Distributed environment / azure functions

[promentol]

Thank you for this library

We are trying to run this in azure functions, where filesystem is not persistent.
Is there anyway we can avoid working_dir ? If we attach our own vector db, key value db and graph db, do we need then working directory at all ?

In other words, if indexing and querying are happening in different instances, what can we do ?

[gusye1234]

Yeah, that's a problem.
Right now, nano-graphrag gonna create a working dir no matter the storages are using or not. Will open a PR about this.

[gusye1234]

Hi @promentol ,
check out the latest commit where I add an option to skip the creating of working dir. You can use always_create_working_dir=False to do that(detail in Advances: some setup options)

[alexandergirardet]

Bumping this since I am implementing a similar solution on AWS lambda and Ions. Happy to add code as example once done :)