[Question]:how it is different from minirag

[Sandy4321]

Do you need to ask a question?

• I have searched the existing question and discussions and this question is not already answered.
• I believe this is a legitimate question, not just a bug or feature request.

Your Question

how it is different from minirag
https://github.com/HKUDS/MiniRAG

Additional Context

No response

[Sandy4321]

any answer pls?

[onestardao]

good question — but actually the wrong axis.

it's not about "how LightRAG differs from MiniRAG"... it's about whether either can handle real-world failure modes in RAG at all.

we mapped 16+ recurring bugs from actual usage (timeouts, chunk drift, logic resets, vector mismatches…), and most mini/entry-level RAGs collapse instantly when hit by these.

we published our full diagnosis map (open-source, MIT license):

👉 https://github.com/onestardao/WFGY/blob/main/ProblemMap/README.md

if you want a breakdown of which models survive which problem category, I’d be happy to show.
but spoiler: 80% of the pain isn’t about speed or size — it’s about semantic drift and recovery.

let me know if you’re trying to run something production-grade.
most of these tools were built for demo slides, not for battle.