What to do if input text is too long for the LLM used?

[jbdatascience]

I experimented with your knowledge graph generator and it works fine!
But I did not use an OpenAI LLM but a GROQ LLM (which is very fast!).
The GROQ LLMs are mostly compatible with OpenAI's client libraries, so that an easy transformation in the code.

GROQ MODELS
See:
https://console.groq.com/docs/overview
https://console.groq.com/docs/quickstart
https://console.groq.com/docs/models

I used GROQ model: llama-3.3-70b-versatile

That works fine (and incredibly fast!).

But experimenting with longer texts did not work as well.
What to do if the input text is too long for the LLM used?

Perhaps this could be a solution to that problem:

CAN WE SPLIT THE TEXT UP IN CHUNKS THAT DO FIT INTO GROQ LLM'S CONTEXT, AND THEN COMBINE THE SUB GRAPHS INTO THE MAIN KNOWLEDGE GRAPH? BUT HOW TO DO THAT EXACTLY AND WOULD THAT WORK?

Also I found an interesting glitch in the Knowledge Graph based on the text on Albert Eindstein.
In that graph there was this strange relation:

Albert Einstein ---spouse ---> Franklin D. Roosevelt

This is the Knowledge Graph

Albert Einstein KNOWLEDGE GRAPH.PNG

[earayu]

Hello!
I noticed you're looking for a better solution to integrate knowledge graphs with LLMs. My recent project, ApeRAG (https://github.com/apecloud/ApeRAG), is specifically focused on this. It provides a comprehensive RAG framework and performs well in handling complex knowledge structures.

Would you be interested in giving ApeRAG a try to see if it can help you with the challenges you're facing? I'm happy to provide any assistance and support.