Growth Portal

A memory-enabled personal growth AI agent that remembers goals, preferences, and conversations to provide context-aware guidance using CockroachDB and AWS.

Comment

Inspiration

Most AI assistants can give useful advice in a single conversation, but personal growth happens over time. I wanted to build an AI coach that could remember a user's goals, preferences, and previous conversations and use that context later instead of starting from scratch.

What it does

Growth Portal is a memory-enabled personal growth AI agent. Users can manage goals and interact with an AI Coach that retrieves relevant information from previous interactions and structured data before responding.

The portal provides an overview of goals and previous discussions, while the AI Coach uses persistent memory to provide more personalized guidance.

How I built it

Growth Portal uses CockroachDB as the persistent memory and application data layer.

Structured information such as goals and conversations is stored in CockroachDB. Long-term memories are stored with vector embeddings, allowing relevant memories to be retrieved through semantic similarity rather than only exact text matching.

CockroachDB Distributed Vector Indexing supports semantic memory retrieval. The CockroachDB Cloud Managed MCP Server is also integrated for agent-accessible retrieval of structured CockroachDB data.

On AWS, the application uses Amazon Bedrock for embeddings and AI responses and AWS Lambda for serverless application execution. The deployed application connects securely from AWS to CockroachDB Cloud.

When a user talks to the AI Coach, the application retrieves relevant persistent memories and structured context, supplies that context to the model, stores the conversation, and extracts durable information for future interactions.

Challenges

One challenge was designing memory so that the agent did not simply store conversation history, but could retrieve useful information later. I implemented semantic retrieval and duplicate-memory checks to make the memory layer more useful.

Another challenge was securely connecting the AWS-hosted application to CockroachDB Cloud. This required configuring SSL certificate verification correctly in the deployed environment.

What I learned

This project helped me understand how persistent memory changes the architecture of an AI application. Instead of treating the database only as storage, CockroachDB becomes part of the agent's reasoning context by combining structured application data with vector-based long-term memory.

Built With

Share this project:

Updates

Submission history