Merge pull request #119 from aqsaaqeel/feat/truncate-big-diffs

feat: handle large diffs by truncating before LLM input

Author: DFanso

cmd/cli/createMsg.go

@@ -91,6 +91,41 @@ func CreateCommitMsg(Store *store.StoreMethods, dryRun bool, autoCommit bool) {
 		return
 	}
 
+	//  Large diff handling
+	const maxDiffChars = 8000 // can change as needed
+	const maxDiffLines = 300
+
+	diffLines := strings.Split(changes, "\n")
+	diffTooLarge := len(changes) > maxDiffChars || len(diffLines) > maxDiffLines
+
+	if diffTooLarge {
+		pterm.Warning.Println("The diff is very large and may exceed the LLM's context window.")
+		pterm.Info.Printf("Diff size: %d lines, %d characters.\n", len(diffLines), len(changes))
+		pterm.Info.Println("Only the first part of the diff will be used for commit message generation.")
+
+		// Truncate the diff for LLM input, preserving whole lines and UTF-8 safety
+		truncatedLines := make([]string, 0, len(diffLines))
+		totalChars := 0
+
+		for i, line := range diffLines {
+			lineLen := len([]rune(line)) + 1 // +1 for newline, using rune count for UTF-8 safety
+
+			// Stop if we've reached max lines or adding this line would exceed max chars
+			if i >= maxDiffLines || (totalChars+lineLen) > maxDiffChars {
+				break
+			}
+
+			truncatedLines = append(truncatedLines, line)
+			totalChars += lineLen
+		}
+
+		changes = strings.Join(truncatedLines, "\n")
+		actualLineCount := len(truncatedLines)
+
+		pterm.Info.Printf("Truncated diff to %d lines, %d characters.\n", actualLineCount, len(changes))
+		pterm.Info.Println("Consider committing smaller changes for more accurate commit messages.")
+	}
+
 	// Handle dry-run mode: display what would be sent to LLM without making API call
 	if dryRun {
 		pterm.Println()