Restore RAG details in Local Dashboard

[Marco Russo (marcorusso97)]

With some prior PR, the detailed view of the RAG attack on the local dashboard was lost. This PR restores it.

[Copilot]

Pull request overview

Restores the detailed RAG (and related indirect injection) trace rendering in the local dashboard, and adjusts Textual snapshot tests to avoid path-with-spaces import failures when running snap_compare.

Changes:

• Route "rag" attack types through the existing indirect prompt injection detail renderer in the dashboard (trace view + history report detail).
• Update history detail pre-parsing so RAG uses the raw serialized trace list (to be rendered via the indirect injection view).
• Update TUI snapshot tests to instantiate apps directly instead of relying on pytest-textual-snapshot’s file-path import flow that can break on repo paths containing spaces.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
tests/unit/cli/tui/test_view_snapshots.py	Reworks snapshot test loading to bypass import_app/shlex.split path issues by importing the module and creating an app instance directly.
hackagent/server/dashboard/_trace_analysis_mixin.py	Restores RAG detail rendering in the attack-specific trace loader by routing "rag" to _render_indirect_injection_view.
hackagent/server/dashboard/_run_history_results_mixin.py	Ensures RAG history detail data uses serialized traces so the indirect injection view can render details.
hackagent/server/dashboard/_reports_mixin.py	Restores RAG goal-detail rendering in the history/report detail panel via _render_indirect_injection_view.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

❌ Patch coverage is 74.57627% with 15 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ent/server/dashboard/_run_history_results_mixin.py	72.72%	15 Missing ⚠️

📢 Thoughts on this report? Let us know!

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

tests/unit/cli/tui/test_view_snapshots.py:45

• _load_app_instance always creates a second Textual App instance because the snapshot app modules themselves already construct a module-level app = ...App() during exec_module(). Returning a new ...App() here doubles initialization work and can introduce extra side effects/global state in Textual. Consider returning the module’s app when available (and only falling back to instantiating class_name).

    path = _APPS / app_file
    spec = importlib.util.spec_from_file_location(path.stem, path)
    assert spec is not None and spec.loader is not None
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

tests/unit/cli/tui/test_view_snapshots.py:46

• _load_app_instance imports the snapshot module (which already constructs a module-level app = ...App()), then instantiates a second App via getattr(module, class_name)(). This creates an unused extra App instance and may introduce side effects/flakiness; it also diverges from how pytest-textual-snapshot historically consumes these files (module-level app). Prefer returning module.app when present, and only fall back to constructing the class if needed.

    spec = importlib.util.spec_from_file_location(path.stem, path)
    assert spec is not None and spec.loader is not None
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)
    return getattr(module, class_name)()

[Marco Russo (marcorusso97)]

Copilot fix merge conflicts

[Copilot]

Copilot fix merge conflicts

Merge conflicts resolved in cb21e8e. Merged origin/main into this branch and combined the crescendo branch (from main) with the rag/indirect_prompt_injection handling (from this PR) in _build_history_goal_detail_data. Dashboard, snapshot, and crescendo tests all pass.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (3)

tests/unit/cli/tui/test_view_snapshots.py:46

• _load_app_instance imports the snapshot module which already creates a module-level app = ... instance, but then it instantiates the App class again. This creates an extra unused Textual App per test case and can introduce unintended side effects during import. Reuse the module-level app when it matches the requested class, and only instantiate as a fallback.

    spec = importlib.util.spec_from_file_location(path.stem, path)
    assert spec is not None and spec.loader is not None
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)
    return getattr(module, class_name)()

tests/unit/server/dashboard/test_rag_history_dispatch.py:202

• _build_history_goal_detail_data has a dedicated crescendo dispatch branch, but the "all branches" test suite doesn't cover it. This leaves the extracted dispatch helper partially untested and makes the class docstring inaccurate.

    def test_pair(self):
        page, result = self._build("pair")
        self.assertEqual(result, {"r1": "PAIR_RESULT"})

    def test_tap(self):

tests/unit/server/dashboard/test_rag_history_dispatch.py:132

• _AllParsersHistoryResultsPage is intended to stub every _parse_*_traces collaborator so the dispatch helper can be fully exercised, but it currently lacks _parse_crescendo_traces. Adding the stub prevents accidental AttributeError if a crescendo test is added (and makes the double match the production interface).

This issue also appears on line 198 of the same file.

    def _parse_pair_traces(self, traces):
        self.calls.append(("pair", traces))
        return "PAIR_RESULT"

    def _parse_tap_traces(self, traces):

[Nicola Franco (franconicola)]

LGTM