Mac Installation

[1of13]

I am having difficulty setting this up on a Mac. I've tried multiple python version, but they all result in failed installations. It would be helpful if you could specify what version to use (as well as hardware requirements).

[yjwang39]

Same on Windows. I've tried Python 3.8.6, 3.10.0 and 3.12

[jianghuyihei]

I am having difficulty setting this up on a Mac. I've tried multiple python version, but they all result in failed installations. It would be helpful if you could specify what version to use (as well as hardware requirements).

Can you provide specific error information? My local Python version is 3.10

[hjzccc]

After ~3 hours of debugging, Somehow I managed to run it.

Issues and Fixes

1. Redundant Dependency: delft

   • Issue: The delft dependency appears to be unnecessary and caused most of the dependency conflicts.
   • Fix: Commented out the delft dependency in the requirements.txt (or the relevant dependency file).

2. Incorrect is_azure Flag in LLM.py

   • Issue: In LLM.py line 56, the is_azure flag is set to True by default. Since I'm using the OpenAI API, this caused configuration issues.
   • Fix: Changed the is_azure flag to False to align with OpenAI API usage.

3. Empty Default Environment Variables

   • Issue: The default environment variables are empty, leading to empty values being used during model initialization.
   • Fix: Commented out the unused environment variables, retaining only OPENAI_API_KEY which is necessary for my setup.

4. Logic Bug in agents.py

   • Issue 1: In agents.py lines 326-376, the variable search_paper is not an array.

     • Fix: Renamed search_paper to search_papers and removed the redundant line at agents.py line 349.

   • Issue 2: The while loop logic at agents.py line 330 seems to attempt finding related papers based on each reference until some are found. However, the condition conflicts with the outer loop. It appears the intended logic is to process one reference at a time. If the result is irrelevant, proceed to the next reference. Finally, if the results are insufficient, fall back to a forward pass. Not sure if I understand it correctly

[LidongBing]

After ~3 hours of debugging, Somehow I managed to run it.

Issues and Fixes

1. Redundant Dependency: delft

   • Issue: The delft dependency appears to be unnecessary and caused most of the dependency conflicts.
   • Fix: Commented out the delft dependency in the requirements.txt (or the relevant dependency file).

2. Incorrect is_azure Flag in LLM.py

   • Issue: In LLM.py line 56, the is_azure flag is set to True by default. Since I'm using the OpenAI API, this caused configuration issues.
   • Fix: Changed the is_azure flag to False to align with OpenAI API usage.

3. Empty Default Environment Variables

   • Issue: The default environment variables are empty, leading to empty values being used during model initialization.
   • Fix: Commented out the unused environment variables, retaining only OPENAI_API_KEY which is necessary for my setup.

4. Logic Bug in agents.py

   • Issue 1: In agents.py lines 326-376, the variable search_paper is not an array.

     • Fix: Renamed search_paper to search_papers and removed the redundant line at agents.py line 349.

   • Issue 2: The while loop logic at agents.py line 330 seems to attempt finding related papers based on each reference until some are found. However, the condition conflicts with the outer loop. It appears the intended logic is to process one reference at a time. If the result is irrelevant, proceed to the next reference. Finally, if the results are insufficient, fall back to a forward pass. Not sure if I understand it correctly

Hello hjzccc, are you interested in joining the project and develop it？

[jianghuyihei]

hjzccc

I'm really sorry for the trouble caused to you. I will fix the code as soon as possible. Thank you for your feedback, and you can also submit your PR to solve the related problem.

[jianghuyihei]

Reference in new

Regarding 4. Issue 2, your understanding is correct. There is a break statement that ends the loop when relevant reference is retrieved. The condition for the outer loop was a remnant from previous code that I did not clean up properly; it was established to address other issues but is no longer necessary. I have now removed it. Thank you for your suggestion, and I apologize for any confusion this may have caused you.

[RunningStone]

Maybe worth a docker version?

[OrangePhi]

May I ask if it is possible to create a more user-friendly version? For example, does it have a suitable UI interface like other AI software? It can be used immediately by beginners as well. Compared to the work you are doing, this should not be difficult.

[jianghuyihei]

OrangePhi

Sure, I will try to build a simple Gradio frontend later

[jianghuyihei]

Maybe worth a docker version?

I am not very familiar with Docker, so I will try to create one in the future