We at Open WebUI have been building something new, and are happy to see some early pioneers carve out their digital corners on the Open WebUI Community...
Three weeks ago, we launched community posts on the Open WebUI platform, a shared space where users can publish tools, prompts, and discussions alongside the main application.
This newsletter is our first attempt to surface what's worth your attention. We'll aim to be selective rather than comprehensive, honest about what's working and what isn't, and useful enough that you're glad you opened this.
This Week's Most Useful#
If you install one thing from this newsletter, consider these:
@Fu-Jie's Markdown Normalizer#
A filter that fixes malformed markdown before it reaches the renderer.
If you've seen code blocks render incorrectly, LaTeX formulas break mid-output, or Mermaid diagrams fail to display, this is likely your fix. The tool intercepts model output and corrects common syntax errors—unclosed fences, mismatched delimiters, improper escaping.
It's not flashy work. It's the kind of infrastructure that makes everything else more reliable. We've tested it across several models and it handles edge cases gracefully.
@skyzi000's Sub-Agent#
A solution for context window constraints.
When you're working with a limited context size but need to perform token-heavy operations—like summarizing a long document or running multi-step analysis—this tool delegates those tasks to a secondary agent. You get the outcome without consuming your primary conversation's context.
The implementation is deliberately simple: it does one thing, explains what it's doing, and returns clean results. Useful if you're running local models with smaller context windows, or if you want to keep your main conversation focused.
Installing tools: Visit the post link, click "Get," and it appears in your Open WebUI instance. If you're new to community tools, see the documentation.
Worth Exploring#
These are more specialized, worth a look depending on your workflow.
@Fu-Jie's Prompt Plus#
Fu-Jie appears twice in this newsletter, and it's earned. Prompt Plus adds a spotlight-style interface to Open WebUI: press Cmd+Shift+P to open a searchable prompt library. What makes it interesting is the input system—prompts can include sliders, dropdowns, date pickers, and text fields, turning them into structured forms.
This is useful if you reuse prompts frequently, share them with colleagues, or want to parameterize workflows without explaining where to edit the text each time. It's more of a power-user tool, but if you fit that profile, it's well-made.
@newnol's YouTube Transcript Provider#
Fetches full transcripts from YouTube videos and feeds them into your chat context. The use case is straightforward: you want to analyze, summarize, or ask questions about video content without leaving your conversation.
Works with most videos that have available captions. Doesn't require API keys or external services—it pulls directly from YouTube's transcript data.
@h4nn1b4l's EasyImage#
If you use Open WebUI for image generation, this filter handles the prompt-engineering layer for you. It takes simple descriptions and expands them with style, composition, and technical parameters before passing them to the image model. It also includes a post-generation review step that checks for common issues.
Worth trying if you generate images regularly and want more consistent results without manually crafting detailed prompts each time.
A Conversation Worth Reading#
@spark1 started a discussion about OpenAI's reported plans to claim a share of customers' AI-assisted discoveries.
The thread doesn't pretend to have answers—it asks questions. How would this be enforced? What constitutes a "discovery"? What does this mean for researchers, developers, and companies using these tools in production?
The responses have been measured and thoughtful. It's the kind of discussion we hope becomes common here: informed speculation, not outrage.
What the Numbers Show (Early Data)#
A note on methodology: usage statistics are opt-in, and we haven't widely promoted the sync feature yet. The numbers below represent a self-selected subset of the community, likely more engaged users who've been clicking random buttons on our community platform. Treat these as directional signals, not census data.
With that caveat, here's what we're seeing:
Most-used models this month:
- gpt-5.2 — OpenAI's current flagship
- gpt-4.1-mini — The cost-performance sweet spot for many users
- gemma3:12b — A local model running entirely on user hardware
That third entry is interesting even with limited data. Seeing it appear alongside commercial offerings suggests at least a notable portion of engaged users value local inference.
As more users opt into statistics sharing, we'll have a clearer picture. For now, the cloud/local balance is one of the trends we're watching.
Top regions by message volume: South Korea, United States, Italy.
(Again, this reflects who's opted in, not necessarily where Open WebUI is most used overall.)
New on the Platform#
We've shipped several features to help good work surface and give contributors recognition.
Upvotes and Downvotes#
Posts can now be upvoted and downvoted. Higher-rated content is more discoverable; lower-rated content fades. This is the simplest form of community curation, if something's useful, vote it up.
Model Ratings#
You can now rate the models you use on our community platform. Over time, this should help newcomers find models that work well and give the community a shared sense of what's reliable.
Leaderboards#
A new leaderboard shows the most active and well-received contributors. It's based on post engagement rather than just posting volume, we wanted to reward quality, not just frequency.
If you're new here: The community platform lives at openwebui.com, separate from your self-hosted Open WebUI instance. It's where tools, prompts, and discussions are shared. Your local installation can pull from it, but the platform itself is a standalone community space.
How to Get Featured#
We select content editorially. There's no automated system or submission queue—we browse the community, test things that look promising, and choose what to include.
What catches our attention:
- Real problems. Tools built because you needed them, not because they'd look good in a portfolio.
- Reliability. We test before featuring. If it breaks, we skip it.
- Clear explanation. Posts that explain what the tool does, why it exists, and how to use it.
If you've built something and want us to see it, post in o/openwebui. That's where we look first.
Looking Ahead#
This is the first issue of what we hope becomes a regular practice. We'll publish when there's enough worth saying—probably every week or two, not on a rigid schedule.
Our goal is curation, not comprehensiveness. The community platform has a lot of activity; this newsletter is for people who want someone else to filter.
If you have feedback—on the newsletter, the platform, or anything else—post in o/openwebui or reach out directly.
We're early. We're learning. Thanks for being here.
The Open WebUI Team
