[Issue]: Conversation-mode group selfies always use the first character

[Romuromylus]

Summary

In Conversation-mode group chats, selfies generated by any character always depict the first character in the group. The image — and its tags — match the first character regardless of which character actually took the selfie.

Expected behavior

A selfie should depict (and be tagged for) the character that took it.

Actual behavior

Every character's selfie is generated from the first group member's appearance, and the selfie's tags carry the first character's name.

Steps to reproduce

1. Create a Conversation-mode group chat with two or three visually distinct characters.
2. Have each character take a selfie.
3. Observe that all selfies match the description of the first character, and expanding the selfie shows the first character's name in the tags.

Environment

• Marinara Engine version: 2.0.5
• Install type: Installer
• OS + version: Windows 11
• Browser or app shell: Chrome

Logs, screenshots, or video

Reported in the community Discord.

Additional context

In Conversation-mode group chats the default response order resolves to "merged" generation (packages/server/src/routes/generate.routes.ts): a single generation produces every character's turn, and all commands parsed from that response — including [selfie] — are attributed to characterIds[0]. The selfie handler then builds the image prompt from that character's appearance/name. Conversation mode does not emit <speaker> tags, so the originating character is not currently recoverable at the command-attribution point; fixing this needs per-character attribution for merged-mode commands.

Template check

Please uncheck (untick) the box below before submitting so we know you read the template. It is intentionally pre-checked:

• I DID NOT read this template and provide the requested details.