Skip to main content

Explore and extend

Congratulations on completing the setup and testing your chat agent! It’s time to dive deeper into the code and experiment with its features. This course provides a functional boilerplate to explore memory handling and strategies for enhancing chat interactions.

Explore the codebase​

The project is structured to give you a solid foundation for experimenting with Chromia-based memory storage and retrieval. Here are some starting points:

Memory strategies​

  • Short-term memory: Examine how recent interactions store and retrieve data. Consider scenarios where you might need to adjust the amount of information in short-term memory.
  • Long-term memory: Discover how the system preserves significant memories over sessions. Experiment with the criteria for transferring short-term memories to long-term memory.

Agent behavior​

  • Investigate how the agent generates responses using stored memories. Tweak the way the agent utilizes these memories to enhance context-rich responses.

Try custom scenarios​

  • Modify the prompts or objectives associated with the agent to observe how it affects its behaviour.
  • Introduce additional memory fields or logs to capture specific interactions or metadata.

Suggested experiments​

Here are a few ideas to kickstart your experimentation:

  1. Customize memory limits
    Please adjust the number of short-term memories stored and see how it impacts the interaction flow. For instance, consider changing the logic to retain only the last five interactions instead of ten.

  2. Change memory utilization
    Experiment with how long-term memories summarize and update. Explore what happens if you increase the update frequency or alter the stored content.

  3. Analyze agent logs
    Dive into the LLM_LOG table to evaluate how requests and responses log. Analyze the data to gain insights into optimizing memory retrieval and response generation.

Take it further​

Feel free to experiment beyond the initial setup:

  • Modify the backend: Adjust the database schema or Rell operations to meet specific project needs.
  • Integrate new features: Enhance the agent’s capabilities by integrating new APIs or adding tools.
  • Refactor memory strategies: Implement advanced memory cleanup, prioritization, or tagging strategies to improve performance.

Need an overview?​

If you need a refresher on how the project works, check out the README in the repository for diagrams and a high-level explanation.

This is your playground—experiment, break things, and rebuild! The possibilities are endless when you work with Chromia’s relational blockchain and memory-centric chat agents.