Scattering Knowledge Every-Where
ASHTRA (Scattering knowledge everywhere using AI) is an application designed to facilitate understanding and dissemination of research papers through podcasts generated by artificial intelligence. With ASHTRA, users can upload research papers and listen to dynamic discussions between virtual hosts discussing the content of the paper. By leveraging advanced technologies such as Advance Rag, Vectara, and Deepgram, ASHTRA transforms complex academic literature into accessible audio content.
- Upload research papers to generate podcasts.
- Dynamic discussions between virtual hosts (Mark and Leslie) analyzing the paper's content.
- Variable podcast duration, ranging from 2 minutes to 10 hours, to accommodate user preferences.
- Utilizes Advance Rag, Vectara, and Deepgram for content retrieval and voice generation.
- Includes a chat bot for answering user queries based on the paper content.
- Streamlined interface for playing, pausing, and navigating podcast content.
- Streamlit
- Python
- React (webpage)
ASHTRA's backend consists of three parallel-running Python scripts:
- LLM.py: Handles communication with Podcast.py and generates scripts for the host and guest.
- Podcast.py: Communicates with LLM.py, uses Deepgram for voice generation, and creates podcast audio files.
- server.py: Creates an endpoint to stream audio files to the frontend application.
Communication between LLM.py and Podcast.py is facilitated through the 'Communication/communication.txt' file, enabling the exchange of script data and coordination of podcast generation.
Upon uploading a research paper, ASHTRA's backend scripts orchestrate the creation of a podcast discussing the paper's content. The resulting podcast can be accessed and played through the frontend application, where users have control over playback options.
To get started with ASHTRA, follow these steps:
- Clone the repository.
- Install the necessary dependencies.
- Run the backend scripts (
LLM.py
,Podcast.py
,server.py
). - Access the frontend application (
home.py
) to interact with ASHTRA and listen to generated podcasts.
- Akhil Songa (
AS
) - Sai Harsha (
SH
) - Tanmai Raavi (
TRA
)
This project is licensed under the MIT License.