Week 1

Financial Advisor Chatbot with AWS Bedrock

This week is about connecting a simple AI chatbot with some custom prompting to an LLM hosted on AWS Bedrock, specifically using the Llama 3 70B Instruct model. The bot acts as a financial advisor that offers actionable and clear financial advice based on user input.

What It Does

• Initializes a connection to AWS Bedrock using the boto3 SDK.
• Uses the converse API to interact with the Llama 3 model, which is fine-tuned for instruction following and conversational interactions.
• Sets up a chat loop where users can ask financial questions or request advice.
• Maintains a conversation history to allow for context-aware responses.
• Applies a system prompt that defines the role of the AI as a helpful, concise, and knowledgeable financial advisor.
• Implements conversation memory management to keep only the most recent exchanges, preventing excessive token usage and ensuring smooth performance.

Features

Conversational Financial Advice

The AI responds to queries about budgeting, saving, investing, retirement planning, debt management, and more — all in plain, understandable language.

Persistent Conversation History

The chat keeps track of the conversation flow by storing messages in a list (conversation_history) and passing them back to the model with each new message.

Memory Management

Limits the number of stored messages (default: last 6 exchanges = 12 messages) to avoid hitting model token limits and reduce latency/cost.