Google Service (Drive v3 or Excel V4) added as a parameter #5147
Conversation
|
|
||
| file = service.files().get(fileId=id, supportsAllDrives=True).execute() | ||
| request = service.files().get_media(fileId=id) | ||
| file = self._service().files().get(fileId=id, supportsAllDrives=True).execute() |
There was a problem hiding this comment.
can we do service = self._service() and then use in both these lines?
There was a problem hiding this comment.
I have updated the pull request accordingly. To be noted that I have not been able to test it since I have a poetry installation error on uvtools which claims not to be available on a windows environment. Even though I cannot run the unit tests, I suspect they will fail since googleapiclient must be added to poetry.lock which is something I don't know how to do.
As a workaround to this pull request, I have the rather easy option to create temporary credentials.json and token.json files before calling the GoogleDriveLoader. So really I won't mind if you decide to reject this pull request for simplicity.
I very much enjoy discovering LLM and AI through LangChain and I appreciate it must be quite intense to monitor such a booming community. Thank you for driving this great project!
…removed from private definitions
…oader (doc, sheets, slides or file)
|
II have further improved this Pool request to add support to Slides and allow load_file_from_ids to check for the appropriate loader. |
|
cross-referencing: #5135 |
|
The pull request 5135 resolve this. |
dosubot
bot
added
Ɑ: doc loader
|
#9999 resolves it. |
Modify GoogleDriveLoader so that it can accept a Google Service instead of relying on a path to credential files.
I am deploying LangChain in serverless environment where I use Redis for chat memory and security token store. This pull request allows to pass a Google Drive or a Google Sheet Service instead of a path to local files.
I would also consider moving _load_sheet_from_id into a different class than GoogleDriveLoader since it is not based on Google Drive API.
Dear @eyurtsev, please advise,