fix(autoplay): reject long tracks, boost Spotify, strip tribute/duration noise

[LucasSantana-Dev]

What

Three root causes behind the issues visible in the screenshot (7-hour video queued, duplicates of now-playing, all tracks from YouTube):

1. Hard-reject tracks > 15 minutes

calculateRecommendationScore previously gave tracks over 7 min a -0.2 penalty — a 7-hour looped YouTube upload still scored high enough to win. Now anything above 15 min returns -Infinity and is dropped at scoring time, covering all candidate paths (seed search, Last.fm, broad fallback).

2. Spotify boost +0.15 → +0.4

The old boost was too weak — when Spotify search returned results they were barely beating YouTube candidates. With +0.4, a Spotify result wins decisively unless the track has a blocking reason (disliked, blocked artist, etc.).

3. Low-quality upload penalty -0.4

Tracks whose resolved title still contains legendado / traduzido / tradução / legendas penalised by -0.4. Eliminates YouTube fan-upload junk even when Spotify fallback fires.

4. New noise patterns (dedup key improvement)

• (Tributo ao Batman) / [Tributo a Led Zeppelin] stripped → core key now matches the clean title
• (Homenagem a X) stripped
• (HH:MM:SS) duration annotations stripped (e.g. (07:05:14) from looped uploads)
• legendas added to bareTitleNoise

These ensure purgeDuplicatesOfCurrentTrack and isDuplicateCandidate correctly match tribute/fan-annotated versions of the now-playing track.

Tests

12 new tests — 2188 total, all green.

Summary by CodeRabbit

• New Features

  • Implemented track duration filtering to exclude songs exceeding 15 minutes from queue recommendations.
  • Enhanced Spotify source prioritization in music selection.
  • Added low-quality upload detection for tracks with dubbing and subtitle markers.
  • Improved title cleaning to remove tribute annotations and duration notations.

• Tests

  • Added comprehensive test coverage for duration filtering and quality detection mechanisms.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
lucky	Ready	Preview, Comment	Apr 13, 2026 10:14pm

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

📝 Walkthrough

Walkthrough

This PR enhances the bot's music queue processing by expanding noise term handling with "legendas", implementing track duration rejection (>15 minutes), penalizing low-quality uploads matching subtitle/translation patterns, increasing Spotify source preference scoring, and adding title cleaning for tribute and duration annotations.

Changes

Cohort / File(s)	Summary
Noise Terms Configuration packages/bot/src/utils/music/noiseTerms.json	Added "legendas" to the bareTitleNoise array to expand noise tokens applied during bare title processing.
Queue Manipulation Logic & Scoring packages/bot/src/utils/music/queueManipulation.ts	Implemented duration rejection (tracks >15 min return score -Infinity), added low-quality upload penalty (0.4 reduction) for titles matching subtitle/legend/tribute patterns, increased Spotify source boost from 0.15 to 0.4, and adjusted formatting in candidate extraction and seed query building.
Queue Manipulation Tests packages/bot/src/utils/music/queueManipulation.spec.ts	Added test cases verifying duration rejection and low-quality upload penalty logic; reformatted Spotify mock structures and adjusted string expectations without changing assertion behavior.
Search Query Cleaning Logic packages/bot/src/utils/music/searchQueryCleaner.ts	Added regex patterns to NOISE_PATTERNS for stripping tribute/homage parentheticals and duration-like annotations (HH:MM:SS format) from titles during cleanTitle processing.
Search Query Cleaning Tests packages/bot/src/utils/music/searchQueryCleaner.spec.ts	Added new test suite validating cleanTitle behavior for tribute tags, homage annotations, and duration-marker stripping while preserving MM:SS-like patterns.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• fix(bot): improve autoplay source diversity #486: Modifies calculateRecommendationScore in the same file, adjusting recommendation penalty logic that now overlaps with the duration rejection and low-quality upload penalties.
• feat(bot): prioritize Spotify source in autoplay search and scoring #585: Changes Spotify preference scoring boost in calculateRecommendationScore, directly related to the adjusted Spotify source weight change from 0.15 to 0.4.
• fix(bot): eliminate same-song repetition in autoplay #575: Expands title-cleaning logic in searchQueryCleaner.ts with hyphenated-suffix handling, touching the same noise pattern and text normalization code paths.

Suggested labels

bot, size/m

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately and concisely summarizes the three main changes: rejecting long tracks, boosting Spotify preference, and stripping tribute/duration noise from titles.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/autoplay-quality-v2

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
86.8% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud