AI YouTube Playlist Summarization and Analysis with Redis

# AI-Powered YouTube Playlist & Video Summarization and Analysis Chat Bot

This N8N workflow transforms YouTube playlists or individual videos into interactive knowledge bases. Instead of watching videos, users can chat with the workflow to get summaries and answers based on the video transcripts.

## How it Works:

1.  **Chat Interaction & Intent Detection:**
    * The workflow starts via a chat interface (`Chat Trigger`), prompting the user for a YouTube playlist or video URL.
    * It retrieves any previous conversation context stored in Redis.
    * A Google Gemini AI agent (`Message Intent`) analyzes the user's input to determine if it contains a valid YouTube playlist URL, a video URL, or neither (`NONE`). It extracts the URL, the corresponding ID (Playlist ID or Video ID), and checks if the user specified a limit for the number of playlist videos to process.
    * The workflow maintains context (like the current URL, ID, and processing status) using Redis (`Update Context Intent`, `Get Previous Context Intent`).

2.  **Routing & Pre-processing Checks:**
    * Based on the detected intent and whether the content needs processing, a `Switch` node (`Route Message Intent`) directs the workflow.
    * If a valid URL is provided and hasn't been processed yet (`status != 'DONE'`), it proceeds to the processing pipeline.
    * It checks if embeddings for the given ID already exist in the Qdrant vector store (`Qdrant Vector Store2`, `If`). If they do, it skips processing and moves to the query handling stage.
    * If the input is a playlist URL but no video limit is specified (`limit=0`), it prompts the user to provide one (`Numb of Videos` agent).

3.  **Video/Playlist Processing Pipeline:**
    * **Data Fetching:**
        * For **Playlists**: Fetches the playlist page (`Playlist HTTP Request`), extracts video details using custom code (`Get Playlist Videos Data`), limits the videos if specified (`Limit`), and then fetches the transcript for each video (`YouTube Transcript`).
        * For **Videos**: Fetches the video page (`Video HTTP Request`), extracts title/description (`Get Title and Desc`), and fetches the transcript (`YouTube Transcript1`).
    * **Transcript Processing & Summarization:**
        * Transcripts are retrieved using the `youtubeTranscripter` node.
        * Transcripts are concatenated (`Concatenate`, `Concatenate1`).
        * Data (like video ID, title, transcript text) is merged and structured (`Merge`, `Edit Fields`).
        * Each video's transcript is summarized by a Google Gemini AI chain (`Summarize & Analyze Transcript`).
    * **Embedding & Storage:**
        * Any existing data for the playlist/video ID is cleared from the Qdrant vector store (`Delete Collection`).
        * The processed transcripts (potentially alongside summaries/metadata) are loaded (`Default Data Loader`).
        * Text is split into manageable chunks (`Recursive Character Text Splitter`).
        * Google Gemini generates vector embeddings for the text chunks (`Embeddings Google Gemini`).
        * These embeddings are stored in a Qdrant collection, indexed by the playlist or video ID (`Qdrant Vector Store`).
    * **Status Update & Final Summary:**
        * The context status in Redis is updated to 'DONE' (`Update Context Process Done1`).
        * All individual video summaries are combined (`Full Summary`).
        * An AI agent (`AI Agent`) may generate a final, comprehensive summary or response based on the processed data, using a tool to query the vector store (`Answer questions with a vector store1`).

4.  **Query Handling (Chatting with Content):**
    * If the user's input is identified as a query rather than a new URL to process (or if processing for a URL is complete), the workflow activates the main query handler.
    * A conversational AI agent (`Handle Queries`), powered by Google Gemini and equipped with chat memory (`Chat Buffer Memory`), answers user questions.
    * This agent uses a specialized tool (`Answer questions with a vector store`) that retrieves relevant information directly from the Qdrant vector store based on the user's query and the stored embeddings.