AWS Architecture updates

The current implementation of the AI features uses Bedrock and Opensearch. Since May 2025, this has cost over $10k without any return on investment (customer using and paying for this service). This is in part due to the AI features being implemented as a beta or trial feature, and not suitable or ready for production use. This job is created to reduce the AWS bill by replacing the current Bedrock and OpenSearch setup with pgvector (on the database we already pay for) and the Anthropic API. - Clear out the leftover resources from the older builds (before the current app that has been implemented) - there are currently over 30 AWS services being paid for, many of which are not used - here is a spreadsheet of AWS services currently being charged by AWS https://docs.google.com/spreadsheets/d/1NH_STy8nRbW3bvQq1hmz9v2g8xSG_AAHJ302JHRqLWc/edit?gid=1514425589#gid=1514425589 - Provide technical documentation on AWS services used, the new AWS architecture implemented and how it was implemented for company records - Implement automated tests to ensure the updated implementation is production ready (ie clients can use it and we can justify charging extra for this service if used) and won't interfere with current clients using the app on live deals On completion, this will remove the ~$1500 / month Opensearch cost from the AWS bill for the next billing period, and keep all AI features working (Note: AI features include the AI assisted RFI Auto answer and Document Review).