Skip to content

Transition handling for cached workspaceId arguments after the snake_case schema migration #373

Description

@JunkaiWang-TheoPhy

A ChatGPT connector created against v1.0.8 can retain a workspaceId tool argument after the server upgrades to the canonical workspace_id schemas. I reproduced an authenticated read call using the previous field name against beta.4: it returned Input validation error: Invalid arguments for tool read: workspace_id: Invalid input: expected string, received undefined. The corresponding canonical call succeeded. The workspace was valid; neither the path nor authentication was the cause.

#339 already contains a report of this upgrade/schema mismatch, and #303 explicitly chose not to add compatibility aliases. This issue proposes a narrower migration boundary for that identified cached client: translate an old workspaceId argument for workspace-scoped tool calls immediately before normal validation. Keep advertised schemas canonical, reject conflicting old/new values, and preserve all tool authorization and path checks. It does not propose legacy aliases throughout the domain model or changing model/host behavior.

Refreshing tools or recreating a connector is the existing client-side workaround. A transition adapter would allow existing conversations to keep working while the host refreshes its descriptors. I have a local reproduction and will submit a small companion PR with the old call failing before the change and succeeding after it, plus conflicting/invalid-input coverage. This is a proposal for maintainer review; it is not a claim that the documented snake_case migration was accidental.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions