Support Ticket Handler (STH) agent

• 

  strnd-supprt-aid

Inspiration

  Small, micro, or new firms may lack the necessary manpower or tools to tackle the support tickets that build up via phone calls, emails, or third party support handling tool. The success of a firm big or small can not be measured by the amount of sales, services , or revenue, but by the number of support tickets it is able to resolve, engage with the customer.

YouTube link - https://www.youtube.com/watch?v=gMSXtP40c2Q

What it does

The Support Ticket Handler (STH) agent uses a RAG system where the existing high priority tickets and possible solutions for recurring product or service problems are stored as two separate JSON files in S3 bucket. The agent goes through the information in the JSON file to identify high priority tickets that need to be resolved within the current month and cross reference, analyzes a second JSON file to match the most appropriate solution for the associated customer problem. The response is stored as a JSON file in an S3 bucket also, a response log is stored in the local system via a docker file.

How we built it

The Support Ticket Handler (STH) agent is built with Strands SDK in Type Script. The agent code utilizes a model from a huge list of active foundation models in Amazon bedrock. The strands agent is built with custom tools that execute tasks in a sequential manner and has session management capability to store the agent responses or retrieve content to and from the Amazon S3 buckets. The information in the buckets are secure with right IAM policies restricting the availability of buckets to only selected services and resources associated with the model that is configured in Bedrock. Guardrails in Bedrock model ensure that the agent is contextually grounded. Agent deployment is in docker and a cron job runs the agent everyday, except weekends at a specific time to get information and store responses in a log file and an S3 bucket to assist the support team and engage with the customer.

Challenges we ran into

Challenges were many from engineering the right prompt to get the right details from the S3 bucket to setting up the right custom tool to call and retrieve information stored in the buckets. Model would run successfully locally but required right authorization to run, once deployed in docker which resembled production like deployment. The everyday cron job is also a work in progress.

What we learned

The learning journey spans working with Strands SDK, Bedrock guardrails, setting custom tools, and docker deployment.

What's next for Support Ticket Handler (STH) agent

The agent can be configured to be deployed in AWS Bedrock Agentcore, if needed.

Built With

• amazon-web-services
• bedrock
• json
• lambda
• s3-buckets
• strands-sdk
• typescript