The aiAgent project has been in development for some time, and our team is excited to share the progress made so far. This post will provide a detailed update on the project, focusing on the areas currently under active development. Our goal is to have a stable version 1 release of aiAgent by the end of May, which we will then begin implementing into production systems.

Refining Core Classes for Enhanced Functionality

DocuStore Improvements

We are in the process of refining the DocuStore class to make it more versatile and adaptable. To achieve this, we are moving it to use a base class and implementing Factory mechanisms for class hierarchies. Currently, DocuStore employs PostgreSQL as its database and the psycopg2 Python library for connecting to it.

With the new improvements, client code will utilize the Factory class to obtain an implementation based on the deployment configuration, which may not necessarily be the PostgreSQL DocuStore. This will allow for greater flexibility and easier integration with various data storage systems.

AiAgent Class Enhancements

The AiAgent class is being updated to provide more customizable query results. We are introducing an output level parameter to the query methods, allowing the response to be one of the following:

  1. Simple text from the NLP, similar to ChatGPT
  2. Simple text from the semantic matching network (smnet)
  3. A JSON object containing both NLP and smnet output, along with additional statistics about the responses. These include the data chunks used to generate the query and the matching score from the vector search.

Introducing the aiAgent UI for Seamless Integration

In addition to the core class updates, we have been developing a base UI library and demo application to simplify the integration of aiAgent into your front-end projects. This will provide a user-friendly interface for managing and configuring the aiAgent according to your specific requirements.

Here are a couple of screenshots showcasing the current UI design. The version in development now features a side tab view layout, allowing users to easily access configuration settings, manage their data stores, and make use of other capabilities as they are introduced over time.

Above: The aiAgent UI showing input only view. When user hits enter after their query or if they hit ESC, the UI will extend.

Above: The aiAgent UI showing the fully extended view. The user can close the UI by clicking outside of the agent view or use ESC key.

For more information and up-to-date documentation, visit the aiAgent website. We are eager to bring you the first production-ready release of aiAgent and look forward to your feedback as we continue refining and expanding its capabilities.