February was the month Ceivo's AI story grew teeth. Across three releases, we shipped features that make the library visual, explorable, and tunable — and we put real operational controls in the hands of the admins running it. It also happened to be the month we took the stage at the HPA Tech Retreat to show this all off live. Here's what landed.
The Library Graph — explore your content visually
The headline feature of February: the Library Graph. Instead of browsing your library as a flat list of folders and files, you can now open an interactive graph view that shows how your content relates to itself — complete with thumbnail previews, cluster groupings, and navigable links between similar assets.
Alongside it, a new Similar Files feature surfaces content that's related to whatever you're looking at, so finding the other angle, the other take, or the other version of a clip becomes a one-click operation instead of a search expedition.
For big archives especially, this turns "I know it's in there somewhere" into "I can see where it is."
Scene-level tags and scene search
Ceivo's tagging got dramatically more precise this month. You can now add, edit, and manage tags on individual scenes within a video — not just on the file as a whole. Those scene tags flow through into search results, so a query can now take you directly to a single moment inside a long-form asset, not just to the file it lives in.
On top of that, file-level tags picked up their own set of improvements: a featured-tag system that surfaces the tags you use most, so tagging is faster and more consistent across your team.
Choose your AI — and how it thinks
One of the most requested features of the quarter landed in February: AI Provider Selection. When you analyze a file, you can now choose which AI provider to use for that specific analysis — overriding the organization default whenever you need to — so different content types can be run through different models without changing your global settings.
Alongside that, Custom Prompt Templates let you create and save your own prompts for AI analysis. That means your team can standardize how you describe sports, news, scripted content, or any other format — and get consistent, predictable metadata back every time.
And because every scene description now carries history, you can see how a description evolved over time, across models, across prompts. The foundation for full AI provenance is now in place.
MCP session logs and the start of agent observability
As Ceivo's MCP connector started getting real usage, we needed real visibility into it. MCP Session Logs landed in February: you can now view and filter activity logs from your MCP connector sessions — and, in a nice recursive touch, those logs are also accessible through MCP itself.
For anyone running agents against their Ceivo library, this is the difference between trusting the magic and actually seeing what it's doing.
The first Diagnostics Dashboard for admins
February also introduced the Diagnostics Dashboard, a new home for admins to view organization health at a glance — file and scene statistics, files with missing descriptions, and the ability to trigger bulk analysis on a filtered set of content in a single click. It was the foundation for the much richer, multi-tab dashboard that shipped in March.
Email sharing and guest view
Building on the file sharing foundation from January, February added two big upgrades:
- Share Files via Email — enter recipient email addresses directly in the share dialog and Ceivo sends a branded email with a link to the shared files.
- Guest View — the people you share with can view and download files directly in the browser, with no account required.
Together, they make sharing out of Ceivo feel as easy as sharing out of any modern cloud app — without giving up the governance that makes Ceivo, Ceivo.
Organization settings and management, finally in the UI
Two infrastructure features that power users had been waiting for:
- Organization Settings — a new screen where you can view and edit your org's settings directly, instead of routing through support.
- Organization Management — super admins can now manage every organization from a single central screen.
Faster processing for long-form content
Under the hood, scene detection got a major upgrade: it can now run on dedicated long-running infrastructure (ECS Fargate), removing previous time limits and allowing much larger and longer videos to be fully processed. If you were holding back on running analysis across your feature-length content, that constraint is gone.
Top 10 notable improvements
Beyond the headline features, ten smaller-but-noticeable items worth calling out:
- Faster playlist rendering — render jobs now start processing immediately when created.
- Cleaner re-analysis — re-analyzing a file now properly replaces old tags instead of piling up duplicates.
- Caption timing fix — captions with broadcast timecode offsets now display correctly during playback.
- Scene Detection toggle — you can turn scene detection on or off in the analysis modal.
- Bigger clipping indicators in the video player for faster visual feedback.
- Improved MCP search with better relevance.
- Fixed MCP connector URL display on the connectors screen.
- Fixed shared-link expiration — links can no longer be set beyond the maximum allowed duration.
- Fixed OpenAI embedding compatibility when using OpenAI for generating embeddings.
- Expanded automated test coverage across the platform, plus a long list of stability and dependency updates.
Why this month mattered
February was the month Ceivo's AI story stopped being about what the model can do and started being about how you control it, how you observe it, and how you make it fit your team's content. Provider selection, custom prompts, scene-level tags, MCP logs, and the first diagnostics dashboard aren't individually flashy — but together they turn Ceivo into something an enterprise media operation can actually deploy with confidence.
It's also what made our HPA Tech Retreat demos land the way they did. The moment you can show an agent safely querying a real library, with typed tags and full logs, the conversation in the room changes.
What's next
March built on every one of these threads — notifications, typed tags, ChatGPT-ready MCP, and a much richer diagnostics view. Reach out if you want to see any of it live, or to ask about where Ceivo is heading from here.