Merge pull request #2239 from asimurka/feature/fix-safety-identifier

LCORE-3192: Fix safety identifier

Author: tisnik

pyproject.toml

@@ -248,7 +248,7 @@ extend-exclude = ["tests/profiles/syntax_error.py"]
 
 [tool.ruff.lint]
 extend-select = ["TID251", "UP006", "UP007", "UP010", "UP017", "UP035", "RUF100", "B009", "B010", "DTZ005", "D202", "I001", "PLR1733"]
-
+ignore = ["UP040", "UP047", "UP045", "BLE", "S", "C", "RUF", "SIM", "B017", "TRY004", "TRY201", "TRY203", "TRY401", "UP008", "UP012", "UP024", "UP041", "B008", "EXE001", "FURB129", "G201", "ISC004", "LOG014", "PERF402", "PIE790", "PLW1510", "PYI034", "PYI064", "RET501", "TC004"]
 
 [tool.ruff.lint.flake8-tidy-imports.banned-api]
 unittest = { msg = "use pytest instead of unittest" }

scripts/gen_doc.py

@@ -67,8 +67,6 @@ def generate_documentation_on_path(path: Path) -> None:
     ----------
         path (str or os.PathLike): Directory in which to generate the README.md file.
     """
-    directory = path
-
     # directory can be skipped if it's part of DIRS_TO_SKIP global list.
     if path.as_posix() in DIRS_TO_SKIP:
         print(f"[gendoc] Skipping {path}")

scripts/vulnerability_report.py

@@ -40,9 +40,8 @@
 from statistics import median
 from typing import Any
 
-from dateutil import parser
-
 import matplotlib.pyplot as plt
+from dateutil import parser
 from matplotlib.figure import Figure
 
 type DependabotAlert = dict[str, Any]

src/app/endpoints/responses.py

@@ -582,7 +582,9 @@ async def handle_streaming_response(
         inference_start_time = time.monotonic()
         try:
             response = await context.client.responses.create(
-                **api_params.model_dump(exclude_none=True)
+                **api_params.model_dump(
+                    exclude_none=True, exclude={"safety_identifier"}
+                )
             )
             generator = response_generator(
                 stream=cast(AsyncIterator[OpenAIResponseObjectStream], response),
@@ -903,6 +905,9 @@ async def response_generator(
                 chunk_dict["response"]["conversation"] = normalize_conversation_id(
                     api_params.conversation
                 )
+                chunk_dict["response"][
+                    "safety_identifier"
+                ] = api_params.safety_identifier
                 _sanitize_response_dict(
                     chunk_dict["response"],
                     configured_mcp_labels,
@@ -1082,7 +1087,9 @@ async def handle_non_streaming_response(
             api_response = cast(
                 OpenAIResponseObject,
                 await context.client.responses.create(
-                    **api_params.model_dump(exclude_none=True)
+                    **api_params.model_dump(
+                        exclude_none=True, exclude={"safety_identifier"}
+                    )
                 ),
             )
             _record_response_inference_result(
@@ -1182,6 +1189,7 @@ async def handle_non_streaming_response(
     response = ResponsesResponse.model_validate(
         {
             **response_dict,
+            "safety_identifier": api_params.safety_identifier,
             "available_quotas": available_quotas,
             "conversation": normalize_conversation_id(api_params.conversation),
             "completed_at": int(completed_at.timestamp()),

tests/e2e/features/llama_stack_disrupted.feature

@@ -276,4 +276,4 @@ Feature: Llama Stack connection disrupted
     {"question": "How do I list files?"}
     """
     Then The status code of the response is 503
-    And The body of the response contains Llama Stack
+    And The body of the response contains OGX