Agora: A Distributed Language Model Framework With API-Call Support for Integrated Climate Forecasting
We introduce Agora, a Generative AI-driven system that delivers expert answers and recommendations on climate and agriculture, transforming complex data into clear, natural language explanations. While built for the rural domain, Agora is highly adaptable and can be deployed across various domain ap...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10993389/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | We introduce Agora, a Generative AI-driven system that delivers expert answers and recommendations on climate and agriculture, transforming complex data into clear, natural language explanations. While built for the rural domain, Agora is highly adaptable and can be deployed across various domain applications. It operates as a “mixture-of-experts” language model system, selectively utilizing multiple fine-tuned large language models for inference. By dynamically integrating external data through API calls, Agora ensures real-time, contextually relevant responses. Agora is built for extensibility—it seamlessly integrates new APIs and domains without requiring a full system retrain. Developed entirely with open-source large language models from the LLaMA family, Agora remains open and adaptable, allowing anyone to extend and enhance its capabilities. Optimized for accessibility, Agora runs efficiently on commodity GPUs without compromising performance. By eliminating the need for expensive hardware like NVIDIA’s A100, it makes text generation more affordable and widely accessible. Agora outperforms closed-source models, achieving 78% accuracy on our question-answering benchmark. This result is achieved via dynamic API integration, which pulls in real-time external data, making responses more adaptive, precise, and context-aware. |
|---|---|
| ISSN: | 2169-3536 |