Architecting LLM-Based Chatbots using free open-source LLMs
summary
Argues for using free open‑source LLMs instead of black‑box APIs, outlines a concrete chatbot pipeline: UI → sanitize input → LLM engine → postprocess → delivery, and recommends fine‑tuning on a controlled dataset, adding fallbacks and monitoring. Warns of real failures: researcher‑produced hallucinated GitHub repos that got forked, and common risks like prompt injection, data leakage and training‑data poisoning; cites OWASP top‑10 LLM vulnerabilities and Hugging Face leaderboards.
AI-written summary · about
chapters (8)
- 0:00Introduction and setup
- 3:40Why choose open-source LLMs
- 7:09LLM basics and chatbot architecture
- 10:22Key components of a chatbot
- 11:48Challenges in chatbot development
- 13:13Security risks in LLM chatbots
- 15:01Use cases and best practices
- 15:43Future of LLM chatbots
related talks
42mKeynote: Designing LLM Native systems
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
26mHow Bad Actors AbUse LLMs The Mechanisms That Enable Misuse and What We Can Do About
10mMastering LLM Chatbot Testing: Metrics, Methods and Mistakes to Avoid
01h 49mAI engineering with open access LLMs that lie, curse, and steal
47mCan You Trust Your ChatBot: Techniques for Testing LLM Responses
24mHow to Keep Your LLM Chatbots Real - A Metrics Survival Guide
15mOpenAssistant's Opensource LLMs Saga