NeuraBill

• 

  NeuraBill

• 

  Application diagram

• 

  elastic workflow

• 

  elastic AI agent

• 

  slack integration #1

• 

  slack integration #2

• 

  slack integration #3

Inspiration

We replicated a real scenario that I face everyday at work, with a situation really close to what I made up for this hackaton.

What it does

NeuraBill is a solution integrated with Slack, Elastic workflows and Elastic AI agent to analyze our invoice system. Let's say that we have an series of Java microservices used to generate and read customer invoices, and all the application logs are sent to Elastic. With NeuraBill you can just write on Slack a command specifying the number of an invoice that the customer complains he did not receive. NeuraBill will be triggered, searching in the log the main reason why the invoice has not been generated, and scraping to check if the error is related to a buggy application tag, if it is systematic or a temporary failure, and so on. It will respond with all the information needed directly in a dedicated Slack channel.

How we built it

We generated a series of fake Java application logs to put into an Elastic index. We then create a new Elastic workflow for the log analysis. The workflow is integrated with several kind of different steps: it executed some search into the Elastic index using ES|QL, it calls 3 different AI agent to analyze the logs, it sends message to a dedicated Slack channel, and it executes some HTTPS calls to scrape information form GitHub. Finally, we created a new Slack application which trigger the Elastic workflows using a specific command, so in this way NeuraBill is completely integrated with a Slack chat and can be triggered from outside.

Challenges we ran into

The 3 agent prompt definition has been challenging, trying to create the best prompt to define very well the boundaries of the agent duties. The other most difficoult challenge was the orchestration, using different kind of steps into the workflow and making them work good all together.

Accomplishments that we're proud of

We created a smart bot that can be used directly from a Slack channel. This is really great: in a real world scenario, if a customer complains that he didn't receive an invoice, we built a tool ready to use and accessible to all the organization that, with an easy command, give back to you the failure reason within a minute. It does not require any technical skill to use it, and speed up the time required for the analysis, making it a great tool for both technical and not technical workers.

What we learned

It was the first time we were using the Elastic AI Agent builder. We learned how to use it and how to integrate it with the Elastic environment with all the other features.

What's next for NeuraBill

NeuraBill can be improved to search for other specific kind of errors. It can also be improved to work autonomously instead of called manually: for example, we can set it up to be triggered when the percentage of application errors rise a given threeshold. Moreover, we can also set it up more integrations, so for example NeuraBill can be able to open ticket on our ticketing system in case it is needed or be integrated with GitLab in order to fix an error that is found in the logs.

Built With

• elastic
• elastic-workflow
• slack
• yaml