The article also discusses the usage costs of aiide, explaining that it provides a simple interface to track the tokens and costs of API calls. It also mentions that aiide has a simple interface to define JSON schemas, which is useful for defining structured outputs and tools. The article concludes by acknowledging the repositories from which aiide has taken inspiration and code, and provides a link to a Gist that demonstrates how to use an LLM as context to write aiide copilots.
Key takeaways:
- aiide is a framework designed to build Language Learning Model (LLM) copilots, providing full control over content sent to the LLM and simplifying streaming for real-time applications.
- The framework includes a simple interface for defining JSON schemas, which is useful for defining structured outputs and tools that might change based on the context of the conversation.
- aiide also offers first-class support for memory, abstracting the chat history into a Pandas DataFrame, and includes a feature for tracking tokens and API costs.
- Tools, defined as classes and passed as instances to the Aiide instance, are a central part of aiide. They can range from simple function calls to complex hierarchical LLM agents.