feat: log retriever endpoint #2601
Conversation
dosubot
bot
added
size:L
Pull Request Validation ReportThis comment is automatically generated by Conventional PR Whitelist Report
Result Pull request does not satisfy any enabled whitelist criteria. Pull request will be validated. Validation Report
Result Pull request satisfies all enabled pull request rules. Last Modified at 09 Jul 24 04:02 UTC |
github-actions
bot
added
enhancement
There was a problem hiding this comment.
Thanks for this work.
My opinion is that the common use case is to follow the logs in "real time". The endpoint is not designed for that and I find it very hard to implement on client side and also not efficient. Pagination for real time logs isn't a common pattern.
Polling the server again and again results in adding overhead.
I think we plan to add this feature to the UI so it has to be lightweight as possible.
My suggestion is that we implement the endpoint as unbounded stream of text.
The client could decide to start from the beginning or tail to the end (getting only new logs).
Plus, in this way we get rid of:
- pagination handling (which is expensive in terms of cpu/mem - you're doing a lot of memory allocations)
- sessions management
Additionally, I think we should stream to file to not impact memory usage at all and make all the logs available for the client. (currently is limited to the buffer size)
| return session_id | ||
|
|
||
| def get_page(self, session_id: str, page: int) -> Dict[str, Any]: | ||
| self.cleanup_sessions() |
There was a problem hiding this comment.
we need to setup a timer to run every x time to check for sessions to cleanup. get_page could not be called for a while
There was a problem hiding this comment.
session and pagination is removed
| if session_id not in self.sessions: | ||
| return {"error": "Invalid or expired session ID"} | ||
|
|
||
| session = self.sessions[session_id] |
There was a problem hiding this comment.
we need a lock for accessing the sessions
| total_logs = len(logs) | ||
| total_pages = ceil(total_logs / self.page_size) | ||
|
|
||
| if page < 1 or page > total_pages: |
There was a problem hiding this comment.
does page start from 1 or 0 ?
if the page is not valid, we should return error
There was a problem hiding this comment.
pagination removed
github-actions
bot
added
enhancement
* log retriever endpoint * disabled by default * realtime log stream via http/2 SSE * read and write lock * unit test (cherry picked from commit 81849d5)
Implemented a feature to allow all LangFlow logs to be retrieved through an API
/logs/logsendpoint with pagination support/logsendpoint to support the last number linesLimitation, this endpoint does not filter out logs based on flow, nor user, therefore it's disabled by default. Once it's enabled the logs can be retrieved by any user.