src/core/generator/KeepChangelogGenerator.ts (1)

149-151: Repeated dynamic import() hurts performance

processEntriesWithExistingChangelog imports ChangelogParser on every invocation.
Instantiate the parser once at the top-level (there’s already a past comment about this).

.github/chatmodes/WG Code Sentinel.chatmode.md (1)

68-68: Remove trailing blank line for consistency.

-

.github/chatmodes/WG Code Alchemist.chatmode.md (1)

74-74: Remove trailing blank line for consistency.

-

src/core/generator/ChangelogParser.ts (1)

199-206: Case-sensitive magicr indicator may yield false negatives

content.includes('Generated by Magic Release (magicr)') fails if the comment is lower-cased or reformatted.
Use a case-insensitive regex to be resilient to manual edits.

-return content.includes(magicrIndicator);
+return new RegExp(magicrIndicator, 'i').test(content);

src/core/generator/KeepChangelogGenerator.ts (1)

606-615: Redundant case 'no-conversion'

switch already has a default; keeping an explicit no-conversion branch triggers Biome’s
noUselessSwitchCase warning.

-      case 'no-conversion':
-      default:
+      default:

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

src/core/generator/KeepChangelogGenerator.ts (5)

460-470: Direct file write without backup risks data loss

The writeChangelog method writes directly to the file without creating a backup first. If the write operation fails partway through, the original changelog could be corrupted.

Consider implementing atomic write operations or creating a backup before writing:

 async writeChangelog(content: string, workingDir: string): Promise<void> {
   const filename = this.config.changelog?.filename ?? 'CHANGELOG.md';
   const changelogPath = path.join(workingDir, filename);

   try {
+    // Write to temporary file first
+    const tempPath = `${changelogPath}.tmp`;
+    writeFileSync(tempPath, content, 'utf8');
+    
+    // Atomically rename temp file to actual file
+    const fs = await import('fs/promises');
+    await fs.rename(tempPath, changelogPath);
+    
     logger.info(`Changelog written to: ${changelogPath}`);
   } catch (error) {
     throw new ChangelogError(`Failed to write changelog: ${error}`);
   }
 }

---

27-28: appendOnly flag is never used in the main generation flow

The appendOnly option is declared but the generate() method doesn't check this flag to decide whether to use generateAppendOnly(). This renders the append-only feature inaccessible.

Apply this fix to honor the appendOnly flag:

 async generate(entries: ChangelogEntry[], workingDir: string): Promise<string> {
   logger.debug('Generating Keep a Changelog format', {
     entriesCount: entries.length,
     includeLinks: this.options,
   });

   const existingChangelog = await this.loadExistingChangelog(workingDir);

+  // Use append-only mode if flag is set and existing changelog exists
+  if (this.options.appendOnly && existingChangelog) {
+    return this.generateAppendOnly(entries, existingChangelog, workingDir);
+  }
+
   // Handle the changelog generation with proper merging logic
   const finalEntries = await this.processEntriesWithExistingChangelog(
     entries,
     existingChangelog,
     workingDir
   );

Also applies to: 65-105

---

84-96: Missing blank lines between changelog entries

The formatEntry() calls don't add blank line separators between entries, which could cause them to merge together and reduce readability.

Add blank line separators after each entry:

 // Add unreleased section if needed
 const unreleasedEntry = finalEntries.find(entry => entry.version === 'Unreleased');
 if (unreleasedEntry && this.hasChanges(unreleasedEntry)) {
-  content += this.formatEntry(unreleasedEntry);
+  content += this.formatEntry(unreleasedEntry) + '\n';
 }

 // Add released versions (sorted by version, newest first)
 const releasedEntries = finalEntries
   .filter(entry => entry.version !== 'Unreleased')
   .sort((a, b) => this.compareVersions(b.version, a.version));

 for (const entry of releasedEntries) {
-  content += this.formatEntry(entry);
+  content += this.formatEntry(entry) + '\n';
 }

---

149-149: Dynamic imports cause unnecessary runtime overhead

Using dynamic import() inside frequently called methods introduces performance overhead. Each call re-imports the module.

Import the ChangelogParser once at the top of the file:

+import ChangelogParser from './ChangelogParser.js';
 import type { ChangelogEntry, Change, ChangeType, MagicReleaseConfig } from '../../types/index.js';

Then replace all dynamic imports:

-    const parser = new (await import('./ChangelogParser.js')).default();
+    const parser = new ChangelogParser();

Also applies to: 476-476, 500-500, 633-633, 741-741

---

137-238: Significant overlap between changelog processing methods

Both processEntriesWithExistingChangelog and processEntriesWithScenario handle similar changelog merging logic with different approaches. This duplication increases maintenance complexity.

Consider consolidating the merging logic into a single, flexible method that can handle all scenarios. This would reduce code duplication and simplify maintenance.

Also applies to: 480-611

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

src/core/generator/KeepChangelogGenerator.ts (7)

239-258: Reuse parser.extractVersions to avoid regex duplication

You already have a robust extractor in ChangelogParser. Delegate to it for consistency and fewer edge cases.

-  extractDocumentedVersions(changelog?: string): Set<string> {
-    const versions = new Set<string>();
-    if (!changelog) {
-      return versions;
-    }
-    // Match version headers: ## [1.0.0] - 2023-01-01 or ## [Unreleased]
-    const versionRegex = /^##\s*\[([^\]]+)\]/gm;
-    let match;
-    while ((match = versionRegex.exec(changelog)) !== null) {
-      if (match[1]) {
-        versions.add(match[1]);
-      }
-    }
-    logger.debug(`Found ${versions.size} documented versions`, Array.from(versions));
-    return versions;
-  }
+  async extractDocumentedVersions(changelog?: string): Promise<Set<string>> {
+    if (!changelog) return new Set<string>();
+    const parser = new (await import('./ChangelogParser.js')).default();
+    const versions = parser.extractVersions(changelog);
+    logger.debug(`Found ${versions.size} documented versions`, Array.from(versions));
+    return versions;
+  }

And update the caller:

-    const documentedVersions = this.extractDocumentedVersions(existingChangelog);
+    const documentedVersions = await this.extractDocumentedVersions(existingChangelog);

---

595-610: Redundant 'no-conversion' switch case removed — good; drop it from the union type too

Type still lists 'no-conversion' though default handles it.

-      | 'no-conversion',
+      // 'no-conversion' handled by default

---

811-845: Always-true condition; simplify insertion logic

The condition foundUnreleased || !foundUnreleased is tautological. Insert before the first version encountered (or just below Unreleased).

@@
-        // If we've passed unreleased, this is where we insert
-        if (foundUnreleased || !foundUnreleased) {
-          insertionIndex = i;
-          break;
-        }
+        // Insert before the first version (below Unreleased if present)
+        insertionIndex = i;
+        break;

---

650-700: Normalize trailing newline in appendToNonMagicrChangelog()

Ensure single trailing newline to match generate() behavior.

-    return result.join('\n');
+    return `${result.join('\n').trimEnd()}\n`;

---

332-338: Guard optional commits array

If Change.commits can be undefined, this throws. Use optional chaining.

-    if (this.options.includeCommitLinks && change.commits.length > 0) {
+    if (this.options.includeCommitLinks && change.commits?.length > 0) {

---

861-862: Remove leftover placeholder comment

Dead comment; drop for cleanliness.

-  // ...existing code...

---

434-455: Pre-release ordering nuance

compareVersions ignores pre-release tags; 1.0.0-beta will sort equal to 1.0.0. If you care about semver precedence, incorporate pre-release comparison.

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

README.md (2)

11-14: Consolidate duplicate Demo sections.

You added a Demo GIF at Lines 11-14 but there’s another “😎 Demo” section at Lines 69-72 claiming “Coming soon!”. Keep only one to avoid conflicting messaging.

Apply this diff to remove the second section:

@@
-## 😎 Demo
-
-> 📹 Coming soon! In the meantime, try it out in your project.
-

Also applies to: 69-72

---

30-31: Update the beta version in the Important notice.

README still says v0.1.0-beta, while this PR is v0.1.1-beta. Align the text.

Apply this diff:

-> This project is currently in beta (v0.1.0-beta). While fully functional, expect occasional improvements and refinements. Report any issues you encounter - your feedback helps make it better! 🙏
+> This project is currently in beta (v0.1.1-beta). While fully functional, expect occasional improvements and refinements. Report any issues you encounter - your feedback helps make it better! 🙏

CHANGELOG.md (1)

36-38: Deduplicate repeated tokens in a single entry.

Example shows “([#4], [#4])”. Collapse duplicates during generation to avoid noisy output.

src/core/generator/KeepChangelogGenerator.ts (5)

461-471: Use atomic write with backup to avoid partial writes/corruption.

Write to a temp file, optionally back up existing, then rename.

Apply this diff (and update imports per next comment):

   async writeChangelog(content: string, workingDir: string): Promise<void> {
     const filename = this.config.changelog?.filename ?? 'CHANGELOG.md';
     const changelogPath = path.join(workingDir, filename);
 
     try {
-      writeFileSync(changelogPath, content, 'utf8');
+      const tmpPath = `${changelogPath}.tmp`;
+      const bakPath = `${changelogPath}.${Date.now()}.bak`;
+      if (existsSync(changelogPath)) {
+        copyFileSync(changelogPath, bakPath);
+      }
+      writeFileSync(tmpPath, content, 'utf8');
+      renameSync(tmpPath, changelogPath);
       logger.info(`Changelog written to: ${changelogPath}`);
     } catch (error) {
       throw new ChangelogError(`Failed to write changelog: ${error}`);
     }
   }

---

812-836: Always-true condition; simplify insertion logic.

This block inserts at the first non-Unreleased version; the condition is redundant.

Apply this diff:

-        // If we've passed unreleased, this is where we insert
-        if (foundUnreleased || !foundUnreleased) {
-          insertionIndex = i;
-          break;
-        }
+        insertionIndex = i;
+        break;

---

476-479: Avoid mixed import styles for ChangelogParser.

Use the existing static import instead of dynamic imports here to reduce overhead and keep consistency.

Apply this diff:

-  static async isMagicrGenerated(content: string): Promise<boolean> {
-    const parser = new (await import('./ChangelogParser.js')).default();
+  static async isMagicrGenerated(content: string): Promise<boolean> {
+    const parser = new ChangelogParser();
     return parser.isMagicrGenerated(content);
   }

-    // Create parser instance for this method
-    const parser = new (await import('./ChangelogParser.js')).default();
+    // Reuse static import
+    const parser = new ChangelogParser();

Also applies to: 740-743

---

438-456: SemVer comparison ignores pre-release identifiers.

For correct ordering of pre-releases (e.g., 1.0.0-beta < 1.0.0), consider using a semver library or parsing pre-release/build metadata.

---

240-259: Leverage parser.extractVersions to avoid duplication.

You re-implement version extraction; reuse ChangelogParser.extractVersions for consistency and fewer regex divergences.

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro