GPT-4 Turbo retrieval documentation maker converts documentation into easily chunkable JSON format

Docs maker will take in documentation and will use the new GPT-4 turbo to convert it into easily chunkable and more accurately retrievable JSON object. We will explore 3 files from basic to more advanced. in the second and third files, we will try to iteratively improve the generated documentation.

Code files are available at Patreon, includes all code files at $30+ levels:

Basic code files are available at $9+ Patreon levels:

Search 200+ echohive videos and code download links:

Everything GPT API Masterclass:

Quick start if you are new to coding and GPT API:

Voice controlled Auto AGI with swarm and multi self launch capabilities:

Chat with us on Discord:

CHAPTERS:
00:00 INTRO
02:00 why this is useful
03:28 Basic code review
06:53 Iterative improvement Code review
08:46 More advanced code review

#gpt4 #gpt4turbo #vectordb #retrievalaugmentedgeneration #openaitutorial #openai #openaiapi #pythonprojects #artificialintelligence #largelanguagemodels

[echohive]

Code files are available at Patreon, includes all code files at $30+ levels:

Basic code files are available at $9+ Patreon levels:



Search 200+ echohive videos and code download links:


Everything GPT API Masterclass:


Quick start if you are new to coding and GPT API:

Voice controlled Auto AGI with swarm and multi self launch capabilities:


Chat with us on Discord:

[vesper]

I wish you had spent a bit of that video talking about how splitting and formatting your knowledge in this way has improved retrieval with your custom GPT, showing before and after and how much better it is with json chunking.. that's the part I was most eager to learn about when I saw your video title. Thanks!

[andydataguy]

Would love to explore an ensemble (swarm) approach as opposed to an iterative approach to finding missing info.

The aggregate accuracy could be estimated and then that error rate published for downstream applications.

Could use GPT3.5 for the analysis and then GPT4 for final review and edit. Then the bulk of token use would be on the input which is cheaper than repeated GPT4 output.

[micbab-vgmu]

Thank you for the great video - In business environment we mostly start the AI journey by connecting data to LLMs - without proper data preparation and RAG we fail. Nothing change data is king.

[ocscmike]

Really interesting approach. Thanks for this. I've been having trouble with inserting Excel and CSV files into Pinecone - there are some gems in here to try. Have you had success with either Excel or CSV files?

[nghitran]

Hi will this works for academic research paper? Assume I have loaded it in text

[nuclear_AI]

As I see it, hths is one of the most powerful uses of LLM's.
Effortlessly making unstructured data, structured 🤯🧠

How frequently did you come across hallucinations using this?