As it turns out, that wasn't really the fix.
With each new 'page load', the same things happens. But if I reload the same conversation after visiting like the admin page, and resend the chat, then it works.
Maybe next week I'll have time to debug and create an issue on their github
Consider this closed. I did get the error again, but then I read I should turn embedding engine on and off.
I had the “ Bypass Embedding and Retrieval” On. I turned it off. And that seemed to have fixed the problem.
See https://docs.openwebui.com/troubleshooting/rag/
“
Go to: Admin Settings > Documents > Embedding Model
Save the embedding model again—even if it's already selected. This forces a recheck/download.
Hmm. Re-upgraded to 3.0.5 and it's working. I'll update to 3.0.6 and see if it's still working
Ignore the log error. That looked to be present in 3.0.4 as well.
After updating to 3.0.5 (or 3.0.6) I get the following response on all chats:
'NoneType' object has no attribute 'get'
I looked at the updating log, and the only issue I saw was this:
Dec 22 16:12:03 ==> Setting permissions
Dec 22 16:12:03 invalid command \".
Dec 22 16:12:03 ERROR: invalid input syntax for type json
Dec 22 16:12:03 LINE 1: UPDATE config SET data= E'{"version":0,"ui":{"default_locale...
Dec 22 16:12:03 ^
Dec 22 16:12:03 DETAIL: Token ""Check the following sentence for grammar and clarity: \"[sentence]" is invalid.
Dec 22 16:12:03 CONTEXT: JSON data, line 1: ...ng sentence for grammar and clarity: \"[sentence]
Dec 22 16:12:03 ==> Update OIDC config
Dec 22 16:12:03 ==> Starting OpenWebUI
Dec 22 16:12:04 /app/code/backend/open_webui/env.py:425: SyntaxWarning: invalid escape sequence '\d'
Dec 22 16:12:04 "^(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[^\w\s]).{8,}$",
Dec 22 16:12:04 /app/code/backend/open_webui/env.py:433: SyntaxWarning: invalid escape sequence '\d'
Dec 22 16:12:04 "^(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[^\w\s]).{8,}$"
Dec 22 16:12:07 INFO [alembic.runtime.migration] Context impl PostgresqlImpl.
Dec 22 16:12:07 INFO [alembic.runtime.migration] Will assume transactional DDL.
Dec 22 16:12:07 INFO [alembic.runtime.migration] Running upgrade 3e0e00844bb0 -> 6283dc0e4d8d, Add channel file table
Dec 22 16:12:07 INFO [alembic.runtime.migration] Running upgrade 6283dc0e4d8d -> 81cc2ce44d79, Update channel file and knowledge table
Dec 22 16:12:07 INFO [alembic.runtime.migration] Running upgrade 81cc2ce44d79 -> c440947495f3, Add chat_file table
Dec 22 16:12:07 WARNI [open_webui.env]
Dec 22 16:12:07 2025-12-22T21:12:07Z
Dec 22 16:12:07 WARNING: CORS_ALLOW_ORIGIN IS SET TO '*' - NOT RECOMMENDED FOR PRODUCTION DEPLOYMENTS.
Dec 22 16:12:07 2025-12-22T21:12:07Z
Dec 22 16:12:10 => Healthcheck error: Error: connect ECONNREFUSED ...:8080
Dec 22 16:12:13 WARNI [langchain_community.utils.user_agent] USER_AGENT environment variable not set, consider setting it to identify your requests.
Dec 22 16:12:15 2025-12-22T21:12:15Z
Anyone else have a problem?
Will cloudron have an External “Sidecar” Python task runner? Trying to understand migration of my pyodide report-generator node.
Just FYI. If you have use pyodide to run python scripts via n8n, wait until the 1.102.0 release:
https://github.com/n8n-io/n8n/releases/tag/n8n%401.102.0
Code Node: Only Block os.system vs Blocking import os
I ran into this one, and now have set n8n to be a manual-upgrade 
https://github.com/n8n-io/n8n/issues/16532#issuecomment-3045792709
It's like they used a time-machine and preemptively fixed this in their main 2 days ago. LOL. Sorry I didn't notice it was so recent.
We can hang tight and wait till the next release!
Sorry, it doesn't seem to be picking up the settings, even though they are in custom_settings.yml. I posted a question to stirling github about that. I'm not sure that it merges custom_settings.yml like you'd assume.
https://github.com/Stirling-Tools/Stirling-PDF/discussions/3058
Thanks for the quick attention, but doesn’t look fixed yet.
I updated, but I don’t see those settings written to the settings.yml, and it still says they are missing (looking in default paths)
Is it hard to allow/run an env.sh? Then maybe we can fix the key problem as well by using environmental variables.
Another question I had. How do I get the API key to persist across restarts? When testing it doesn't save. Under Settings -> Account Settings -> Your API Key'
I poked around seeing if there was an environmental variables env.sh script, but there wasn't on. Otherwise I could try export SYSTEM_CUSTOMPATHS_OPERATIONS_WEASYPRINT=/usr/bin/weasyprint
I think these need to be set in the settings.yml. I can't edit it since it get's refreshed on restart.
system:
customPaths:
operations:
weasyprint: "/usr/bin/weasyprint" #Defaults to /opt/venv/bin/weasyprint
unoconvert: "/usr/bin/unoconv" #Defaults to /opt/venv/bin/unoconvert
Here's the log:
Feb 24 13:08:22 18:08:22.650 [main] WARN s.s.SPDF.config.ExternalAppDepConfig - Missing dependency: /opt/venv/bin/weasyprint - Disabling group: Weasyprint (Affected features: Markdown To PDF, Url To PDF, Html To PDF)
Feb 24 13:08:22 18:08:22.656 [main] WARN s.s.SPDF.config.ExternalAppDepConfig - Missing dependency: /opt/venv/bin/unoconvert - Disabling group: Unoconv (Affected features: File To PDF)
Feb 24 13:08:22 18:08:22.857 [main] INFO s.s.S.config.EndpointConfiguration - Total disabled endpoints: 6. Disabled endpoints: book-to-pdf, file-to-pdf, html-to-pdf, markdown-to-pdf, pdf-to-book, url-to-pdf
Hello,
I just went through fixing my install to the latest 9.0.2. In logs it said "Your EspoCRM version (8.4.2) is not supported. Required version: 9.0.1." I finally realized that since my extensions were not compatible with 9.0, the upgrade script never did it's work.
Steps:
Repair -> Enable Recovery ModeTerminal, you'll need to first run the following commands:mkdir -p /run/espocrm/cache/application
chmod -R 777 /run/espocrm/cache
chown -R www-data:www-data /run/espocrm/cache
Terminal, use button to upload new extensions to /app/data, and then upgrade with php command.php extension --file=/app/data/YOUR-EXTENSION.zipRecovery Mode. Upgrade script will now completeEnjoy!