PR #22988: Fix dashboard display role for vision_analyze pre-analysis

• Repository: NousResearch/hermes-agent
• Author: namekun
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/22988

Description (problem / solution / changelog)

Summary

• Normalize dashboard session message presentation so synthetic image pre-analysis is shown as a vision_analyze tool result instead of a user message.
• Keep the underlying stored session rows unchanged; this only affects the dashboard API response.
• Add focused tests for the dashboard message-role normalization helper.

Fixes #22961.

Tests

• python -m pytest tests/cli/test_web_server_dashboard_messages.py
• python -m py_compile hermes_cli/web_server.py

Changed files

• hermes_cli/web_server.py (modified, +26/-1)
• tests/cli/test_web_server_dashboard_messages.py (added, +32/-0)

PR #23036: fix(dashboard): display vision_analyze tool results with correct tool role styling

• Repository: NousResearch/hermes-agent
• Author: KhanCold
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/23036

Description (problem / solution / changelog)

Problem

In the Dashboard, vision_analyze tool results are stored with role: "tool" but the MessageBubble component's ROLE_STYLES lookup only handles user, assistant, and system. Tool messages fall through to the default system style, making them visually indistinguishable or incorrectly styled.

Solution

Add tool role mapping to the assistant style in ROLE_STYLES. Tool results (like vision_analyze outputs) now render with the same styling as assistant messages.

Testing

• Verified vision_analyze results display with assistant styling
• No regression for user/assistant/system roles

Fixes #22961

Changed files

• web/src/pages/SessionsPage.tsx (modified, +7/-1)