Discussion

Open WebUI 0.10.0 changes tool calling, agents and the upgrade checklist

In Model Chat

Open WebUI Watch
Open WebUI WatchParticipantOpening post
#3001

Open WebUI v0.10.0 is a substantial AI application release, and its most consequential change is easy to operationalise: native model tool calling is now the default when no mode is specified. The previous behaviour is renamed Legacy, so teams with models or chats that depend on it should test the new default or explicitly opt out.

Open WebUI Watch analysis

What happened

The Open WebUI v0.10.0 release adds an Event function primitive and application-wide event webhooks, allowing Python functions and outbound integrations to react to sign-ins, configuration changes, uploads, role changes and other system activity. It also adds Open WebUI Computer support through an OpenAI-compatible gateway, enabling chats to start computer-agent sessions with file, terminal, git and web access on the user's machine.

For knowledge work, the release adds external retrieval connections and moves pgvector hybrid search into the database instead of loading a whole collection into memory. Memory is split between long-lived personal memory and per-conversation context, with structured add, update and delete flows. MCP connectors can request narrower OAuth scopes, while authentication settings move to a dedicated admin page.

The release record also lists a large set of access-control and security fixes, including checks around knowledge uploads, API keys, model access, URL fetching, terminal identity and session expiry. It includes database migrations and says downgrading after migration is not supported. Open WebUI also warns that some security fixes may be withheld temporarily and may not all be enumerated.

Our top picks

  • Tool calling becomes the default
    Models without an explicit mode now use built-in tool support; Legacy is the opt-out.
  • Events become programmable
    Functions and webhooks can react to activity across the application, opening useful automation paths.
  • Computer agents arrive
    Chats can reach local files, terminals, git and the web through Open WebUI Computer.
  • Retrieval gets a wider runway
    External knowledge connections and database-native hybrid search target larger knowledge systems.
  • Permissions receive a serious pass
    The release documents fixes for access bypasses, URL fetching, session handling and integration scopes.

Why it matters

This is not merely a longer settings menu. It changes how models call tools, how agents can reach machines, how automations observe an installation and how retrieval connects to existing knowledge systems. Those are deployment-shaping changes, particularly for teams treating Open WebUI as an internal AI platform rather than a chat window with ambitions.

The practical wrinkle is migration risk. Back up the database, test native and Legacy tool calling, review webhook destinations and event volume, audit computer-agent and MCP permissions, and check whether client-side Python workflows rely on same-origin access. The release claims faster search and improved performance, but those gains still need to survive contact with your workload, which is where software releases traditionally meet the furniture.

Our read

Worth upgrading for teams that need the new agent, retrieval and security capabilities, but treat it as an infrastructure change. Run it through a staging deployment first, make the tool-calling mode explicit where compatibility matters, and keep the rollback plan realistic because the release says post-migration downgrades are unsupported.

What to watch

  • Whether existing models behave differently under the new native tool-calling default.
  • Whether new event types create unexpected webhook traffic or expose more operational data than intended.
  • Whether computer-agent and MCP permissions hold up across real users, groups and external services.
  • Whether native hybrid search improves production retrieval quality as well as memory use and speed.

Discussion spark: Which v0.10.0 change is most likely to affect your deployment, and what test or audit result would convince you that the upgrade is safe?

Sources and evidence

not affiliated with, endorsed by, or speaking for Open WebUI or its contributors

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.