fix(llms): send UiPath token via Authorization Bearer header in UiPathOpenAI

AzureOpenAI emits the api_key argument as the Azure-style `api-key` header,
which AgentHub's LLMProxy [Authorize] pipeline (JwtBearer/S2S schemes) does
not read, resulting in 401 with no Authorization header on the request.
Bedrock and Vertex wrappers already attach `Authorization: Bearer <token>`
explicitly; align UiPathOpenAI by setting it in default_headers and passing
a placeholder api_key to satisfy the AzureOpenAI constructor.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: ionmincu

packages/uipath-llamaindex/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "uipath-llamaindex"
-version = "0.5.12"
+version = "0.5.13"
 description = "Python SDK that enables developers to build and deploy LlamaIndex agents to the UiPath Cloud Platform"
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.11"

packages/uipath-llamaindex/src/uipath_llamaindex/llms/_openai.py

@@ -50,7 +50,9 @@ def __init__(
         api_version: str = "2024-10-21",
         **kwargs: Any,
     ):
+        token = os.environ.get("UIPATH_ACCESS_TOKEN")
         default_headers_dict = {
+            "Authorization": f"Bearer {token}",
             "X-UiPath-LlmGateway-ApiFlavor": "auto",
             "X-UiPath-LlmGateway-RequestingProduct": "uipath-python-sdk",
             "X-UiPath-LlmGateway-RequestingFeature": "llama-index-agent",
@@ -74,7 +76,7 @@ def __init__(
             "model": model_value,
             "deployment_name": model_value,
             "azure_endpoint": f"{base_url}/{vendor_endpoint}",
-            "api_key": os.environ.get("UIPATH_ACCESS_TOKEN"),
+            "api_key": "uipath-gateway",
             "api_version": api_version,
             "is_chat_model": True,
             "default_headers": default_headers_dict,