Gnomly: Your AI-Powered Web & Video Content Analysis Assistant

Transform Complex Content into Clear Insights

Why You Need This Tool

Do these scenarios sound familiar?

• Facing 20-page research reports but needing only core findings
• Saving 3-hour tutorial videos with no time to watch
• Comparing website perspectives with information overload
• Struggling with technical documentation needing plain-language explanations

Meet Gnomly – the Chrome extension that solves these problems through three core capabilities:

1. Intelligent extraction of web/video content
2. Precise summarization and analysis
3. Real-time Q&A for deeper exploration

Performance tests: Processes 300-page PDFs in 2 minutes, achieves 92% accuracy on YouTube video summarization (Llama2 model)

Core Feature Breakdown

🌐 Universal Content Processing

🚀 Six Technical Breakthroughs

1. Adaptive Chunking System
   Splits long texts while preserving logical flow

2. Cross-Platform Format Retention
   Maintains tables, code blocks, and special characters

graph LR
A[Original Content] --> B{Length Check}
B -->|Exceeds Limit| C[Smart Chunking]
B -->|Within Limit| D[Direct Processing]
C --> E[Sequential AI Processing]
E --> F[Progress Tracking]
F --> G[Result Compilation]

3. URL-Triggered Automation
   Auto-applies presets:

   • reddit.com/* → Focuses on comment analysis
   • youtube.com/watch → Prioritizes transcripts
   • github.com/* → Highlights code segments

4. Real-Time Token Monitoring
   Displays in sidebar:

   • Token consumption
   • Remaining capacity
   • Limit warnings

5. Element Targeting
   Visually select page areas:

   // Example: Extract Reddit comments
   document.querySelector('[data-testid="comment"]')
   

Step-by-Step Installation Guide

Prerequisites

1. Essential software:

   • Latest Chrome browser
   • Ollama local service
   • Recommended models: mistral (speed) or llama2 (accuracy)

2. Terminal commands:

   git clone https://github.com/your/gnomly-repo.git
   cd gnomly-repo
   npm install && npm run build
   

Browser Setup

1. Navigate to chrome://extensions/
2. Enable Developer mode (top-right)
3. Click Load unpacked extension
4. Select the generated /dist folder

Initial Configuration

sequenceDiagram
    User->>Gnomly: Click settings icon ⚙️
    Gnomly->>User: Display configuration panel
    User->>Gnomly: Enter server address
    Gnomly->>Ollama: Connection test
    Ollama-->>Gnomly: Return model list
    Gnomly->>User: Show model dropdown
    User->>Gnomly: Select model → Save

Pro tip: Use http://localhost:11434 for local setups

Practical Usage Scenarios

Case 1: Academic Paper Analysis

1. Open research paper
2. Click toolbar fox icon 🦊
3. Select Get Page Content
4. Click AI Summary
5. Ask follow-ups: “Explain methodology in simple terms”

Case 2: YouTube Learning

1. Open educational video
2. Activate Gnomly sidebar
3. Click Get Transcript
4. System auto-detects video ID
5. Generate timestamped key moments

Case 3: Custom Prompts

| Step                   | Example                          |
|------------------------|----------------------------------|
| 1. Open prompt manager | Creating recipe analyzer         |
| 2. Click "Get URL"     | Auto-fills: `cooking-site.com/*` |
| 3. Write prompt        | "Extract ingredient list"        |
| 4. Element targeting   | Lock ingredient table            |
| 5. Set as default      | Applies to all recipes           |

Advanced: Use *.gov/* for government sites

Technical Architecture

graph LR
    A[Webpage] --> B{Gnomly}
    B --> C[Content Extractor]
    C --> D[Chunk Processor]
    D --> E[Ollama Connector]
    E --> F[AI Processing]
    F --> G[Result Generator]
    G --> H[Chat Interface]

Core Technologies:

• Extraction: Enhanced Readability.js
• Chunking: Dynamic token calculation
• Connectivity: Custom HTTP headers
• Output: Original format preservation

Frequently Asked Questions

❓ Is this free?

Completely open-source! Requires self-hosted Ollama (local/cloud)

❓ Supported AI models?

All Ollama-compatible models:

• mistral: Speed/accuracy balance
• llama2-uncensored: Unfiltered version
• deepseek-coder: Technical specialist

❓ Content length limits?

Depends on model context window (4K-32K tokens). Longer content auto-chunks.

❓ Privacy concerns?

Local processing by default. Remote servers require your own infrastructure.

❓ Deepseek integration?

Free-tier available:

1. Register at https://platform.deepseek.com/
2. Get API key
3. Switch provider in settings

Development Roadmap

UI Enhancements

• Light/dark mode toggle
• Auto-scroll content loading
• Cross-page transcript saving

Feature Priorities

pie
    title User-Requested Features
    "Multi-AI Support" : 45
    "Enhanced Interaction" : 30
    "Auto-Scraping" : 15
    "Model Switching" : 10

Code Improvements

• Expanded test coverage
• Stricter linting rules
• Architectural optimization

Get Started Now

1. [Install Ollama](https://ollama.ai/download)
2. Run: `ollama pull mistral`
3. [Get extension](https://github.com/your/repo)
4. Follow Chapter 3 installation

License: Apache 2.0 (commercial use permitted)