Introducing ExyData Labs: Query to InfluxDB in natural language
Introduction
Part of our DNA at ExyData is to always be at the forefront of technology. We are always looking for new ways to improve introduce tools that helps to improve our services and make them more accessible to our customers. That's why we are excited to introduce ExyData Labs, a new initiative to bring the power of AI / Machine Learning to ExyData services.
Our first project is a natural language interface to query InfluxDB using Flux. This new tool will allow you to ask questions in plain English and get answers from your InfluxDB database. No need to learn a new query language or spend hours writing complex queries. Just ask a question and get an answer.
How it works
For now, we have a prototype of what we called Flux GPT, that you can access from here: ExyData Labs. You can ask questions like "What is the average temperature in the last 24 hours?" or "Show me the top 5 servers by CPU usage". The tool will then translate your question into a Flux query and execute it against our InfluxDB database that we are filling with synthetic data.
Things to consider
This is a prototype and there are still many things to improve. The tool is still learning and may not always understand your questions or doesn't have the best performance. So, be patient :D
How behind the scenes looks like for this project?
Basically, what we did was to train a Llama model with a dataset of questions and their corresponding Flux queries. We then fine-tuned the model with our own dataset and deployed it to a web interface. The model is running on a server and receives the questions from the web interface, translates them into Flux queries, and executes them against our InfluxDB database.
The architecture, looks like this:
Conclusion
We are excited about the potential of this new tool and we hope that it will help you to get more insights and make it accessible to more people. We are planning to continue working on this project and improve it over time. Is not yet decided if we will release it as a product, but we are open to your feedback and suggestions. So, feel free to try it and let us know what you think.