This repo was updated in December 2024, but originally it contains materials that were discussed in "Beginner to Master Ollama & Build a YouTube Summarizer with Llama 3 and LangChain".
This repo will teach you how to:
- Use LLM local or API via Ollama and again via LangChain
- Use Llama models (Llama 3l2-3B as of Dec 2024)
- Build UI with Gradio
- Use case = "Summarize YouTube"
- Summarizatin using LangChain's map-reduce approach.
- [added] Customize prompts during the map and combine steps.
Specifically, we will first retrieve a transcript of the target YouTube video (directed via a URL), then we will as the Llama model to summarize it.
- Create conda environment with python=3.12
- Install these modules
pip install gradio openai langchain langchain-community youtube_transcript_api tiktoken transformers langchain-ollama
- Serve Ollama if it's not run already:
ollama serveIf you don't know how, consult my how-to video or Ollama directly.
- Simply run:
python main.pyOllamato run local LLM APILlama 3.2-3Bfrom Meta, to use as AI brain. See on Ollama page.Gradio, to build UILangChainas framework for LLM apptiktokenlibrary to estimate token counts
Feel free to copy and distribute, but we appreciate you giving us credits.
👍 Like | 🔗 Share | 📢 Subscribe | 💬 Comments | ❓ Questions