"Empowering TYPO3 with Data Sovereign LLMs: An Overview"
summary
Shows running open-source LLMs (Llama/OpenLLaMA family) locally, describing a migration off public APIs to self-hosting, the actual tooling used (Python training scripts and LangChain-style orchestration), and a real run where fine-tuning a 5 MB text corpus drives CPUs to 100% for hours. Demonstrates integration with TYPO3 via t3_cowriter TYPO3 extension, explains how models serve content generation while keeping data on-prem, and outlines practical tradeoffs: performance, resource needs, and data-sovereignty benefits.
AI-written summary · about
chapters (9)
- 0:00Introduction to TYPO3 and LLMs
- 5:22Data sovereignty challenges
- 13:08Hosting solutions for privacy
- 15:07Moving to self-hosted models
- 18:06Introducing Llama models
- 20:26Open-source training tools
- 22:01TYPO3 integration with Llama
- 25:57Content generation demo
- 29:06Q&A and closing remarks
related talks
45mEverything You Need to Know About Running LLMs Locally
37mExpanding the Capabilities of LLMs Through Community Collaboration
01h 49mAI engineering with open access LLMs that lie, curse, and steal- 37mExpanding the Capabilities of LLMs Through Community Collaboration
29mHow to roll your own LLM
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
40mInstructLab: Applying Open Source Methods to Building and Trai...
28mA community based open source LLM and what it can do for your business