Skip to content

Conversation drawer shows the previous attack's conversations after switching attacks #3575

Open

Description

Description

The conversation panel keeps a single component instance across attack switches, and its state reset block only clears the loaded conversations when the new attack id is null. Switching from one attack to another therefore leaves the previous attack's conversation list rendered under the new attack's header until the new fetch resolves, with no loading indicator.

During that window a row click hands the old attack's conversation id to the new attack's context, and the star action can promote an old attack's conversation as the new attack's main conversation. The Chat route keeps the panel mounted across attack changes and forces the drawer open whenever related conversations exist, so the stale window is directly reachable by navigating between attacks.

Expected Behavior

Switching attacks should clear the previous attack's conversations, clear the main id and error, and show the loading state until the new attack's conversations resolve, mirroring the mount behavior.

Reproduction Steps

  1. Open an attack with conversations and the conversation drawer open.
  2. Navigate to a second attack.
  3. The drawer immediately shows the first attack's conversations under the second attack's header.

Minimal Reproduction

The render phase reset in ConversationPanel.tsx gates every setter behind if (!attackResultId).

Environment

PyRIT main.

Activity

  1. TrinityBerserker commented on Oct 27, 2026

    @TrinityBerserker

    Hi! I'd like to work on this issue. I've reviewed the description and have a good idea of where the problem is. I'll submit a PR once I have a fix.

  2. ManoharPaturi commented on Oct 27, 2026

    @ManoharPaturi
    ContributorAuthor

    🐦[T3T]Berserker 🐦 🐍 (@TrinityBerserker) Thanks for your interest in helping with this issue!

    Just to clarify, this issue was already assigned to me by Roman Lutz (@romanlutz), and I've submitted PR #3576 with the fix and a regression test.

    Since the issue is already being addressed, I'd appreciate it if we could avoid duplicate work. Would you mind closing your PR #3583?

    Thanks for your understanding!

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

Metadata

Metadata

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions