Use LLM generate Marp slides
This is a retrieval augmented generation application.
We can have a pre-defined slide outlines.
For each outlines, we will have prompt to guide for generation. (we will have a page that can configure this)
In addition, we can have agents that download related figure or extract figure from the document.
- For each document we will use LlamaIndex for vector storage.
- Prompts for each page, for example "What is the motivation of this document, summarize into 3 bullet points."
- We will use Marp to generate slides and can easily convert to different slides format.
Copy example.env
to .env
, and fill your OpenAI keys. (Or you can input manually in the Streamlit UI)
pip install -r requirements.txt
streamlit run Main.py
Use Black formatter in this project.
For the import require in specific order, use # noqa
at the end of the line.
- Most basic "Upload document -> LlamaIndex -> LLM -> Marp" pipeline
- Customized LLM prompt config page (or config file)
- CLI tool
- Online demo
- Multi-modal
- Able to add image from document
- Able to add image from the web
- Able to capture table from document