Operations & Systems
Youtube Intelligence Agent
This n8n workflow functions as a YouTube intelligence agent, regularly scraping video transcripts via Apify and ingesting them as embeddings into a Qdrant vector database. It then provides a web UI and API for users to search these video transcripts, using an LLM to summarize…
Built · ~6.5 hours saved per week
YouTube Intelligence Agent
Overview
This workflow builds a complete Retrieval-Augmented Generation (RAG) application that functions as an intelligent search engine for a specific YouTube channel. It automates the entire process from data ingestion to providing a user-facing web interface for searching video content.
The workflow operates in two distinct phases:
1. Data Ingestion & Indexing: On a schedule (or manually), it fetches the latest videos from the n8n YouTube channel, downloads their transcripts (subtitles), breaks them into smaller chunks, converts them into vector embeddings using AI, and stores them in a Qdrant vector database. This creates a searchable knowledge base of the channel's video content. 2. Interactive Search & Summarization: It hosts a web UI where a user can type a question. The workflow takes this question, finds the most relevant segments from all video transcripts using vector search, and then uses a Large Language Model (LLM) to both extract the precise snippets with timestamps and generate a concise, direct answer to the user's query.
This automation transforms a video library into a queryable database, allowing users to find specific information within seconds without manually scrubbing through hours of video.
Technologies Used
- n8n: The core automation platform, used for workflow orchestration, backend APIs, and hosting the web UI. - Apify: A web scraping and automation platform used to reliably fetch YouTube channel video lists and their corresponding transcripts. - OpenAI: Provides the AI models for generating text embeddings (vector representations of text) and for language understanding and generation tasks. - Qdrant: A high-performance vector database used to store the video transcript embeddings and perform fast semantic searches. - Redis: Used to implement a simple rate-limiting mechanism for the public-facing search API. - HTML, CSS, htmx: Standard web technologies used to create the dynamic, single-page web application that serves as the user interface.
AI Integration
AI is central to this workflow, powering its ability to understand and search content based on meaning, not just keywords.
1.
Tags: Reports & Analytics, RAG, Perplexity, Data Transformation, QDRANT, AI Automation
Integrations: HTTP, Vector Store, OpenAI, Webhook, Redis, Sub-workflow, Schedule