteCodexModelForTier(model, serviceTier),
+                confidence: "partial",
+                metrics: {
+                  tokensInput: deltaInput > 0 ? deltaInput : void 0,
+                  tokensCachedInput: deltaCached > 0 ? deltaCached : void 0,
+                  tokensOutput: deltaOutput > 0 ? deltaOutput : void 0,
+                  tokensReasoningOutput: deltaReasoning > 0 ? deltaReasoning : void 0,
+                  tokensTotal: deltaTotal > 0 ? deltaTotal : void 0,
+                  modelContextWindow: usage.modelContextWindow,
+                  reasoningEffort
+                }
+              }), { filePath, sourcePathHash, lineNumber, topType, payloadType, options }));
+            }
+          } else {
+            events.push(withBackfillRefs(baseCodexEvent({
+              ts,
+              type: "model.usage",
+              sessionId,
+              turnId: currentTurnId,
+              cwd,
+              project,
+              model: rewriteCodexModelForTier(model, serviceTier),
+              confidence: "partial",
+              metrics: {
+                tokensInput: input,
+                tokensCachedInput: cached,
+                tokensOutput: output,
+                tokensReasoningOutput: reasoning,
+                tokensTotal: total,
+                modelContextWindow: usage.modelContextWindow,
+                reasoningEffort
+              }
+            }), { filePath, sourcePathHash, lineNumber, topType, payloadType, options }));
+          }
+          lastTotalTokenUsage = { input, cached, output, reasoning, total };
         }
         break;
       }
@@ -5284,7 +5332,7 @@ function headlessCodexUsage(raw) {
   const cached = numberField(usage, "cached_input_tokens") ?? numberField(usage, "cache_read_input_tokens") ?? numberField(usage, "cached_tokens") ?? 0;
   const reasoning = numberField(usage, "reasoning_output_tokens") ?? numberField(usage, "reasoning_tokens") ?? 0;
   const totalField = numberField(usage, "total_tokens");
-  const total = totalField !== void 0 && (totalField > 0 || input + output + reasoning === 0) ? totalField : input + output + reasoning;
+  const total = totalField !== void 0 && (totalField > 0 || input + output === 0) ? totalField : input + output;
   if (input === 0 && cached === 0 && output === 0 && reasoning === 0 && total === 0) {
     return;
   }
@@ -5293,7 +5341,7 @@ function headlessCodexUsage(raw) {
     tokensCachedInput: cached || void 0,
     tokensOutput: output || void 0,
     tokensReasoningOutput: reasoning || void 0,
-    tokensTotal: total,
+    tokensTotal: total || void 0,
     modelCalls: 1
   };
 }
@@ -5320,16 +5368,26 @@ function headlessCodexTimestamp(raw) {
 }
 function tokenUsageFromPayload(payload) {
   const info = objectField(payload, "info");
-  const usage = objectField(info, "last_token_usage");
+  const totalUsage = objectField(info, "total_token_usage");
+  const usage = Object.keys(totalUsage).length > 0 ? totalUsage : objectField(info, "last_token_usage");
   if (Object.keys(usage).length === 0) {
     return;
   }
+  const input = numberField(usage, "input_tokens") ?? 0;
+  const cached = numberField(usage, "cached_input_tokens") ?? 0;
+  const output = numberField(usage, "output_tokens") ?? 0;
+  const reasoning = numberField(usage, "reasoning_output_tokens") ?? 0;
+  const totalField = numberField(usage, "total_tokens");
+  const total = totalField !== void 0 && (totalField > 0 || input + output === 0) ? totalField : input + output;
+  if (input === 0 && cached === 0 && output === 0 && reasoning === 0 && total === 0) {
+    return;
+  }
   return {
-    tokensInput: numberField(usage, "input_tokens"),
-    tokensCachedInput: numberField(usage, "cached_input_tokens"),
-    tokensOutput: numberField(usage, "output_tokens"),
-    tokensReasoningOutput: numberField(usage, "reasoning_output_tokens"),
-    tokensTotal: numberField(usage, "total_tokens"),
+    tokensInput: input || void 0,
+    tokensCachedInput: cached || void 0,
+    tokensOutput: output || void 0,
+    tokensReasoningOutput: reasoning || void 0,
+    tokensTotal: total || void 0,
     modelContextWindow: numberField(info, "model_context_window")
   };
 }
@@ -5666,8 +5724,7 @@ async function parseCopilotSessionFile(filePath, options) {
           const cacheReadTokens = numberField(usage, "cacheReadTokens") || 0;
           const cacheWriteTokens = numberField(usage, "cacheWriteTokens") || 0;
           const reasoningTokens = numberField(usage, "reasoningTokens") || 0;
-          const totalInput = inputTokens + cacheReadTokens + cacheWriteTokens;
-          const total = totalInput + outputTokens + reasoningTokens;
+          const total = inputTokens + outputTokens;
           if (total <= 0) {
             continue;
           }
@@ -5682,7 +5739,7 @@ async function parseCopilotSessionFile(filePath, options) {
             model: modelName,
             confidence: "exact",
             metrics: {
-              tokensInput: totalInput || void 0,
+              tokensInput: inputTokens || void 0,
               tokensOutput: outputTokens || void 0,
               tokensCachedInput: cacheReadTokens + cacheWriteTokens || void 0,
               tokensCacheReadInput: cacheReadTokens || void 0,
@@ -6199,20 +6256,15 @@ async function loadToolInputsFromUpdates(updatesPath) {
   return map;
 }
 function metricsFromSignals(signals) {
-  const contextTokens = numberField(signals, "contextTokensUsed");
   const toolCallCount = numberField(signals, "toolCallCount");
   const turnCount = numberField(signals, "turnCount");
   const durationSec = numberField(signals, "sessionDurationSeconds");
   const linesAdded = numberField(signals, "agentLinesAdded");
   const linesRemoved = numberField(signals, "agentLinesRemoved");
-  if (contextTokens === void 0 && toolCallCount === void 0 && turnCount === void 0 && durationSec === void 0) {
+  if (toolCallCount === void 0 && turnCount === void 0 && durationSec === void 0) {
     return void 0;
   }
   const metrics = {};
-  if (contextTokens !== void 0 && contextTokens > 0) {
-    metrics.tokensTotal = contextTokens;
-    metrics.tokensInput = contextTokens;
-  }
   if (toolCallCount !== void 0) {
     metrics.toolCalls = toolCallCount;
   }
@@ -6651,13 +6703,14 @@ function opencodeUsageFromInfo(info) {
   const cacheRead = Math.max(0, numberField(cache, "read") || 0);
   const cacheWrite = Math.max(0, numberField(cache, "write") || 0);
   const totalInput = input + cacheRead + cacheWrite;
-  const total = Math.max(0, numberField(tokensObj, "total") || totalInput + output + reasoning);
+  const totalOutput = output + reasoning;
+  const total = Math.max(0, numberField(tokensObj, "total") || totalInput + totalOutput);
   if (total <= 0) {
     return void 0;
   }
   return {
     tokensInput: totalInput || void 0,
-    tokensOutput: output || void 0,
+    tokensOutput: totalOutput || void 0,
     tokensReasoningOutput: reasoning || void 0,
     tokensCachedInput: cacheRead + cacheWrite || void 0,
     tokensCacheReadInput: cacheRead || void 0,
@@ -8825,28 +8878,29 @@ function workbuddyUsageMetrics(record) {
   if (!input && !output && !total && !cacheRead && !cacheCreate && !reasoning) {
     return void 0;
   }
+  const alignedTotal = total || input + output;
   return {
-    tokensInput: input,
+    tokensInput: input || void 0,
     tokensOutput: output,
     tokensCachedInput: cacheRead + cacheCreate,
     tokensCacheCreationInput: cacheCreate,
     tokensCacheReadInput: cacheRead,
     tokensReasoningOutput: reasoning,
-    tokensTotal: total || input + output,
+    tokensTotal: alignedTotal,
     modelCalls: 1
   };
 }
 function workbuddyCachedTokens(usage, rawUsage) {
-  const rawCache = numberField(rawUsage, "cache_read_input_tokens") ?? numberField(rawUsage, "prompt_cache_hit_tokens") ?? numberField(rawUsage, "cached_tokens");
-  if (rawCache !== void 0) {
-    return rawCache;
-  }
   const details = usage.inputTokensDetails;
-  if (Array.isArray(details) && isPlainObject(details[0])) {
-    return numberField(details[0], "cached_tokens") || 0;
-  }
   const rawDetails = objectField(rawUsage, "prompt_tokens_details");
-  return numberField(rawDetails, "cached_tokens") || 0;
+  return Math.max(
+    0,
+    numberField(rawUsage, "cache_read_input_tokens") || 0,
+    numberField(rawUsage, "prompt_cache_hit_tokens") || 0,
+    numberField(rawUsage, "cached_tokens") || 0,
+    Array.isArray(details) && isPlainObject(details[0]) ? numberField(details[0], "cached_tokens") || 0 : 0,
+    numberField(rawDetails, "cached_tokens") || 0
+  );
 }
 function workbuddyReasoningTokens(usage, rawUsage) {
   const completionDetails = objectField(rawUsage, "completion_tokens_details");
@@ -9249,6 +9303,7 @@ async function readZCodeRows(dbPath) {
       'reasoning_tokens',reasoning_tokens,
       'cache_creation_input_tokens',cache_creation_input_tokens,
       'cache_read_input_tokens',cache_read_input_tokens,
+      'provider_total_tokens',provider_total_tokens,
       'computed_total_tokens',computed_total_tokens,
       'tool_call_count',tool_call_count
     ) from model_usage order by started_at, id;`,
@@ -9280,16 +9335,24 @@ function modelMetrics(row) {
   const reasoning = numberField(row, "reasoning_tokens") || 0;
   const cacheCreation = numberField(row, "cache_creation_input_tokens") || 0;
   const cacheRead = numberField(row, "cache_read_input_tokens") || 0;
-  const total = numberField(row, "computed_total_tokens") || input + output + reasoning;
+  const cached = cacheCreation + cacheRead;
+  const providerTotal = numberField(row, "provider_total_tokens");
+  const computedTotal = numberField(row, "computed_total_tokens");
+  const totalWithoutSeparateCache = input + output;
+  const totalWithSeparateCache = input + cached + output;
+  const comparisonTotal = providerTotal ?? computedTotal ?? 0;
+  const cacheIsSeparate = comparisonTotal > 0 && Math.abs(comparisonTotal - totalWithSeparateCache) < Math.abs(comparisonTotal - totalWithoutSeparateCache);
+  const alignedInput = input > 0 ? input + (cacheIsSeparate ? cached : 0) : cached;
+  const alignedTotal = alignedInput + output;
   const durationMs = numberField(row, "duration_ms");
   return {
-    tokensInput: input,
+    tokensInput: alignedInput || void 0,
     tokensOutput: output,
     tokensReasoningOutput: reasoning,
-    tokensCachedInput: cacheCreation + cacheRead,
+    tokensCachedInput: cached,
     tokensCacheCreationInput: cacheCreation,
     tokensCacheReadInput: cacheRead,
-    tokensTotal: total,
+    tokensTotal: alignedTotal,
     modelDurationMs: durationMs,
     durationMs,
     modelCalls: 1,
@@ -10284,7 +10347,7 @@ function totalTokensFromEvent(event) {
   if (typeof explicit === "number" && explicit > 0) {
     return explicit;
   }
-  return Math.max(0, event.metrics?.tokensInput || 0) + Math.max(0, event.metrics?.tokensOutput || 0) + Math.max(0, event.metrics?.tokensReasoningOutput || 0);
+  return Math.max(0, event.metrics?.tokensInput || 0) + Math.max(0, event.metrics?.tokensOutput || 0);
 }
 function lineStatsFromEvent(event) {
   const files = event.fileActivities || [];

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@yhong91/vibetime",
   "type": "module",
-  "version": "0.1.27",
+  "version": "0.1.28",
   "description": "vibetime CLI — install AI-agent hooks (Claude Code, Codex, OpenCode, Pi) and report activity to vibetime.",
   "license": "MIT",
   "publishConfig": {