big problem!

[zhoup15]

这个项目存在很大问题！！！
1.项目默认使用Mineru，却没有提示安装该是项目。
2.完全按照项目要求配置了，最后生成的海报却是AI自己编的，根本没有用我给的文档。
3.按要求提供了pdf论文，rag却没有任何输出。
这个项目做得真差！

[xlrrrr]

Hi, thank you for the feedback. I re-ran the project immediately after seeing this issue and confirmed that everything works correctly on my end. Let me address your concerns one by one.

1. MinerU Installation

MinerU is already declared in the project's requirements.txt:

mineru[core]==2.6.4

Following the README instructions and running pip install -r requirements.txt will install it along with all other dependencies.

2 & 3. Poster content not from your paper / RAG has no output

These two issues are very likely caused by the same root cause — LightRAG did not work properly during the RAG stage, so the subsequent stages had no real paper content to use, and the LLM fell back to generating from its own knowledge.

To help diagnose the problem, could you please check the following files under your outputs/<project_name>/paper/ directory:

• rag_output/<your_pdf_name>/auto/<your_pdf_name>.md — This is the markdown file generated by MinerU after parsing your PDF. If this file is missing or empty, the PDF parsing step failed.
• rag_storage/ — This directory contains the LightRAG index files (e.g., kv_store_text_chunks.json, vdb_chunks.json). If these files are empty or missing, the RAG indexing step did not complete successfully.
• normal/checkpoint_rag.json — This file contains the RAG query results. Please check whether the "answer" fields contain content actually from your paper, or generic/unrelated text.

I have uploaded a working outputs example here for your reference:
👉 Sample outputs (Google Drive)

You can compare your local output files against this to identify where things went wrong.

Also, please share any error messages or logs from your run (you can add --debug for more detailed output), and I'll be happy to help you troubleshoot further.