Refactor markdown report generation in GenerationStats for improved readability and consistency

Author: pelikhan

packages/core/src/usage.ts

@@ -411,24 +411,24 @@ export class GenerationStats {
 
   /**
    * Generates a compact markdown report suitable for GitHub comments.
-   * 
+   *
    * The report contains:
    * - A collapsible `<details>` section with aggregate usage statistics in the summary
    * - A table showing individual LLM call details including model, tokens, costs, and duration
    * - Proper formatting for tokens (t, kt, Mt) and costs (¢, $)
    * - Duration formatting (ms, s, m, h)
-   * 
-   * @returns A markdown string with a details section containing aggregate results 
+   *
+   * @returns A markdown string with a details section containing aggregate results
    *          as summary and a table with individual LLM call usage, tokens, and costs.
-   * 
+   *
    * @example
    * ```typescript
    * const stats = new GenerationStats("openai:gpt-4", "main");
    * stats.addUsage({ prompt_tokens: 100, completion_tokens: 50, total_tokens: 150, duration: 1000 }, 1000);
-   * 
+   *
    * const child = stats.createChild("openai:gpt-3.5-turbo", "helper");
    * child.addUsage({ prompt_tokens: 200, completion_tokens: 100, total_tokens: 300, duration: 2000 }, 2000);
-   * 
+   *
    * const report = stats.toMarkdownReport();
    * // Returns:
    * // <details>
@@ -443,17 +443,17 @@ export class GenerationStats {
   toMarkdownReport(): string {
     const accumulated = this.accumulatedUsage();
     const totalCost = this.cost();
-    
+
     // Create summary with aggregate statistics
     const summaryParts = [
       `💰 Usage Report`,
-      prettyTokens(accumulated.total_tokens) || "0t",
+      prettyTokens(accumulated.total_tokens, "both"),
       prettyCost(totalCost),
       prettyDuration(accumulated.duration),
-    ].filter(part => !!part);
-    
+    ].filter((part) => !!part);
+
     const summary = summaryParts.join(" ");
-    
+
     // Collect all usage data (parent + children)
     const usageData: Array<{
       Model: string;
@@ -464,47 +464,50 @@ export class GenerationStats {
       $: string;
       "⏱️": string;
     }> = [];
-    
+
     // Add parent stats if it has usage
     if (this.usage.total_tokens > 0) {
       // For parent, calculate cost from actual usage or use existing cost calculation
-      const parentCost = this.chatTurns.length > 0 
-        ? this.chatTurns
-            .filter(turn => !turn.cached)
-            .map(turn => estimateCost(turn.model, turn.usage) ?? estimateCost(this.model, turn.usage))
-            .reduce((a, b) => (a ?? 0) + (b ?? 0), 0)
-        : estimateCost(this.model, this.usage);
-        
+      const parentCost =
+        this.chatTurns.length > 0
+          ? this.chatTurns
+              .filter((turn) => !turn.cached)
+              .map(
+                (turn) =>
+                  estimateCost(turn.model, turn.usage) ?? estimateCost(this.model, turn.usage),
+              )
+              .reduce((a, b) => (a ?? 0) + (b ?? 0), 0)
+          : estimateCost(this.model, this.usage);
+
       usageData.push({
         Model: this.resolvedModel,
         Label: this.label || "-",
         [CHAR_UP_ARROW]: prettyTokens(this.usage.prompt_tokens) || "0t",
-        [CHAR_DOWN_ARROW]: prettyTokens(this.usage.completion_tokens) || "0t", 
+        [CHAR_DOWN_ARROW]: prettyTokens(this.usage.completion_tokens) || "0t",
         [CHAR_UP_DOWN_ARROWS]: prettyTokens(this.usage.total_tokens) || "0t",
         $: prettyCost(parentCost) || "-",
         "⏱️": prettyDuration(this.usage.duration) || "-",
       });
     }
-    
+
     // Add children stats
     for (const child of this.children) {
       const childUsage = child.accumulatedUsage();
       // Calculate cost for child - try existing cost calculation first, then estimate from usage
-      const childCost = child.chatTurns.length > 0 
-        ? child.cost() 
-        : estimateCost(child.model, childUsage);
-      
+      const childCost =
+        child.chatTurns.length > 0 ? child.cost() : estimateCost(child.model, childUsage);
+
       usageData.push({
         Model: child.resolvedModel,
         Label: child.label || "-",
         [CHAR_UP_ARROW]: prettyTokens(childUsage.prompt_tokens) || "0t",
         [CHAR_DOWN_ARROW]: prettyTokens(childUsage.completion_tokens) || "0t",
-        [CHAR_UP_DOWN_ARROWS]: prettyTokens(childUsage.total_tokens) || "0t", 
+        [CHAR_UP_DOWN_ARROWS]: prettyTokens(childUsage.total_tokens) || "0t",
         $: prettyCost(childCost) || "-",
         "⏱️": prettyDuration(childUsage.duration) || "-",
       });
     }
-    
+
     // If no usage data, add a placeholder row
     if (usageData.length === 0) {
       usageData.push({
@@ -517,10 +520,10 @@ export class GenerationStats {
         "⏱️": "-",
       });
     }
-    
+
     // Generate markdown table
     const table = dataToMarkdownTable(usageData);
-    
+
     return details(summary, table);
   }
 }