Repair Python E2E target propagation and notebook examples, correct the
Responses vision payload, and isolate and harden frontend E2E validation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Repair scanner isolation, provider resource cleanup, Hugging Face downloads, Windows-safe output, and executable notebook examples discovered by exhaustive post-merge validation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Use the newly configured adversarial_chat target in the benchmark end-to-end smoke test after confirming registration and a clean 7/7 full scenario run.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Remove the generic scorer lifecycle method and threshold-wrapper delegation while retaining Azure Content Safety's explicit resource cleanup.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Store the audited execution results for all runnable documentation notebooks so output cells can be reviewed directly in the branch changeset.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Retain Realtime WebSocket cleanup as a public API while moving ordinary OpenAI teardown behind a private hook and removing validation-only cleanup calls from documentation notebooks.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…lidation branch

Resolved notebook conflicts in doc/code/converters/1_text_to_text_converters and doc/code/executor/1_single_turn by regenerating executed outputs from the merged .py sources. Verified affected tests (110 passed), full pre-commit, and notebook cell-output audit (0 errors).

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…-2157

Preserve refreshed executed outputs prior to removing split-owned code
changes and merging latest main.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Merge latest origin/main (27d20ee). Resolutions:
- 26 A-G split-owned code/test files reset to origin/main (now covered by
  PRs microsoft#2208-2214), so this branch carries no production-code duplication.
- matrix_atomic_attack_builder(.py/test) taken from main (orphaned core
  refactor does not belong on a notebook-validation branch).
- Executed notebook .ipynb outputs kept from this branch (regenerated next).

Result vs main: notebooks + notebook-source hardening + docs + frontend e2e only.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

The first TextAdaptive() cell in 3_adaptive_scenarios.ipynb runs many
objectives across several attempts and exceeds the default 600s per-cell
timeout. Give that notebook a 1800s per-cell timeout via a per-notebook
override, leaving other scenario notebooks at the 600s default.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Sync validation branch with latest main. Resolved doc/scanner/airt.ipynb
conflict (both sides changed source) by regenerating the notebook from the
merged jupytext-paired airt.py.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Revert 40 notebooks whose cell SOURCE is byte-identical to origin/main
(only re-executed LLM outputs / execution_count / metadata differed) plus
2 jupytext-paired .py files that differed only by a jupytext_version header
stamp. pyrit/** is identical to origin/main, so these diffs cannot reflect
any behavior change. Keeps only files with real source/doc/test changes.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

These 4 notebooks carried real .ipynb source edits (added Context Compliance
section, reworded modality-feedback objective/scorer, heading numbering,
garak encoding-doc trim) that were not reflected in their jupytext-paired
.py. Regenerated each .py from its .ipynb so the pair is consistent,
preserving each file's original jupytext_version stamp so the diff shows
only the real source change.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Main dropped numeric prefixes from notebook H1 titles. Sync the validation
branch:
- Un-number kept notebooks (real source edits remain):
  3_image_converters ('# Image Converters'),
  8_non_llm_targets ('# Azure Blob Storage Targets').
- After un-numbering, 1_loading_datasets and 2_openai_responses_target
  differ from main only by re-executed output noise, so revert both
  pairs (.ipynb + .py) to origin/main; they drop from the changeset.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Remove the DejaVu-Sans font_manager workaround (and the try/except unwrap)
from doc/code/converters/3_image_converters.{ipynb,py}, restoring main's
version. The underlying cause -- AddTextImageConverter defaulting to the
non-cross-platform "helvetica.ttf" -- is being fixed at the source (default
to None -> Pillow bundled default) in a separate session, after which this
notebook runs cleanly on its plain defaults with no font warning.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…nce executor)

The branch added a "Context Compliance" section importing
ContextComplianceAttack from pyrit.executor.attack, plus its supporting
adversarial-target setup and narrative. That class no longer exists in main
(context compliance was migrated out of executor/attack), so the section is
broken against main and the change is unnecessary. Restore main's version of
doc/code/executor/1_single_turn.{ipynb,py}.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…orkaround)

5_workflow.ipynb is explicitly CI-skipped (test_executor_notebooks.py skipped_files) because it needs a live Azure Storage Account + local AI-recruiter service. The branch had added runtime storage_available/recruiter_available guards and re-executed the notebook in an infra-less environment, which replaced main's preserved XPIAStatus.SUCCESS demonstration outputs with 'example not run: unavailable' skip messages. Main's version still resolves against current API, so restore it verbatim.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…fix)

The validation branch's 8_modality_feedback carried a simplified 5-criteria
objective (plus a text-only profile and max_turns=8) that predated and
discarded microsoft#2188 "FIX: Preserve iterative image state in Crescendo". Restore
the origin/main version so the notebook keeps the detailed albatross
objective, 7-criteria scorer, and iterative-image-state fix.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…aults doc)

The validation branch's garak notebook trimmed the "Aggregate techniques:
ALL / DEFAULT" markdown block that microsoft#2190 "Standardize garak.encoding
defaults" added on main. Restore the origin/main version so the DEFAULT/ALL
aggregate-technique documentation is preserved.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…soft#2222)

The Playwright host-resolution change on this validation branch duplicates
and conflicts with PR microsoft#2222 "FIX Stabilize Playwright frontend host
resolution", which is the dedicated PR for playwright.config.ts. Revert to
origin/main so microsoft#2222 owns that file. (The e2e timing-stabilization tweaks in
api.spec.ts / flows.spec.ts are independent and left in place.)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…ion into PR microsoft#2222)

api.spec.ts and flows.spec.ts carried e2e timing-stabilization tweaks
(explicit /api/health timeouts, a toBeVisible wait, a beforeAll setTimeout).
Per decision to consolidate all Playwright e2e stabilization into PR microsoft#2222,
revert these to origin/main here; the tweaks are folded into microsoft#2222 alongside
the playwright.config.ts host-resolution fix.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…ch main)

Undo the verbose TargetRegistry.get_registry_singleton().instances.get(...)
pattern in round_robin_target, airt, and benchmark; restore main's canonical
OpenAIChatTarget() usage. These registry rewrites (and the entangled
validation-only bounding tweaks that referenced the registry vars) were
scaffolding added during pre-release validation and are not wanted in docs.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Undo the validation-only auth workaround (explicit endpoint/model + Entra
get_azure_openai_auth fallback). Main's canonical bare OpenAITextEmbedding()
resolves to Entra auth when no embedding key is set; the Entra path belongs in
the environment (.env), not hardcoded in the doc notebook.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Undo validation-only scaffolding: main lists/looks up scenarios from
import-time class registration (no init needed) and builds the target
directly via OpenAIChatTarget(), initializing late with just
LoadDefaultDatasets(). The added TargetInitializer/ScorerInitializer/
TechniqueInitializer are unused here and only add env-config fragility.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

…meout)

Restores the unbounded production-default adaptive example to match origin/main;
keeps the 1800s per-cell override in test_notebooks_scenarios.py so the auto-run
scenarios integration test has headroom instead of the default 600s.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

2_resiliency: drop unused Target/Scorer/Technique initializers (same scaffolding
reverted in 1_class_registry). 8_non_llm_targets: restore AZURE_SQL backend and the
unconditional Azure Blob example, dropping the storage-availability gating workaround.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

1_pyrit_scan: drop dynamic-port + explicit config-path scaffolding.
10_1_playwright_target: drop Windows sys.platform gating (skip-listed, not auto-run).
4_openai_video_target: drop unneeded scorer/temp-file cleanup + var extraction and the
+10313-line re-executed output committed during validation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448

Readers may not know what 'the initializer cleanup' refers to; frame it as the
consolidation/rename leading up to v1.0.0 so pre-v1.0.0 configs know when the old
names (simple/airt/scenario_technique/objective_list) were removed.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 52d2518d-ebe5-4888-8189-6146b6503448