Social Media Performance Analysis

---

Overview

This project analyzes social media performance using:

• Langflow for workflow creation and GPT integration.
• DataStax Astra DB for database operations.

The key tasks include:

1. Database Upload
2. Post Analysis
3. Video Analysis

---

Task Breakdown

1. Database Upload

• Flow Description:
  1. A predefined JSON dataset simulating social media engagement (e.g., likes, shares, comments, post types) is prepared.
  2. The dataset is chunked into 512-character segments with 50-character overlaps.
  3. These chunks are uploaded to Astra DB along with OpenAI embeddings.

2. Post Analysis

• Workflow:

  1. A chat input queries Astra DB using search parameters.
  2. OpenAI embeddings refine the query and parse relevant data.
  3. Insights are generated through Langflow’s GPT integration, providing detailed performance metrics.

• Example Outputs:

  • Carousel posts have 20% higher engagement than static posts.
  • Reels drive 2x more comments compared to other formats.

3. Video Analysis

• Transcript Analysis:

  1. Video scripts or transcripts are extracted using Whisper AI.
  2. The transcript is processed in Langflow, where GPT evaluates strengths, weaknesses, and suggestions.

• Thumbnail Analysis:

  1. A thumbnail image is uploaded and analyzed using the Sharp package.
  2. Key metrics include:
     • Brightness: 75.84%
     • Contrast: 65.26%
     • Color Variety: 25.84%
     • Visual Complexity: 15.53%
     • Topic Relevance: 20%
  3. Suggestions:
     • Good contrast.
     • Consider using more diverse colors and visual elements.
     • Use YouTube’s recommended 16:9 aspect ratio.