Welcome to Eternity AI's Rtrafactor workspace!
Rtrafactor is a Python library developed at the Indian Institute of Technology, Patna, as part of the Real Time Retrieval Argumentation (RTRA) architecture. It integrates real-time access tools like search, retrieval, summarization, and argumentation into AI models via a Python library.
To read the research paper detailing RTRA, click here.
Install Rtrafactor into your project using pip:
pip install rtrafactor
Ensure the authenticity of the installed package:
pip show rtrafactor
Rtrafactor is available as a L2LM (Language-to-Language Model) architecture, currently integrated with HuggingFace's Inference API. Follow these steps to use Rtrafactor:
-
Import RTRAConnector:
from rtrafactor import RTRAConnector
-
Instantiate RTRAConnector:
connector = RTRAConnector(huggingface_model, huggingface_api_token)
-
Query for Answers:
query = "Your question here?" one_shot_answer = connector.compare_answers(query) print(one_shot_answer)
Here are some example queries you can try with Rtrafactor:
- Why is Delhi's CM in jail?
- Who is Dr. Kuldip Singh Patel?
- Who is Udit Akhouri?
While RTRA enhances model capabilities, there are still limitations to address:
- Hallucination & Citation: Addressing false information and improving citation accuracy.
- Latency: Reducing response time and optimizing underlying architecture.
RTRA architecture challenges the conventional approach of relying on vast computational resources by democratizing access to the internet's information. It enables developers, including solo developers and small teams, to build efficient models without requiring extensive computational resources.
This project was developed by Udit Raj, Sanya Gupta, and other fellow researchers at the Indian Institute of Technology, Patna, as part of ongoing research in AI and natural language processing.
For detailed usage examples and code, visit Rtrafactor on Google Colab.