Merge branch 'main' into logging

Author: jgarciao

.pre-commit-config.yaml

@@ -36,13 +36,13 @@ repos:
         exclude: .*/__snapshots__/.*|.*-input\.json$|^semgrep\.yaml$
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.15.6
+    rev: v0.15.7
     hooks:
       - id: ruff
       - id: ruff-format
 
   - repo: https://github.com/gitleaks/gitleaks
-    rev: v8.30.1
+    rev: v8.30.0
     hooks:
       - id: gitleaks
 
@@ -90,7 +90,7 @@ repos:
       - id: actionlint
 
   - repo: https://github.com/DavidAnson/markdownlint-cli2
-    rev: v0.21.0
+    rev: v0.22.0
     hooks:
       - id: markdownlint-cli2
         args:

CONSTITUTION.md

@@ -79,6 +79,7 @@ All code MUST consider security implications.
 - Avoid running destructive commands without explicit user confirmation
 - Use detect-secrets and gitleaks pre-commit hooks to prevent secret leakage
 - Test code MUST NOT introduce vulnerabilities into the tested systems
+- Use `utilities.path_utils.resolve_repo_path` to resolve and validate any user-supplied or parameterized file paths, preventing path-traversal and symlink-escape outside the repository root
 - JIRA ticket links are allowed in PRs and commit messages (our Jira is public)
 - Do NOT reference internal-only resources (Jenkins, Confluence, Slack threads) in code, PRs, or commit messages
 - Do NOT link embargoed or security-restricted (RH-employee-only) tickets

pyproject.toml

@@ -6,7 +6,7 @@ output-format = "grouped"
 extend-exclude = ["utilities/manifests"]
 
 [tool.ruff.lint]
-external = ["E501"]
+external = ["E501", "FCN001"]
 
 [tool.ruff.format]
 exclude = [".git", ".venv", ".mypy_cache", ".tox", "__pycache__", "utilities/manifests"]

tests/llama_stack/conftest.py

@@ -1,6 +1,5 @@
 import os
 from collections.abc import Callable, Generator
-from pathlib import Path
 from typing import Any
 
 import httpx
@@ -810,7 +809,6 @@ def vector_store(
             try:
                 vector_store_upload_doc_sources(
                     doc_sources=doc_sources,
-                    repo_root=Path(request.config.rootdir).resolve(),
                     llama_stack_client=unprivileged_llama_stack_client,
                     vector_store=vector_store,
                     vector_io_provider=vector_io_provider,

tests/llama_stack/utils.py

@@ -20,6 +20,7 @@
 )
 from utilities.exceptions import UnexpectedResourceCountError
 from utilities.opendatahub_logger import get_logger
+from utilities.path_utils import resolve_repo_path
 from utilities.resources.llama_stack_distribution import LlamaStackDistribution
 
 LOGGER = get_logger(name=__name__)
@@ -279,38 +280,32 @@ def vector_store_create_file_from_path(
 
 
 def vector_store_upload_doc_sources(
-    doc_sources: Any,
-    repo_root: Path,
+    doc_sources: list[str],
     llama_stack_client: LlamaStackClient,
     vector_store: Any,
     vector_io_provider: str,
 ) -> None:
     """Upload parametrized document sources (URLs and repo-local paths) to a vector store.
 
-    Resolves each local path under ``repo_root`` and re-resolves directory entries to avoid
-    symlink escape outside the repository.
+    Resolves each local path via ``resolve_repo_path`` and re-resolves directory entries
+    to avoid symlink escape outside the repository.
 
     Args:
         doc_sources: List of URL or path strings (repo-relative or absolute under repo root).
-        repo_root: Resolved repository root; local paths must resolve under this directory.
         llama_stack_client: Client used for file and vector store APIs.
         vector_store: Target vector store (must expose ``id``).
         vector_io_provider: Provider id for log context only.
 
     Raises:
-        TypeError: If ``doc_sources`` is not a list.
-        ValueError: If a local path resolves outside ``repo_root``.
+        ValueError: If a local path resolves outside the repo root.
         FileNotFoundError: If a file or non-empty directory source is missing.
     """
-    if not isinstance(doc_sources, list):
-        raise TypeError(f"doc_sources must be a list[str], got {type(doc_sources).__name__}")
     LOGGER.info(
         "Uploading doc_sources to vector_store (provider_id=%s, id=%s): %s",
         vector_io_provider,
         vector_store.id,
         doc_sources,
     )
-    repo_root_resolved = repo_root.resolve()
     for source in doc_sources:
         if source.startswith(("http://", "https://")):
             vector_store_create_file_from_url(
@@ -319,25 +314,14 @@ def vector_store_upload_doc_sources(
                 vector_store=vector_store,
             )
             continue
-        raw_path = Path(source)  # noqa: FCN001
-        resolved_source = raw_path.resolve() if raw_path.is_absolute() else (repo_root_resolved / raw_path).resolve()
-        if not resolved_source.is_relative_to(repo_root_resolved):
-            raise ValueError(
-                f"doc_sources path must be under repo root ({repo_root_resolved}): {source!r}",
-            )
-        source_path = resolved_source
+        source_path = resolve_repo_path(source=source)
 
         if source_path.is_dir():
             files = sorted(source_path.iterdir())
             if not files:
                 raise FileNotFoundError(f"No files found in directory: {source_path}")
             for file_path in files:
-                file_path_resolved = file_path.resolve(strict=True)
-                if not file_path_resolved.is_relative_to(repo_root_resolved):
-                    raise ValueError(
-                        f"doc_sources directory entry must resolve under repo root "
-                        f"({repo_root_resolved}): {file_path!r} -> {file_path_resolved!r}",
-                    )
+                file_path_resolved = resolve_repo_path(source=file_path)
                 if not file_path_resolved.is_file():
                     continue
                 vector_store_create_file_from_path(

tests/model_serving/model_server/utils.py

@@ -88,7 +88,7 @@ def verify_inference_response(
 
         elif (
             isinstance(inference_service, InferenceGraph)
-            and inference.deployment_mode == KServeDeploymentType.RAW_DEPLOYMENT
+            and inference.deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES
         ):
             assert "x-forbidden-reason: Access to the InferenceGraph is not allowed" in res["output"]
 

utilities/constants.py

@@ -7,8 +7,11 @@
 class KServeDeploymentType:
     SERVERLESS: str = "Serverless"
     RAW_DEPLOYMENT: str = "RawDeployment"
+    STANDARD: str = "Standard"
     MODEL_MESH: str = "ModelMesh"
 
+    RAW_DEPLOYMENT_MODES: tuple[str, ...] = (RAW_DEPLOYMENT, STANDARD)
+
 
 class ModelFormat:
     CAIKIT: str = "caikit"

utilities/general.py

@@ -173,9 +173,9 @@ def create_isvc_label_selector_str(isvc: InferenceService, resource_type: str, r
 
     """
     deployment_mode = isvc.instance.metadata.annotations.get(Annotations.KserveIo.DEPLOYMENT_MODE)
-    if deployment_mode in (
-        KServeDeploymentType.SERVERLESS,
-        KServeDeploymentType.RAW_DEPLOYMENT,
+    if (
+        deployment_mode == KServeDeploymentType.SERVERLESS
+        or deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES
     ):
         return f"{isvc.ApiGroup.SERVING_KSERVE_IO}/inferenceservice={isvc.name}"
 

utilities/inference_utils.py

@@ -118,7 +118,7 @@ def is_service_exposed(self) -> bool:
         """
         labels = self.inference_service.labels
 
-        if self.deployment_mode == KServeDeploymentType.RAW_DEPLOYMENT:
+        if self.deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES:
             if isinstance(self.inference_service, InferenceGraph):
                 # For InferenceGraph, the logic is similar as in Serverless. Only the label is different.
                 return not (labels and labels.get(Labels.Kserve.NETWORKING_KSERVE_IO) == "cluster-local")
@@ -310,7 +310,7 @@ def generate_command(
 
         elif self.protocol == "grpc":
             cmd_exec = "grpcurl -connect-timeout 10 "
-            if self.deployment_mode == KServeDeploymentType.RAW_DEPLOYMENT:
+            if self.deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES:
                 cmd_exec += " --plaintext "
 
         else:
@@ -536,10 +536,9 @@ def get_target_port(self, svc: Service) -> int:
                 and port.protocol.lower() == svc_protocol.lower()
                 and port.name == self.protocol
             ) or (
-                self.deployment_mode
-                in (
-                    KServeDeploymentType.RAW_DEPLOYMENT,
-                    KServeDeploymentType.SERVERLESS,
+                (
+                    self.deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES
+                    or self.deployment_mode == KServeDeploymentType.SERVERLESS
                 )
                 and port.protocol.lower() == svc_protocol.lower()
             ):
@@ -679,15 +678,18 @@ def create_isvc(
         _annotations = {Annotations.KserveIo.DEPLOYMENT_MODE: deployment_mode}
 
     # model mesh auth is set in ServingRuntime
-    if enable_auth and deployment_mode in {KServeDeploymentType.SERVERLESS, KServeDeploymentType.RAW_DEPLOYMENT}:
+    if enable_auth and (
+        deployment_mode == KServeDeploymentType.SERVERLESS
+        or deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES
+    ):
         _annotations[Annotations.KserveAuth.SECURITY] = "true"
 
     # default to True if deployment_mode is Serverless (default behavior of Serverless) if was not provided by the user
     # model mesh external route is set in ServingRuntime
     if external_route is None and deployment_mode == KServeDeploymentType.SERVERLESS:
         external_route = True
 
-    if external_route and deployment_mode == KServeDeploymentType.RAW_DEPLOYMENT:
+    if external_route and deployment_mode in KServeDeploymentType.RAW_DEPLOYMENT_MODES:
         labels[Labels.Kserve.NETWORKING_KSERVE_IO] = Labels.Kserve.EXPOSED
 
     if deployment_mode == KServeDeploymentType.SERVERLESS and external_route is False:

utilities/infra.py

@@ -272,7 +272,7 @@ def wait_for_inference_deployment_replicas(
             # to be set in deployment spec by HPA
             if (
                 isvc.instance.metadata.annotations.get("serving.kserve.io/deploymentMode")
-                == KServeDeploymentType.RAW_DEPLOYMENT
+                in KServeDeploymentType.RAW_DEPLOYMENT_MODES
             ):
                 wait_for_replicas_in_deployment(
                     deployment=deployment,