πŸ“… 2025-05-11 β€” Session: Developed RAG Pipeline and OpenAI API Enhancements

πŸ•’ 14:00–15:10
🏷️ Labels: RAG, Openai, API, Development, Conference
πŸ“‚ Project: Dev
⭐ Priority: MEDIUM

Session Goal

The session aimed to enhance the Retrieval-Augmented Generation (RAG) system and integrate new functionalities using OpenAI’s Assistants API.

Key Activities

  • Explored potential contributions to the LSFA 2025 conference, focusing on semantic pipelines and networking opportunities.
  • Developed a roadmap for the RAG mini-suite, including phases for MVP completion and productization.
  • Implemented a Gradio UI module for RAG, enabling context-based question answering with LLMs.
  • Fixed bugs in the RAG UI, including missing functions and tab configuration issues.
  • Updated the run_llm() function to comply with OpenAI API version 1.0.0.
  • Explored manual vs. native in-model RAG retrieval methods.
  • Created a starter script for an OpenAI-native Assistant-based RAG pipeline.
  • Updated OpenAI Assistants API tool types and methods for message creation with attachments.
  • Addressed file attachment limits in the Assistants v2 API and proposed a hybrid RAG approach using a vector database.

Achievements

  • Completed the implementation of the RAG UI module and fixed related bugs.
  • Successfully updated the run_llm() function and integrated it into the existing system.
  • Developed strategic options for RAG document retrieval using OpenAI Assistants.

Pending Tasks

  • Further exploration and implementation of strategic RAG options for enhanced document retrieval.
  • Continued development of the RAG mini-suite for full productization and outreach.
  • Preparation and submission of proposals for LSFA 2025.