Turn messy Slack threads into perfect Notion docs.
Stop losing tribal knowledge. React to any Slack thread with a 🧠 emoji, and our AI instantly extracts the problem and solution, pushing clean Markdown to your Knowledge Base.
Noise Filtering AI
Our LLM strips out memes and pleasantries. You only get the actual solution saved to your docs.
Notion Integration
Seamlessly pushes perfectly formatted Markdown directly into your Notion database using native APIs.
Zero Behavior Change
Don't force engineers to write docs. Just ask them to click an emoji when a problem is solved.
No workspace connected
Install ThreadExtract in your Slack workspace to start turning threads into Notion docs.
Allow the "ThreadExtract" app to access Slack
This app was created by a member of your workspace, Korrali.
Review app permissions
Information "ThreadExtract" can view
Actions "ThreadExtract" can take
Settings
Configure your workspace integrations and preferences.
Notion configuration
Connect Notion to automatically sync your extracted Slack threads.
Connect your Notion account and we will automatically set up an Extracted Slack Threads database for you.
Extraction triggers
Select which Slack emojis should trigger extraction.
ThreadExtract OAuth
ThreadExtract OAuth
Production API is returning 503 errors on /api/users endpoint since 10:30 AM
Production API is experiencing a spike in 502 errors due to ConnectionTimeout when trying to reach the Redis cache.
CPU on the Redis cluster is pegged at 99%. The new 'Suggested Users' feature is running an uncached KEYS * scan.
Redis Connection Timeout
[View Original Slack Thread](https://slack.com/app_redirect?team=T0B68DVA7TR&channel=C0B68DVJVP1&message_ts=1785692777.968409)
The Problem
The production API is experiencing a spike in 502 errors due to ConnectionTimeout when trying to reach the Redis cache, with CPU on the Redis cluster pegged at 99%.
The Solution
The issue was caused by the new 'Suggested Users' feature running an uncached KEYS * scan against Redis every time a user logs in. Disabling the enable_suggested_users flag in LaunchDarkly resolved the issue. A Jira ticket will be opened to replace the KEYS * call with a proper SCAN loop or a different architecture for the suggestions.