More Sustainable AI Agents on Hermes with GreenPT

Today I set up GreenPT as a custom provider on Hermes, my AI agent framework. This article covers my experience with the setup process, the reasons behind choosing GreenPT, and my initial impressions of using their services.

Setup Process

Initial Setup

The setup process for GreenPT on Hermes was straightforward. I started by:

1. Navigating to the custom provider section in Hermes.
2. Entering the base URL and API key provided by GreenPT.
3. Selecting the models and context window size.
4. Choosing the chat completions option (which I later realized was necessary for the type of interactions I wanted).

Signup Issues got resolved very fast

I encountered some issues during the signup process due to an old account from when GreenPT first started. However, their support team was very responsive and fixed the problem quickly. This positive experience with their support team was a good indication of their commitment to customer service.

Why GreenPT?

Sustainability Focus

One of the main reasons I wanted to test GreenPT was their focus on sustainability. As someone based in the Netherlands (EU), I value local support and insights into electricity consumption. GreenPT's dashboard provides detailed statistics on electricity and CO2 emissions, which aligns with my goal of reducing the environmental impact of my AI workloads.

No Model Lock-in

Another important factor was the absence of model lock-in. GreenPT allows me to use a variety of models, giving me the flexibility to choose the best tool for each task without being tied to a specific provider.

Initial Impressions

Dashboard and Metrics

The GreenPT dashboard is user-friendly and provides valuable insights into electricity and CO2-related statistics. Seeing the actual impact of my AI usage was an eye-opener and reinforced my commitment to sustainable practices.

Performance on Hermes

Setting up GreenPT on my VPS was a breeze. The integration with Hermes was seamless, and I was able to start using their services almost immediately. The performance has been impressive so far, with no noticeable latency or downtime.

Conclusion

Overall, my initial experience with GreenPT has been positive. Their focus on sustainability, local support, and flexibility in model choice make them an attractive option for AI workloads. I look forward to continuing to use their services and exploring the full range of features they offer.

Next Steps

I plan to conduct more extensive tests and monitor the performance and sustainability metrics over time. I will also explore the possibility of integrating GreenPT with other projects and sharing my findings with the community.