v0.2.1

llm-app-interface Version 0.2.1 Release

The llm-app-interface now supports an integration with tactile-guidance (HANS) as seen in this pull request. These changes contain the necessary doc additions to support this integration on the server side, in addition to several other performance/quality changes to the interface.

Full Changelog: v0.2.0...v0.2.1

v0.2.0

llm-app-interface Version 0.2.0 Release

Major Enhancements to Navigation Intelligence, Compatibility, and Memory Context

This release marks a significant evolution from the initial version. In v0.2.0, we've focused on refining the interface’s reasoning quality, expanding its memory capabilities, and ensuring compatibility with the latest LangChain and Pydantic versions. This sets the foundation for even more powerful and user-adaptive behavior in accessibility apps.

Key Improvements:

• Enhanced Reasoning and Prompting: improved the LLM's ability to reason, clarify, and respond through major prompt refactors and expanded navigation logic.
• Memory Summarization Integration: added a summarization node to retain and recall more complex context histories, improving long-term coherence during multi-step interactions.
• New Graph Node for App Call Data: introduced additional test coverage and node logic to better handle external app data, supporting more dynamic scenarios.
• Date & Time Tool Added: simple yet useful utility node now allows the assistant to reason about current temporal context.
• Bug Fixes & Refinements

What's Next?

Many more ideas were discovered during the development of v0.2.0. Refer to the Issues page for more details.

Full Changelog: v0.1.0...v0.2.0

v0.1.0

llm-app-interface Version 0.1.0 Release

This is the initial version of our Flexible Context-Aware Navigation Interface for Accessibitiy Apps. Welcome!

This project is a flexible and intelligent context-aware navigation interface designed to enhance interoperability with accessibility applications for vision-impaired individuals. Utilizing LangChain and Large Language Models (LLMs), it provides a seamless API that supports both human and system inputs, enabling back-and-forth communication and clarifying interactions.

Over the next six months, we anticipate significant development for this project. You can expect major updates to the codebase and the API interface.

Current Features:

• API Interface: Exposed via LangServe for interaction with other apps.
• Context-aware: Uses a LangGraph-based approach for efficient state management and future-proofing.
• Interoperability: Designed to enhance accessibility apps and facilitate communication with LLMs.
• Support for OAuth: User authentication using OAuth2.
• Back-and-Forth Communication: Enables real-time clarifying questions and data exchange between apps and users.

Full Changelog: https://github.com/RillJ/llm-app-interface/commits/v0.1.0