Это видео недоступно.
Сожалеем об этом.
Advanced RAG Techniques with
HTML-код
- Опубликовано: 14 авг 2024
- Retrieval-Augmented Generation (RAG) is a useful method to enhance LLMs with external knowledge, leading to more relevant answers. But how does one go from a RAG demo to a production RAG application? What are the key factors, frameworks, and techniques to keep in mind?
Join Timescale and special guest presenter Laurie Voss, VP DevRel at @LlamaIndex for a deep dive as we go beyond the basics and explore advanced techniques for implementing RAG when building AI applications.
🛠 𝗥𝗲𝗹𝗲𝘃𝗮𝗻𝘁 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀
📌 Free trial of Timescale Vector ⇒ tsdb.co/webina...
📌 Presentation slides ⇒ tsdb.co/llamai...
📌 Getting started with LlamaIndex and Timescale Vector tutorial ⇒ tsdb.co/llamai...
📌 RAG with time-based retrieval ⇒ • LlamaIndex Webinar: Ti...
🐯 𝗔𝗯𝗼𝘂𝘁 𝗧𝗶𝗺𝗲𝘀𝗰𝗮𝗹𝗲
Timescale a mature cloud PostgreSQL platform engineered for demanding workloads like time-series, vector, events and analytics data.
💻 𝗙𝗶𝗻𝗱 𝗨𝘀 𝗢𝗻𝗹𝗶𝗻𝗲!
🔍 Website ⇒ tsdb.co/homepage
🔍 Slack ⇒ slack.timescal...
🔍 GitHub ⇒ github.com/tim...
🔍 Twitter ⇒ / timescaledb
🔍 Twitch ⇒ / timescaledb
🔍 LinkedIn ⇒ / timescaledb
🔍 Timescale Blog ⇒ tsdb.co/blog
🔍 Timescale Documentation ⇒ tsdb.co/docs
📚 𝗖𝗵𝗮𝗽𝘁𝗲𝗿𝘀
00:00 Introduction
02:07 RAG Challenges: Accuracy, Faithfulness, Recency, Provenance
03:44 How to perform RAG: Vector search, hybrid search
06:05 What is LlamaIndex? (Overview)
07:52 Data Ingestion
09:46 Data embedding (vectorization)
10:26 Vector embedding storage
10:49 Embedding querying
12:46 Advanced RAG Strategies
12:51 Sub Question Query Engine
13:54 Small to big retrieval
15:23 Node preprocessing (metadata filtered search)
16:28 Hybrid search
17:21 Time filtered search (time-series)
17:29 Dealing with Complex documents
19:48 Text to SQL
21:50 Agents
23:40 Production deployment
25:04 Recap and Summary
26:21 Demo: Chat with Github Commits
31:52 Questions and Answers
32:34 Nodes vs Indexes in LlamaIndex
33:45 What LLM should I use for my task? (Small vs large models)
36:14 Gemini Support in LlamaIndex
36:38 RAG and SQL
38:36 Security with RAG and SQL database access
39:54 Knowledge Graphs and RAG
41:14 Agents and custom input
42:22 Node Post Processing in LlamaIndex
44:34 Data Schema for vector tables in PostgreSQL and Timescale
45:59 Document Scoring in RAG
46:59 Conclusion and Resources
Hi, nice video on advancing RAG to a new level.
I'm curious about SQLconnectors, how does this work under the hood? Do you only retrieve the schema of the table or does it share similiar functionality to SQLAgent from langchain?
I think so, you can see the LlamaIndex docs for more details (see video description for links)
Hi Sir, Can you suggest me the best approach that suits to build a RAG app for multiple 10K Reports?