Extract YouTube transcripts and metadata for RAG indexing.
npx skills add https://github.com/mindmorass/reflex --skill youtube-harvester
Please help me install this Agent Skill. Skill: youtube-harvester Source: https://github.com/mindmorass/reflex/tree/main/plugins/reflex/skills/youtube-harvester Command: npx skills add https://github.com/mindmorass/reflex --skill youtube-harvester
YouTube hosts a wealth of tutorials and lectures, but transcripts and metadata are often scattered or hard to index. This Skill provides end-to-end extraction, chunking, and enrichment to enable efficient retrieval and reasoning over video content.
Use the YouTube Harvester to ingest a video by URL into your RAG collection: harvest_youtube_video(url='https://youtube.com/watch?v=VIDEO_ID', collection='my_rag', chunk_strategy='time', chunk_size=300)
High-intent search queries and answers about installing and using this skill.
Extract YouTube transcripts using yt-dlp and youtube-transcript-api libraries to pull manual or auto-generated captions from videos. The Skill applies timestamp-aware chunking and ingests transcripts directly into your RAG collection for retrieval and reasoning over video content.
Yes, the Skill supports harvesting playlists and channels with workflows that extract transcripts and metadata from multiple videos and ingest them into a target RAG collection in a single operation.
Time-based, topic-based, and semantic chunking strategies are supported. Time-based preserves timestamp context for precise retrieval, while topic and semantic approaches organize content by meaning for improved relevance in RAG queries.
Yes, metadata enrichment captures title, channel name, upload date, duration, and tags alongside transcripts. This enriched metadata improves search relevance and context preservation when indexing into your RAG collection.
Optional openai-whisper integration enables audio transcription when manual or auto-generated captions are unavailable, ensuring transcript extraction is possible across all video types in your harvesting workflow.
Core dependencies are yt-dlp and youtube-transcript-api for extraction. Optional openai-whisper adds audio transcription capability. Python environment with these libraries installed is the entry requirement before using the Skill.
jMertaDownload videos, audio, subtitles, and transcripts from yt-dlp-supported sites.
vecearConvert documents into chunked JSONL datasets with metadata for embedding search.
ThiNepoCreates Retrieval-Augmented Generation pipelines with Chroma, FAISS, and Pinecone vector stores.
ElliotJLTExtract YouTube video transcripts into timestamped Markdown files with metadata.
