Text-to-Graph

Yes, I believe so. I'm also working on something like this.

Right now, I believe it will require agents, pipelines, fine-tuning, and/or breaking the mold a bit. My plan is to,,,

1. Use one LLM / agent setup decide if that is what the user wants (or would be helpful in the reply)
2. A RAG like system to decide what dataset / type of code agent is needed
3. Another (code specialist) LLM to write the SQL or Python
4. A number of LLMs along the way to determine if the answer / query was correct, or to try again

Search for "sql" and there are a number of pages in the docs

oh cool, they have a PandasQE and a JsonQE (QE: QueryEngine)

generally, my thought is that the LLM system will return JSON to the frontend, which has config+data to determine what/how to render graphs (intermixed with text sections, and citations)

@verdverm I'm also about to have another question dropped in another section. But i'm about to follow up on this one

That would be absolutely sickkkk

This looks like the example to follow / start from: https://docs.llamaindex.ai/en/stable/examples/agent/custom_agent/