A Hybrid LLM based Enterprise Chatbot Solution Weaver Peer by Intellya
summary
Hybrid system merges an intent-and-scenario chatbot with retrieval-augmented generation and LLMs to handle questions reliably while covering edge cases. Architecture uses vector stores (Elastic/Chroma/Milvus), an embedder that is often on‑prem fine‑tuned embeddings giving ~20–30% better MAP, and LLMs for summarization; an NL engine routes queries between scenario stories and vector DB + SQL RAG (v2 generates SQL when needed). A conversational buffer preserves context; main limits are chunked-document independence, privacy/infrastructure tradeoffs, and early on‑prem LLM quality.
AI-written summary · about
chapters (13)
- 0:00Hybrid LLM enterprise chatbot
- 1:07Intelia’s AI ecosystem
- 3:56Weaver’s intent chatbot
- 8:33Retrieval augmented generation
- 12:05Vier Pier V1 architecture
- 14:33Embedding model trade-offs
- 19:12Fine-tuning LLMs multilingual
- 20:19Vier Pier V2 SQL integration
- 22:10Hybrid chatbot Weaver RAG
- 24:18Weaver use cases
- 25:23Q&A licensing RAG
- 28:37Knowledge graphs for RAG
- 29:19Q&A data infrastructure
related talks
17mArchitecting LLM-Based Chatbots using free open-source LLMs
36mLLMs and AI agents demystified
26mHow Does LLM and Generative AI Really Work?
48mUnlocking LLM Potential: From Exploration to Integration
01h 04mBuilding Expertise Into Generated Conversations
42mKeynote: Designing LLM Native systems
01h 39mMastering LLMs: From Prompt Engineering to Agentic AI
39mLeveraging AI and Emerging Technologies for Competitive Advantage