Performance: Optimize token deduplication in FrameAlignedMerger with early-exit reverse loop

[ysdede]

What changed

Replaced the Array.prototype.some usage in FrameAlignedMerger (src/parakeet.js) with a reverse for loop that early-exits based on the calculated time difference.

Why it was needed

During long transcriptions, confirmedTokens continually grows. When processing chunks, overlapping regions iterate over this growing array for deduplication. Profiling showed that scanning the entire array (Array.prototype.some) became an $O(N)$ bottleneck, consuming significant main-thread time as lengths scaled up.

Impact

For confirmedTokens arrays of 100,000 items, the benchmark for overlap deduplication dropped from ~484.1ms to ~0.2ms (a ~2000x speedup). This is effectively an $O(1)$ amortized check, reducing GC overhead and CPU usage without altering core functionality.

How to verify

const { FrameAlignedMerger } = await import('./src/parakeet.js');

const merger = new FrameAlignedMerger();
merger.confirmedTokens = Array.from({length: 100000}, (_, i) => ({
  id: i % 100,
  absTime: i * 0.1
}));

const start = performance.now();
merger.processChunk({
  tokenIds: Array.from({length: 1000}, (_, i) => (100000 + i) % 100),
  frameIndices: Array.from({length: 1000}, (_, i) => 100000 + i)
}, 10000, 100);
console.log(`Time: ${performance.now() - start}ms`);

---

PR created automatically by Jules for task 7204369694129704738 started by @ysdede

Summary by Sourcery

Enhancements:

• Improve confirmed token deduplication performance in FrameAlignedMerger by replacing a full-array scan with an early-exit reverse loop based on time ordering.

Summary by CodeRabbit

• Chores
  • Improved performance optimization for internal token confirmation processing.

[google-labs-jules]

👋 Jules, reporting for duty! I'm here to lend a hand with this pull request.

When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down.

I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job!

For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with @jules. You can find this option in the Pull Request section of your global Jules UI settings. You can always switch back!

New to Jules? Learn more at jules.google/docs.

---

For security, I will only act on instructions from the user who triggered this task.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 995cabb2-fadf-4430-a5d3-dc8406cdefe6

📥 Commits

Reviewing files that changed from the base of the PR and between da00dbd and 114c690.

📒 Files selected for processing (1)

• src/parakeet.js

---

📝 Walkthrough

Walkthrough

The FrameAlignedMerger.processChunk method's "already confirmed" check is optimized by replacing a full-array forward scan with bounded reverse iteration over time-ordered confirmedTokens, exiting early when candidate tokens exceed timeTolerance.

Changes

Cohort / File(s)	Summary
Performance Optimization src/parakeet.js	Optimized the "already confirmed" check in FrameAlignedMerger.processChunk by replacing a full-array some() scan with bounded reverse iteration, enabling early exit when tokens fall outside the time tolerance window.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Suggested labels

type/performance, severity/high, effort/S

Poem

🐰 A backward glance through time's sweet stream,
Confirms the tokens in a faster dream—
No needless scans from start to end,
Just wisdom to the past we send! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The PR description covers key sections (what changed, why needed, impact with metrics, and verification steps) but is missing required template elements like fragile areas checklist, scope guard, and test/risk evidence.	Complete the template by adding scope guard checkboxes, fragile areas touched checklist, verification steps (test evidence, memory/regression checks), risk level, and rollback plan.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: optimizing token deduplication performance in FrameAlignedMerger using an early-exit reverse loop approach.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ 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 perf/merger-token-dedup-7204369694129704738

---

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.}

[kilo-code-bot]

Kilo Code Review could not run — your account is out of credits.

Add credits at app.kilo.ai to enable reviews on this change.

[sourcery-ai]

Hey - I've left some high level feedback:

• This optimization relies on confirmedTokens always being sorted by absTime; consider explicitly documenting or asserting this invariant near the class or where confirmedTokens is mutated so future changes don’t accidentally break the early-exit logic.
• Inside the reverse loop, you can avoid the extra Math.abs call by leveraging the breaking condition (e.g., loop while token.absTime - t.absTime <= this.timeTolerance and then only compare IDs), which keeps the logic equivalent but slightly simplifies and tightens the inner check.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- This optimization relies on `confirmedTokens` always being sorted by `absTime`; consider explicitly documenting or asserting this invariant near the class or where `confirmedTokens` is mutated so future changes don’t accidentally break the early-exit logic.
- Inside the reverse loop, you can avoid the extra `Math.abs` call by leveraging the breaking condition (e.g., loop while `token.absTime - t.absTime <= this.timeTolerance` and then only compare IDs), which keeps the logic equivalent but slightly simplifies and tightens the inner check.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly optimizes the token deduplication process within the FrameAlignedMerger by replacing an inefficient Array.prototype.some call with a more performant reverse for loop. The new implementation includes an early-exit condition based on time difference, transforming a previous O(N) bottleneck into an effectively O(1) amortized check. This change drastically improves performance for long transcriptions, leading to substantial speedups and more efficient resource utilization.

Highlights

• Performance Optimization: Replaced the Array.prototype.some usage in FrameAlignedMerger with a reverse for loop that early-exits based on the calculated time difference to optimize token deduplication.
• Addressing Bottleneck: Addressed an O(N) bottleneck in token deduplication during long transcriptions, where scanning the confirmedTokens array consumed significant main-thread time.
• Significant Speedup: Achieved a ~2000x speedup for overlap deduplication in benchmarks with 100,000 items, effectively making the check O(1) amortized and reducing GC overhead and CPU usage.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request optimizes the FrameAlignedMerger class in src/parakeet.js by replacing an O(N) Array.prototype.some() check with a time-ordered backward loop that includes an early exit condition. This change aims to improve performance by avoiding full array scans when checking for already confirmed tokens, especially for long transcriptions. There are no review comments to address.