FEAT Wire frontend attack view to backend APIs by romanlutz · Pull Request #1371 · Azure/PyRIT

romanlutz · 2026-02-17T04:30:30Z

Add target management and real attack execution to the PyRIT frontend, replacing the echo stub with live backend communication.

Backend:

Add attack CRUD endpoints (create, list, get, messages, add message)
Replace target_unique_name with target_registry_name across models, routes, services, and mappers for clarity
Add TargetInfo model to nest target metadata in AttackSummary
Add target_registry_name to AddMessageRequest so the backend stays stateless (no reverse lookups from attack identifier)
Move lifespan init to pyrit_backend CLI; main.py warns if run directly

CLI:

Add FrontendCore.run_initializers_async() to consolidate initializer resolution used by both pyrit_backend and run_scenario_async
Move all deferred imports to module level in frontend_core.py
Refactor pyrit_backend to use FrontendCore two-step init pattern (initialize_async + run_initializers_async)

Frontend:

Add Config page with target list, create target dialog, and set-active-target flow
Wire ChatWindow to attacksApi: lazy attack creation on first message, send via PromptNormalizer, map backend responses to UI messages
Add messageMapper utils (backend DTO <-> frontend Message conversion)
Add full backend DTO types mirroring pyrit/backend/models
Support simulated_assistant role, error rendering, loading indicators, and media attachments (image/audio/video)

Tests:

Add/update 300+ backend unit tests covering attack service, mappers, target service, API routes, and main lifespan
Add 150+ frontend tests covering ChatWindow, TargetConfig, CreateTargetDialog, MessageList, api service, and messageMapper
Update test_frontend_core patch targets to match top-level imports

pyrit/backend/main.py

pyrit/cli/frontend_core.py

bashirpartovi · 2026-02-17T07:14:53Z

pyrit/backend/main.py

@@ -91,9 +96,3 @@ def setup_frontend() -> None:

 # Set up frontend at module load time (needed when running via uvicorn)
 setup_frontend()


maybe this should go into lifespan?

setup_frontend needs to happen at module load time otherwise uvicorn can't serve the routes. From what I understand, lifespan is for startup/shutdown tasks that are primarily async, not FastAPI route setup. That's why the comment is there right above 🙂 I'm happy to add to it to make this clearer if you have suggestions.

frontend/src/components/Config/CreateTargetDialog.tsx

frontend/src/components/Config/TargetConfig.tsx

hannahwestra25 · 2026-02-17T16:45:17Z

Can you add screenshots for the frontend

frontend/src/components/Config/TargetConfig.tsx

Backend: - Replace private CentralMemory._memory_instance access with try/except around the public get_memory_instance() API in the lifespan handler. Initialization: - Extract run_initializers_async() as a public function in pyrit.setup.initialization so initializer execution can be invoked without redundantly re-loading env files, resetting defaults, and re-creating the memory instance. - FrontendCore.run_initializers_async() now calls the new function directly instead of re-invoking initialize_pyrit_async. - Export run_initializers_async from pyrit.setup. Frontend: - Extract TargetTable into its own component (TargetTable.tsx). - Move makeStyles definitions to co-located .styles.ts files for TargetConfig, TargetTable, and CreateTargetDialog. - Remove redundant explicit generic in useState<string>('') calls. - Use FluentUI Field validationMessage/validationState props for inline field-level validation in CreateTargetDialog. Tests: - Update TestRunScenarioAsync patches to mock run_initializers_async instead of initialize_pyrit_async.

Copilot

Pull request overview

Wires the PyRIT frontend attack experience to live backend APIs, including target management, attack execution, conversation branching, and richer message rendering.

Changes:

Added/updated backend attack + target endpoints and DTOs (including conversation summaries and target metadata).
Refactored initialization flow to shift lifespan init into CLI and consolidate initializer execution.
Expanded frontend UI (config/history/chat) and added extensive unit + E2E coverage.

Reviewed changes

Copilot reviewed 84 out of 88 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
tests/unit/registry/test_converter_registry.py	Adds unit tests for converter registry singleton/metadata behavior.
tests/unit/memory/test_sqlite_memory.py	Adds tests for new conversation stats aggregation in SQLite memory.
tests/unit/cli/test_pyrit_backend.py	Adds tests for CLI config-file forwarding and server startup flow.
tests/unit/cli/test_frontend_core.py	Updates patch targets and initializer execution expectations after refactor.
tests/unit/backend/test_target_service.py	Updates tests for target registry naming changes.
tests/unit/backend/test_main.py	Updates lifespan expectations to “warn only” behavior.
pyrit/setup/initializers/airt_targets.py	Adds extra kwargs support and new/renamed target presets.
pyrit/setup/initializers/airt.py	Switches auth approach to Entra token providers and updates required vars.
pyrit/setup/initialization.py	Extracts `run_initializers_async` to separate initializer execution from memory init.
pyrit/setup/init.py	Exposes `run_initializers_async` as part of setup public API.
pyrit/prompt_target/openai/openai_video_target.py	Enforces single-turn conversation constraint for video target.
pyrit/prompt_target/openai/openai_target.py	Refactors OpenAI base target inheritance/init to align with PromptTarget.
pyrit/prompt_target/openai/openai_response_target.py	Includes target-specific params in identifiers (e.g., extra body params).
pyrit/prompt_target/openai/openai_realtime_target.py	Adjusts inheritance to keep chat-target semantics for realtime target.
pyrit/prompt_target/openai/openai_image_target.py	Enforces single-turn conversation constraint for image target.
pyrit/models/conversation_stats.py	Introduces ConversationStats aggregate model.
pyrit/models/attack_result.py	Adds `attack_result_id` onto domain AttackResult.
pyrit/models/init.py	Exports ConversationStats from models package.
pyrit/memory/sqlite_memory.py	Adds conversation stats query and refactors some filtering helpers/update behavior.
pyrit/memory/memory_models.py	Maps DB primary key into `AttackResult.attack_result_id`.
pyrit/memory/memory_interface.py	Adds conversation stats API and updates attack result insert/update semantics.
pyrit/memory/azure_sql_memory.py	Adds Azure SQL implementation for conversation stats and safer update behavior.
pyrit/cli/pyrit_backend.py	Refactors CLI to use FrontendCore two-step init and adds `--config-file`.
pyrit/cli/frontend_core.py	Moves deferred imports to module-level and adds `run_initializers_async` method.
pyrit/backend/services/target_service.py	Renames target id field to registry name and updates pagination cursor.
pyrit/backend/routes/version.py	Adds database backend info to version response payload.
pyrit/backend/routes/targets.py	Renames path params and docs to registry naming scheme.
pyrit/backend/routes/attacks.py	Expands attack routes to support conversations and changes identifiers to attack_result_id.
pyrit/backend/models/targets.py	DTO rename + adds `supports_multiturn_chat`.
pyrit/backend/models/attacks.py	Adds target metadata nesting, conversation endpoints, and new message request fields.
pyrit/backend/models/init.py	Updates exports for renamed message response DTO.
pyrit/backend/mappers/target_mappers.py	Maps multiturn capability and renames target id field in DTO mapping.
pyrit/backend/mappers/init.py	Renames exported async mapper function.
pyrit/backend/main.py	Removes standalone uvicorn runner and changes lifespan to warn when uninitialized.
frontend/src/utils/messageMapper.ts	Adds backend DTO ↔ UI Message mapping (attachments, reasoning, errors).
frontend/src/types/index.ts	Adds backend DTO type mirrors and expands UI message model.
frontend/src/services/api.ts	Adds targets/attacks/labels API clients and query serialization.
frontend/src/services/api.test.ts	Expands mocked API service tests for new endpoints.
frontend/src/components/Sidebar/Navigation.tsx	Adds navigation views (chat/history/config) and active styling.
frontend/src/components/Sidebar/Navigation.test.tsx	Updates navigation tests for new view switching behavior.
frontend/src/components/Layout/MainLayout.tsx	Shows DB info in version tooltip and wires navigation callbacks.
frontend/src/components/Layout/MainLayout.test.tsx	Updates layout tests for new navigation props and DB tooltip behavior.
frontend/src/components/Labels/LabelsBar.test.tsx	Adds unit tests for labels UI behavior and label fetching.
frontend/src/components/Config/TargetTable.tsx	Adds target list table UI with active-target selection controls.
frontend/src/components/Config/TargetTable.styles.ts	Adds styling for target table and active row highlighting.
frontend/src/components/Config/TargetConfig.tsx	Implements target config page with fetch/retry, refresh, and create dialog.
frontend/src/components/Config/TargetConfig.test.tsx	Adds tests for config page states and interactions.
frontend/src/components/Config/TargetConfig.styles.ts	Adds styling for config page layout and states.
frontend/src/components/Config/CreateTargetDialog.tsx	Adds create-target dialog and validation + submit flow.
frontend/src/components/Config/CreateTargetDialog.test.tsx	Adds tests for create-target dialog validation and submission.
frontend/src/components/Config/CreateTargetDialog.styles.ts	Adds styling for create-target dialog layout.
frontend/src/components/Chat/InputBox.tsx	Adds banners/locking states, ref API, and multiturn warnings for active target.
frontend/src/components/Chat/InputBox.test.tsx	Adds tests for new input-box behaviors (single-turn, ref attachments, banners).
frontend/src/components/Chat/ConversationPanel.tsx	Adds conversation list panel for attacks with promote-to-main and new conversation actions.
frontend/src/components/Chat/ConversationPanel.test.tsx	Adds tests for conversation panel rendering and interactions.
frontend/src/App.tsx	Introduces multi-view app shell, target selection, attack loading, and global labels.
frontend/src/App.test.tsx	Expands app tests for navigation, target selection, and opening historical attacks.
frontend/playwright.config.ts	Splits Playwright projects into seeded vs live modes.
frontend/package.json	Adds e2e scripts for seeded and live test projects.
frontend/eslint.config.js	Adds Node globals for Playwright e2e files.
frontend/e2e/config.spec.ts	Adds e2e coverage for config page and config↔chat flow.
frontend/e2e/api.spec.ts	Adds e2e API smoke tests (targets/attacks) and improves slow-backend handling.
frontend/e2e/accessibility.spec.ts	Updates a11y coverage for new navigation and config table; adjusts expected header text.
frontend/dev.py	Improves dev runner process management, adds detach/logs/config-file support.
frontend/README.md	Documents seeded vs live e2e modes.
.github/workflows/frontend_tests.yml	Runs seeded-only e2e in GitHub Actions.
.gitattributes	Adds union merge strategy for squad log/state files.
.devcontainer/devcontainer_setup.sh	Makes Playwright install failures non-blocking with clearer messaging.

Copilot · 2026-02-28T14:25:05Z

pyrit/memory/sqlite_memory.py

        from pyrit.memory.memory_models import AttackResultEntry, PromptMemoryEntry

-        return exists().where(
+        targeted_harm_categories_subquery = exists().where(


Both _get_attack_result_harm_category_condition and _get_attack_result_label_condition used to return an exists().where(...) condition, but now only assign it to a local variable (*_subquery) without returning it. That will make these helpers return None, breaking filtering logic that expects a SQLAlchemy condition. Return the constructed subquery (or revert to the inline return exists().where(...)).

Copilot · 2026-02-28T14:25:05Z

pyrit/memory/sqlite_memory.py

        from pyrit.memory.memory_models import AttackResultEntry, PromptMemoryEntry

-        return exists().where(
+        labels_subquery = exists().where(


Both _get_attack_result_harm_category_condition and _get_attack_result_label_condition used to return an exists().where(...) condition, but now only assign it to a local variable (*_subquery) without returning it. That will make these helpers return None, breaking filtering logic that expects a SQLAlchemy condition. Return the constructed subquery (or revert to the inline return exists().where(...)).

Copilot · 2026-02-28T14:25:05Z

pyrit/backend/routes/attacks.py

+        tb = traceback.format_exception(type(e), e, e.__traceback__)
+        # Include the root cause if chained
+        cause = e.__cause__
+        if cause:
+            tb += traceback.format_exception(type(cause), cause, cause.__traceback__)
+        detail = "".join(tb)
        raise HTTPException(
            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
-            detail=f"Failed to add message: {str(e)}",
+            detail=detail,
        ) from e


Returning full stack traces in the HTTP 500 detail leaks internal code structure and potentially sensitive runtime data to clients. Log the traceback server-side (optionally include request IDs), and return a generic error message to the client; if you want trace exposure for development, gate it behind a dev/debug flag.

Copilot · 2026-02-28T14:25:06Z

pyrit/prompt_target/openai/openai_target.py



-class OpenAITarget(PromptChatTarget):
+class OpenAITarget(PromptTarget):


Now that OpenAI targets use multiple inheritance elsewhere (e.g., RealtimeTarget(OpenAITarget, PromptChatTarget)), calling PromptTarget.__init__ directly bypasses cooperative initialization and can break MRO-based init chaining. Prefer super().__init__(...) in OpenAITarget.__init__ so all bases in the MRO can initialize correctly.

Copilot · 2026-02-28T14:25:06Z

pyrit/prompt_target/openai/openai_target.py

+        PromptTarget.__init__(
            self,


Now that OpenAI targets use multiple inheritance elsewhere (e.g., RealtimeTarget(OpenAITarget, PromptChatTarget)), calling PromptTarget.__init__ directly bypasses cooperative initialization and can break MRO-based init chaining. Prefer super().__init__(...) in OpenAITarget.__init__ so all bases in the MRO can initialize correctly.

Suggested change

PromptTarget.__init__(

self,

super().__init__(

Copilot · 2026-02-28T14:25:07Z

frontend/src/components/Chat/ConversationPanel.tsx

+
+  useEffect(() => {
+    fetchConversations()
+  }, [fetchConversations, activeConversationId])


Including activeConversationId in this effect dependency will refetch the full conversations list on every conversation selection, creating unnecessary API traffic and UI churn. Consider fetching on attackResultId changes (and after mutating actions like create/promote), but not on local selection changes.

Suggested change

}, [fetchConversations, activeConversationId])

}, [fetchConversations])

Copilot · 2026-02-28T14:25:07Z

tests/unit/memory/test_sqlite_memory.py

+    from pyrit.models import Message
+    from unit.mocks import get_sample_conversations


Several new imports are placed inside test functions. For consistency and readability, prefer moving these imports to the module level (unless there’s a specific reason to defer import side effects).

Copilot · 2026-02-28T14:25:07Z

tests/unit/memory/test_sqlite_memory.py

+    import uuid
+
+    from pyrit.models import MessagePiece


Several new imports are placed inside test functions. For consistency and readability, prefer moving these imports to the module level (unless there’s a specific reason to defer import side effects).

Copilot · 2026-02-28T14:25:07Z

tests/unit/memory/test_sqlite_memory.py

+    import uuid
+
+    from pyrit.models import MessagePiece


Several new imports are placed inside test functions. For consistency and readability, prefer moving these imports to the module level (unless there’s a specific reason to defer import side effects).

Copilot · 2026-02-28T14:25:07Z

tests/unit/memory/test_sqlite_memory.py

+    import uuid
+
+    from pyrit.models import MessagePiece


Several new imports are placed inside test functions. For consistency and readability, prefer moving these imports to the module level (unless there’s a specific reason to defer import side effects).

Copilot

Pull request overview

Copilot reviewed 84 out of 88 changed files in this pull request and generated 4 comments.

Comments suppressed due to low confidence (1)

pyrit/memory/sqlite_memory.py:1

This function used to return the exists().where(...) condition, but now assigns it to labels_subquery and never returns it. That will cause callers to append None to SQLAlchemy conditions and break attack filtering by labels. Return labels_subquery (or inline it back into a return statement).

# Copyright (c) Microsoft Corporation.

Copilot · 2026-02-28T14:36:28Z

pyrit/memory/memory_interface.py

+        from contextlib import closing
+
+        with closing(self.get_session()) as session:
+            from sqlalchemy.exc import SQLAlchemyError
+
+            try:
+                session.add_all(entries)
+                session.commit()
+                # Populate the attack_result_id back onto the domain objects so callers
+                # can reference the DB-assigned ID immediately after insert.
+                for ar, entry in zip(attack_results, entries):
+                    ar.attack_result_id = str(entry.id)
+            except SQLAlchemyError:
+                session.rollback()
+                raise


The new inline insert logic introduces deferred/local imports (closing, SQLAlchemyError). This makes the method harder to reason about and diverges from the project’s import organization (imports at module top). Move these imports to the top of the file (or reuse already-imported equivalents) and keep the method focused on DB operations.

Copilot · 2026-02-28T14:36:28Z

frontend/e2e/api.spec.ts

+  test("should list targets", async ({ request }) => {
+    const response = await request.get("/api/targets?count=50");
+
+    expect(response.ok()).toBe(true);
+    const data = await response.json();
+    expect(data).toHaveProperty("items");
+    expect(Array.isArray(data.items)).toBe(true);
+  });
+
+  test("should create and retrieve a target", async ({ request }) => {
+    const createPayload = {
+      target_type: "OpenAIChatTarget",
+      params: {
+        endpoint: "https://e2e-test.openai.azure.com",
+        model_name: "gpt-4o-e2e-test",
+        api_key: "e2e-test-key",
+      },
+    };
+
+    const createResp = await request.post("/api/targets", { data: createPayload });
+    // The endpoint may not be implemented, may require different schema, or may
+    // return a validation error.  Skip when the backend cannot handle the request.
+    if (!createResp.ok()) {
+      test.skip(true, `POST /api/targets returned ${createResp.status()} — skipping`);
+      return;
+    }


The E2E API tests don’t align with the backend contract: targets list uses limit (not count), and create uses type (not target_type). As written, the create test will likely always skip, reducing the value of the suite. Update these requests to match the API schema and avoid conditional skipping for expected/steady-state flows (use deterministic mocks or dedicated seeded endpoints if needed).

Copilot · 2026-02-28T14:36:29Z

frontend/src/utils/messageMapper.ts

+  const mime = mimeField || defaultMimeForDataType(dataType)
+  const isBase64 = !value.startsWith('data:') && !value.startsWith('http')
+  const url = isBase64 ? buildDataUri(value, mime) : value
+  const prefix = isOriginal ? 'original_' : ''


The base64 detection treats any non-data: and non-http string as base64 and wraps it into a data URI. Backend values can be file paths (e.g., /tmp/x.png, C:\\...) or non-http URLs (blob:, file:, azblob:), which will become invalid data URIs. Consider a stricter base64 check (e.g., regex/length validation) and/or explicit handling for known path/scheme prefixes before deciding to build a data URI.

Copilot · 2026-02-28T14:36:29Z

frontend/src/components/Sidebar/Navigation.tsx

      <Button
-        className={styles.iconButton}
+        className={currentView === 'chat' ? styles.activeButton : styles.iconButton}
        appearance="subtle"
        icon={<ChatRegular />}
        title="Chat"
-        disabled
+        onClick={() => onNavigate('chat')}
+      />
+
+      <Button
+        className={currentView === 'history' ? styles.activeButton : styles.iconButton}
+        appearance="subtle"
+        icon={<HistoryRegular />}
+        title="Attack History"
+        onClick={() => onNavigate('history')}
+      />
+
+      <Button
+        className={currentView === 'config' ? styles.activeButton : styles.iconButton}
+        appearance="subtle"
+        icon={<SettingsRegular />}
+        title="Configuration"
+        onClick={() => onNavigate('config')}
      />


These icon-only buttons rely on the title attribute for labeling. title is not consistently announced by screen readers and isn’t a reliable accessible name. Provide an explicit accessible label (e.g., aria-label) so the buttons have stable names for assistive tech and testing via role/name queries.

- Add run_initializers_async to pyrit.setup for programmatic initialization - Switch AIRTInitializer to Entra (Azure AD) auth, removing API key requirements - Add --config-file flag to pyrit_backend CLI - Use PyRIT configuration loader in FrontendCore and pyrit_backend - Update AIRTTargetInitializer with new target types Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Add conversation_stats model and attack_result extensions - Add get_attack_results with filtering by harm categories, labels, attack type, and converter types to memory interface - Implement SQLite-specific JSON filtering for attack results - Add memory_models field for targeted_harm_categories - Add prompt_metadata support to openai image/video/response targets - Fix missing return statements in SQLite harm_category and label filters Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…nd-attack-api

- Add attack CRUD routes with conversation management - Add message sending with target dispatch and response handling - Add attack mappers for domain-to-DTO conversion with signed blob URLs - Add attack service with video remix support and piece persistence - Expand target service and routes with registry-based target management - Add version endpoint with database info Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Add attack-centric chat UI with multi-conversation support - Add conversation panel with branching and message actions - Add attack history view with filtering - Add labels bar for attack metadata - Add target configuration with create dialog - Add message mapper utilities for backend/frontend translation - Add video playback support with signed blob URLs - Add InputBox with attachment support and auto-expand - Update dev.py with --detach, logs, and process management - Add e2e tests for chat, config, and flow scenarios Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 83 out of 88 changed files in this pull request and generated 4 comments.

Copilot · 2026-02-28T14:57:29Z

pyrit/backend/routes/version.py

+        db_name = None
+        if memory.engine.url.database:
+            db_name = memory.engine.url.database.split("?")[0]
+        database_info = f"{db_type} ({db_name})" if db_name else db_type


Exposing memory.engine.url.database in a public /version response can leak sensitive deployment details (e.g., server/database names, file paths). Consider returning only the backend type (or a redacted form), and/or only including database identifiers when running in a trusted/development mode.

Suggested change

db_name = None

if memory.engine.url.database:

db_name = memory.engine.url.database.split("?")[0]

database_info = f"{db_type} ({db_name})" if db_name else db_type

database_info = db_type

Copilot · 2026-02-28T14:57:29Z

pyrit/backend/models/attacks.py

+    conversation_id: str = Field(..., description="Unique conversation identifier")
+    message_count: int = Field(0, description="Number of messages in this conversation")
+    last_message_preview: Optional[str] = Field(None, description="Preview of the last message")
+    created_at: Optional[str] = Field(None, description="ISO timestamp of the first message")


created_at is modeled as Optional[str] here while other timestamps in the same API use datetime (e.g., AttackSummary.created_at/updated_at, CreateConversationResponse.created_at). Using datetime consistently improves schema clarity and avoids clients guessing formats; Pydantic will serialize it as ISO 8601 anyway.

Suggested change

created_at: Optional[str] = Field(None, description="ISO timestamp of the first message")

created_at: Optional[datetime] = Field(None, description="ISO timestamp of the first message")

Copilot · 2026-02-28T14:57:30Z

frontend/src/services/api.test.ts

      }),
    },
+    targetsApi: {
+      listTargets: jest.fn(async (limit = 50, cursor?: string) => {
+        const params: Record<string, string | number> = { limit };
+        if (cursor) params.cursor = cursor;
+        const response = await mockApiClient.get("/targets", { params });
+        return response.data;
+      }),


This test suite mocks the module under test (./api) and re-implements the production logic inside the mock, which means it won’t catch regressions in frontend/src/services/api.ts. Prefer importing the real targetsApi/attacksApi and mocking only apiClient (e.g., via jest spies or an axios mock adapter) so the tests validate the actual implementation.

Copilot · 2026-02-28T14:57:30Z

pyrit/prompt_target/openai/openai_image_target.py

+        request = text_pieces[0]
+        messages = self._memory.get_conversation(conversation_id=request.conversation_id)
+
+        n_messages = len(messages)
+        if n_messages > 0:
+            raise ValueError(
+                "This target only supports a single turn conversation. "
+                f"Received: {n_messages} messages which indicates a prior turn."
+            )


This loads the entire conversation just to check whether any prior turn exists. A cheaper approach is to query only for existence/first row (or count with a limit), which avoids pulling full histories into memory when a user accidentally reuses a conversation ID.

bashirpartovi reviewed Feb 17, 2026

View reviewed changes