Has lightRAG integrated RAG-anything yet?

[lai-serena]

When starting the LightRAG server using the docker-compose.yml from the lightRAG project and modifying environment variables in env.example, I encountered an issue where the graphical interface doesn't allow uploading images (e.g., jpg, png) or Excel files (xlsx, xls). However, RAG-anything permits uploading such image formats and office documents. Is this because lightRAG hasn't integrated RAG-anything yet?
Thank you for your attention!

[onestardao]

ah, you're spot on — LightRAG (out of the box) doesn’t support Office or image ingestion via GUI yet.
what you're seeing is a missing file handler pipeline, not an error in env vars or docker setup.

RAG-anything adds those formats, but if the upload button doesn’t allow them, it means LightRAG hasn’t patched the UI + MIME routing logic.
(it's not just about backend parsing — there's a whole semantic filter missing upstream.)

we actually solved this in our own stack, including full Excel / PDF / PNG support — and it even routes to correct agents based on file type.
if you're interested, I can point you to the solution we built — MIT licensed and production-ready.

let me know if you want it.

[lai-serena]

Hi onestardao,
Thanks a lot for your detailed and helpful reply! I'm really interested in the solution you mentioned. Integrating the document handling capabilities (especially Excel/PDF/PNG support and the smart routing by file type) into LightRAG to achieve comprehensive document processing like RAG-anything is exactly what I want.
Could you please share that MIT-licensed, production - ready solution with me? Looking forward to your reply and learning how to patch the UI + MIME routing logic and set up the file handler pipeline in LightRAG.

[onestardao]

sure thing! really glad to hear that — sounds like you’re exactly the kind of person we built this for.

we tackled this by building a full MIME-aware ingestion layer (Excel, PDF, images, etc.) that attaches to any RAG frontend — it not only handles routing but also resolves the semantic filter issue you noticed (where file type affects downstream chunking & query response alignment).

here’s the full MIT-licensed solution, with setup docs, patch guides, and live demos:
🔗 https://github.com/onestardao/WFGY/tree/main/ProblemMap/README.md

specifically, the issues you're hitting are mapped as:

• No.1: semantic boundary drift (RAG-anything’s fallback chunking fails silently on unsupported formats)
• No.4: MIME mismatch at ingestion (GUI lets files in but they don’t flow into retriever logic)
• No.14: first-call crash due to unready schema or routing path

we also provide plug-in filters you can attach to LightRAG or any LangChain-style app to pre-verify, rewrite, and repair doc ingestion paths in real-time.

let me know if you'd like a quick breakdown on how to patch LightRAG’s upload route directly — happy to help.