Feature: Add comma-separated input support for keywords and channels

[crrrowz]

Overview: This PR introduces support for comma-separated inputs when adding keywords and channels.

User Experience: Users can now paste or type multiple entries at once, separated by commas, making the filtering process much faster and more efficient.

Technical Change: Improved the input parsing logic to handle string splitting and whitespace trimming for a cleaner data set.

[varshneydevansh]

Thank you for this patch.

Frankly I didn't thought of having comma based entry
for filtering. This is a great addition.

I will first test this as currently I am adding category
based filtering using the same Keyword UI in Tab UI.

Moreover, one user also asked for CSV based import
and export and now I can see what they meant.

EDIT: The chanegs are looking good.

[crrrowz]

Thank you for the feedback. I’m glad the changes align with the project's goals. I believe using comma-based entries simplifies the user's workflow while keeping the UI clean and intuitive.

Thinking ahead, this structure also opens the door for some powerful AI integrations. Instead of users manually brainstorming keywords, we could implement a 'Natural Language Intent' feature. For example, a user could type: 'I want to avoid games and distractions that hinder my productivity,' and an LLM could automatically parse that into a comprehensive list of comma-separated filtering keywords.

Even better, we could abstract this entirely: the AI could treat that sentence as a 'Behavioral Persona' or a 'Global Prompt' that filters content in the background, without the user ever needing to see or manage a messy list of keywords.

I’ll be standing by for your thoughts on this vision once you've finished testing the current implementation!

[varshneydevansh]

I actually have plans to intgrate a local
LLM Based solution in the extension
itself. But since the demand of Mobile,
iPad app is substantial along with
Android/Fire TV I have to get those first.

[crrrowz]

To give you a clearer picture of the roadmap I’m proposing, here is a breakdown of the AI Behavioral Filtering Architecture:

1️⃣ The Goal

Moving from a manual "Keyword-based" system to an Intelligent Behavioral Filter. Instead of users managing messy lists like "football, messi, goals", they simply state their intent:

"I want to focus on programming and avoid sports distractions" The system then understands the context and filters accordingly.

2️⃣ The Problems it Solves

• Manual Effort: Eliminates the need for users to brainstorm and type hundreds of keywords manually.
• Context Blindness: Traditional filters might block "Game" even in educational content like "Game Theory." This architecture understands semantics to avoid over-blocking.
• Evolving Content: It catches new terms and trends automatically because it understands the "vibe" or "topic," not just the letters.

3️⃣ The Implementation Roadmap (6 Phases)

• Phases 1-2 (Semantic Understanding): Running Local AI directly in the browser using Transformers.js. This ensures 100% privacy as no data ever leaves the user's device.
• Phases 3-4 (Performance): Leveraging WebGPU to ensure filtering happens in milliseconds (<50ms) without slowing down the UI.
• Phases 5-6 (Hybrid Integration): Merging this AI logic with the existing manual system for maximum reliability.

"Keep the code simple, the UI intuitive, and let the complexity hide within the architectural depth."

I’ll be standing by for your thoughts on this vision once you've finished testing the current implementation!

AI_BEHAVIORAL_FILTERING_ARCHITECTURE.md