📅 2025-04-18 — Session: Explored Modular AI Workflows with PromptFlow and RAG
🕒 18:40–19:20
🏷️ Labels: Ai Workflows, Promptflow, RAG, Architecture, Development
📂 Project: Dev
⭐ Priority: MEDIUM
Session Goal
The session aimed to explore and understand the modular design of AI workflows, focusing on Microsoft’s PromptFlow and OpenAI’s Retrieval-Augmented Generation (RAG).
Key Activities
- Discussed the modular approach in AI workflows, particularly Microsoft’s PromptFlow and OpenAI’s RAG, highlighting the roles of various tools and blocks.
- Reflected on the distinctions and compatibilities between Microsoft’s PromptFlow and OpenAI’s API approaches, focusing on embedding and retrieval tasks.
- Outlined tools in Microsoft’s PromptFlow for embedding generation, vector storage, and retrieval operations, and their integration with RAG pipelines.
- Explored the process of enhancing PromptFlow workflows by integrating existing tools and adding custom abstractions to improve UX and modularity.
- Reviewed a detailed architecture map of PromptFlow, including folder organization related to embedding, vector search, and reusable tools.
- Outlined the architecture map for the FlowPower/AI Lambda Layer paradigm, detailing core block types and orchestration layers.
- Explored the architecture and components of a Composable Semantic Runtime for modular AI workflows.
Achievements
- Completed the README for the FlowPower system, providing a foundational introduction and options for further development.
Pending Tasks
- Further exploration of the architecture maps and tool package creation for PromptFlow.
- Implementation of the sprint plan for the FlowPower/AI Lambda Layer paradigm.
- Development of additional documentation, templates, or starter code for FlowPower.