Lightning Talk Beginner

AgentFlow- A memory layer for local llm’s

Rejected

Session Description

AgentFlow is A memory Layer for localised llms which is built on top Ollama that allows us have localised content and print management to keep conversations and to improve context, I believe a memory layer is essential for localised models since the models through ollama usually are single prompted

Key Takeaways

The key takeaways from this talk could be

Why a memory layer is essential for local llms

How a smaller parameter models can simply become smarter than higher parameter models with a memory layer

References

https://www.linkedin.com/posts/subhash-chandra-bose-lavu-2b822b208_localllm-aiengineering-llms-ugcPost-7422292421855621120-V0dH?utm_source=share&utm_medium=member_ios&rcm=ACoAADS3BewBIYW2Z_O_JKo6IVMMinHjT2Jr1v4

https://slyezil.github.io/subhash-portfolio/

Session Categories

Introducing a FOSS project or a new version of a popular project

Speakers

Subhash Chandra Bose lavu Application development professional

I’m a developer who enjoys building real systems and understanding how things actually work under the hood, from backend services and APIs to how data flows through an application in production. Alongside my day-to-day engineering work, I’ve been exploring local LLMs and agent-style architectures from a systems perspective, especially how memory, context, state and resource constraints shape model behaviour beyond simple single-prompt interactions.

https://slyezil.github.io/subhash-p