datarobot-genai 0.2.12__py3-none-any.whl → 0.2.14__py3-none-any.whl

datarobot_genai/nat/agent.py

@@ -21,7 +21,6 @@ from nat.data_models.api_server import ChatRequest
 from nat.data_models.api_server import ChatResponse
 from nat.data_models.intermediate_step import IntermediateStep
 from nat.data_models.intermediate_step import IntermediateStepType
-from nat.runtime.loader import load_workflow
 from nat.utils.type_utils import StrPath
 from openai.types.chat import CompletionCreateParams
 from ragas import MultiTurnSample
@@ -34,6 +33,8 @@ from datarobot_genai.core.agents.base import InvokeReturn
 from datarobot_genai.core.agents.base import UsageMetrics
 from datarobot_genai.core.agents.base import extract_user_prompt_content
 from datarobot_genai.core.agents.base import is_streaming
+from datarobot_genai.core.mcp.common import MCPConfig
+from datarobot_genai.nat.helpers import load_workflow
 
 logger = logging.getLogger(__name__)
 
@@ -166,17 +167,24 @@ class NatAgent(BaseAgent[None]):
         # Print commands may need flush=True to ensure they are displayed in real-time.
         print("Running agent with user prompt:", chat_request.messages[0].content, flush=True)
 
+        mcp_config = MCPConfig(
+            authorization_context=self.authorization_context,
+            forwarded_headers=self.forwarded_headers,
+        )
+        server_config = mcp_config.server_config
+        headers = server_config["headers"] if server_config else None
+
         if is_streaming(completion_create_params):
 
             async def stream_generator() -> AsyncGenerator[
                 tuple[str, MultiTurnSample | None, UsageMetrics], None
             ]:
-                usage_metrics: UsageMetrics = {
+                default_usage_metrics: UsageMetrics = {
                     "completion_tokens": 0,
                     "prompt_tokens": 0,
                     "total_tokens": 0,
                 }
-                async with load_workflow(self.workflow_path) as workflow:
+                async with load_workflow(self.workflow_path, headers=headers) as workflow:
                     async with workflow.run(chat_request) as runner:
                         intermediate_future = pull_intermediate_structured()
                         async for result in runner.result_stream():
@@ -188,7 +196,7 @@ class NatAgent(BaseAgent[None]):
                             yield (
                                 result_text,
                                 None,
-                                usage_metrics,
+                                default_usage_metrics,
                             )
 
                         steps = await intermediate_future
@@ -197,6 +205,11 @@ class NatAgent(BaseAgent[None]):
                             for step in steps
                             if step.event_type == IntermediateStepType.LLM_END
                         ]
+                        usage_metrics: UsageMetrics = {
+                            "completion_tokens": 0,
+                            "prompt_tokens": 0,
+                            "total_tokens": 0,
+                        }
                         for step in llm_end_steps:
                             if step.usage_info:
                                 token_usage = step.usage_info.token_usage
@@ -210,7 +223,7 @@ class NatAgent(BaseAgent[None]):
             return stream_generator()
 
         # Create and invoke the NAT (Nemo Agent Toolkit) Agentic Workflow with the inputs
-        result, steps = await self.run_nat_workflow(self.workflow_path, chat_request)
+        result, steps = await self.run_nat_workflow(self.workflow_path, chat_request, headers)
 
         llm_end_steps = [step for step in steps if step.event_type == IntermediateStepType.LLM_END]
         usage_metrics: UsageMetrics = {
@@ -234,7 +247,7 @@ class NatAgent(BaseAgent[None]):
         return result_text, pipeline_interactions, usage_metrics
 
     async def run_nat_workflow(
-        self, workflow_path: StrPath, chat_request: ChatRequest
+        self, workflow_path: StrPath, chat_request: ChatRequest, headers: dict[str, str] | None
     ) -> tuple[ChatResponse | str, list[IntermediateStep]]:
         """Run the NAT workflow with the provided config file and input string.
 
@@ -247,7 +260,7 @@ class NatAgent(BaseAgent[None]):
             ChatResponse | str: The result from the NAT workflow
             list[IntermediateStep]: The list of intermediate steps
         """
-        async with load_workflow(workflow_path) as workflow:
+        async with load_workflow(workflow_path, headers=headers) as workflow:
             async with workflow.run(chat_request) as runner:
                 intermediate_future = pull_intermediate_structured()
                 runner_outputs = await runner.result()

datarobot_genai/nat/datarobot_llm_clients.py

@@ -14,6 +14,7 @@
 
 from collections.abc import AsyncGenerator
 from typing import Any
+from typing import TypeVar
 
 from crewai import LLM
 from langchain_openai import ChatOpenAI
@@ -22,12 +23,32 @@ from llama_index.llms.litellm import LiteLLM
 from nat.builder.builder import Builder
 from nat.builder.framework_enum import LLMFrameworkEnum
 from nat.cli.register_workflow import register_llm_client
+from nat.data_models.llm import LLMBaseConfig
+from nat.data_models.retry_mixin import RetryMixin
+from nat.plugins.langchain.llm import (
+    _patch_llm_based_on_config as langchain_patch_llm_based_on_config,
+)
+from nat.utils.exception_handlers.automatic_retries import patch_with_retry
 
 from ..nat.datarobot_llm_providers import DataRobotLLMComponentModelConfig
 from ..nat.datarobot_llm_providers import DataRobotLLMDeploymentModelConfig
 from ..nat.datarobot_llm_providers import DataRobotLLMGatewayModelConfig
 from ..nat.datarobot_llm_providers import DataRobotNIMModelConfig
 
+ModelType = TypeVar("ModelType")
+
+
+def _patch_llm_based_on_config(client: ModelType, llm_config: LLMBaseConfig) -> ModelType:
+    if isinstance(llm_config, RetryMixin):
+        client = patch_with_retry(
+            client,
+            retries=llm_config.num_retries,
+            retry_codes=llm_config.retry_on_status_codes,
+            retry_on_messages=llm_config.retry_on_errors,
+        )
+
+    return client
+
 
 class DataRobotChatOpenAI(ChatOpenAI):
     def _get_request_payload(
@@ -77,7 +98,8 @@ async def datarobot_llm_gateway_langchain(
     config["base_url"] = config["base_url"] + "/genai/llmgw"
     config["stream_options"] = {"include_usage": True}
     config["model"] = config["model"].removeprefix("datarobot/")
-    yield DataRobotChatOpenAI(**config)
+    client = DataRobotChatOpenAI(**config)
+    yield langchain_patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -90,7 +112,8 @@ async def datarobot_llm_gateway_crewai(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["base_url"] = config["base_url"].removesuffix("/api/v2")
-    yield LLM(**config)
+    client = LLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -103,7 +126,8 @@ async def datarobot_llm_gateway_llamaindex(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["api_base"] = config.pop("base_url").removesuffix("/api/v2")
-    yield DataRobotLiteLLM(**config)
+    client = DataRobotLiteLLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -119,7 +143,8 @@ async def datarobot_llm_deployment_langchain(
     )
     config["stream_options"] = {"include_usage": True}
     config["model"] = config["model"].removeprefix("datarobot/")
-    yield DataRobotChatOpenAI(**config)
+    client = DataRobotChatOpenAI(**config)
+    yield langchain_patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -136,7 +161,8 @@ async def datarobot_llm_deployment_crewai(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["api_base"] = config.pop("base_url") + "/chat/completions"
-    yield LLM(**config)
+    client = LLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -153,7 +179,8 @@ async def datarobot_llm_deployment_llamaindex(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["api_base"] = config.pop("base_url") + "/chat/completions"
-    yield DataRobotLiteLLM(**config)
+    client = DataRobotLiteLLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(config_type=DataRobotNIMModelConfig, wrapper_type=LLMFrameworkEnum.LANGCHAIN)
@@ -167,7 +194,8 @@ async def datarobot_nim_langchain(
     )
     config["stream_options"] = {"include_usage": True}
     config["model"] = config["model"].removeprefix("datarobot/")
-    yield DataRobotChatOpenAI(**config)
+    client = DataRobotChatOpenAI(**config)
+    yield langchain_patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(config_type=DataRobotNIMModelConfig, wrapper_type=LLMFrameworkEnum.CREWAI)
@@ -182,7 +210,8 @@ async def datarobot_nim_crewai(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["api_base"] = config.pop("base_url") + "/chat/completions"
-    yield LLM(**config)
+    client = LLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(config_type=DataRobotNIMModelConfig, wrapper_type=LLMFrameworkEnum.LLAMA_INDEX)
@@ -197,7 +226,8 @@ async def datarobot_nim_llamaindex(
     if not config["model"].startswith("datarobot/"):
         config["model"] = "datarobot/" + config["model"]
     config["api_base"] = config.pop("base_url") + "/chat/completions"
-    yield DataRobotLiteLLM(**config)
+    client = DataRobotLiteLLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -212,7 +242,8 @@ async def datarobot_llm_component_langchain(
     config["stream_options"] = {"include_usage": True}
     config["model"] = config["model"].removeprefix("datarobot/")
     config.pop("use_datarobot_llm_gateway")
-    yield DataRobotChatOpenAI(**config)
+    client = DataRobotChatOpenAI(**config)
+    yield langchain_patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -229,7 +260,8 @@ async def datarobot_llm_component_crewai(
     else:
         config["api_base"] = config.pop("base_url") + "/chat/completions"
     config.pop("use_datarobot_llm_gateway")
-    yield LLM(**config)
+    client = LLM(**config)
+    yield _patch_llm_based_on_config(client, config)
 
 
 @register_llm_client(
@@ -246,4 +278,5 @@ async def datarobot_llm_component_llamaindex(
     else:
         config["api_base"] = config.pop("base_url") + "/chat/completions"
     config.pop("use_datarobot_llm_gateway")
-    yield DataRobotLiteLLM(**config)
+    client = DataRobotLiteLLM(**config)
+    yield _patch_llm_based_on_config(client, config)

datarobot_genai/nat/helpers.py

@@ -0,0 +1,87 @@
+# Copyright 2025 DataRobot, Inc. and its affiliates.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from collections.abc import AsyncGenerator
+from contextlib import asynccontextmanager
+
+from nat.builder.workflow import Workflow
+from nat.builder.workflow_builder import WorkflowBuilder
+from nat.data_models.config import Config
+from nat.runtime.loader import PluginTypes
+from nat.runtime.loader import discover_and_register_plugins
+from nat.runtime.session import SessionManager
+from nat.utils.data_models.schema_validator import validate_schema
+from nat.utils.io.yaml_tools import yaml_load
+from nat.utils.type_utils import StrPath
+
+
+def load_config(config_file: StrPath, headers: dict[str, str] | None = None) -> Config:
+    """
+    Load a NAT configuration file with injected headers. It ensures that all plugins are
+    loaded and then validates the configuration file against the Config schema.
+
+    Parameters
+    ----------
+    config_file : StrPath
+        The path to the configuration file
+
+    Returns
+    -------
+    Config
+        The validated Config object
+    """
+    # Ensure all of the plugins are loaded
+    discover_and_register_plugins(PluginTypes.CONFIG_OBJECT)
+
+    config_yaml = yaml_load(config_file)
+
+    add_headers_to_datarobot_mcp_auth(config_yaml, headers)
+
+    # Validate configuration adheres to NAT schemas
+    validated_nat_config = validate_schema(config_yaml, Config)
+
+    return validated_nat_config
+
+
+def add_headers_to_datarobot_mcp_auth(config_yaml: dict, headers: dict[str, str] | None) -> None:
+    if headers:
+        if authentication := config_yaml.get("authentication"):
+            for auth_name in authentication:
+                auth_config = authentication[auth_name]
+                if auth_config.get("_type") == "datarobot_mcp_auth":
+                    auth_config["headers"] = headers
+
+
+@asynccontextmanager
+async def load_workflow(
+    config_file: StrPath, max_concurrency: int = -1, headers: dict[str, str] | None = None
+) -> AsyncGenerator[Workflow, None]:
+    """
+    Load the NAT configuration file and create a Runner object. This is the primary entry point for
+    running NAT workflows with injected headers.
+
+    Parameters
+    ----------
+    config_file : StrPath
+        The path to the configuration file
+    max_concurrency : int, optional
+        The maximum number of parallel workflow invocations to support. Specifying 0 or -1 will
+        allow an unlimited count, by default -1
+    """
+    # Load the config object
+    config = load_config(config_file, headers=headers)
+
+    # Must yield the workflow function otherwise it cleans up
+    async with WorkflowBuilder.from_config(config=config) as workflow:
+        yield SessionManager(await workflow.build(), max_concurrency=max_concurrency)

{datarobot_genai-0.2.12.dist-info → datarobot_genai-0.2.14.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: datarobot-genai
-Version: 0.2.12
+Version: 0.2.14
 Summary: Generic helpers for GenAI
 Project-URL: Homepage, https://github.com/datarobot-oss/datarobot-genai
 Author: DataRobot, Inc.

{datarobot_genai-0.2.12.dist-info → datarobot_genai-0.2.14.dist-info}/RECORD

@@ -100,14 +100,15 @@ datarobot_genai/llama_index/agent.py,sha256=V6ZsD9GcBDJS-RJo1tJtIHhyW69_78gM6_fO
 datarobot_genai/llama_index/base.py,sha256=ovcQQtC-djD_hcLrWdn93jg23AmD6NBEj7xtw4a6K6c,14481
 datarobot_genai/llama_index/mcp.py,sha256=leXqF1C4zhuYEKFwNEfZHY4dsUuGZk3W7KArY-zxVL8,2645
 datarobot_genai/nat/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-datarobot_genai/nat/agent.py,sha256=jDeIS9f-8vGbeLy5gQkSjeuHINx5Fh_4BvXYERsgIIk,10516
+datarobot_genai/nat/agent.py,sha256=DuGrgqt1FzvAE-cRH_P3LTFUlwuClvbVurdwA-RsbuY,11177
 datarobot_genai/nat/datarobot_auth_provider.py,sha256=Z4NSsrHxK8hUeiqtK_lryHsUuZC74ziNo_FHbsZgtiM,4230
-datarobot_genai/nat/datarobot_llm_clients.py,sha256=STzAZ4OF8U-Y_cUTywxmKBGVotwsnbGP6vTojnu6q0g,9921
+datarobot_genai/nat/datarobot_llm_clients.py,sha256=Yu208Ed_p_4P3HdpuM7fYnKcXtimORHpKlWVPyijpU8,11356
 datarobot_genai/nat/datarobot_llm_providers.py,sha256=aDoQcTeGI-odqydPXEX9OGGNFbzAtpqzTvHHEkmJuEQ,4963
 datarobot_genai/nat/datarobot_mcp_client.py,sha256=35FzilxNp4VqwBYI0NsOc91-xZm1C-AzWqrOdDy962A,9612
-datarobot_genai-0.2.12.dist-info/METADATA,sha256=5DnB86Cp4uSS6x5ZjJqblA40CusRaa9V5Jw0kiGSVig,6301
-datarobot_genai-0.2.12.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-datarobot_genai-0.2.12.dist-info/entry_points.txt,sha256=jEW3WxDZ8XIK9-ISmTyt5DbmBb047rFlzQuhY09rGrM,284
-datarobot_genai-0.2.12.dist-info/licenses/AUTHORS,sha256=isJGUXdjq1U7XZ_B_9AH8Qf0u4eX0XyQifJZ_Sxm4sA,80
-datarobot_genai-0.2.12.dist-info/licenses/LICENSE,sha256=U2_VkLIktQoa60Nf6Tbt7E4RMlfhFSjWjcJJfVC-YCE,11341
-datarobot_genai-0.2.12.dist-info/RECORD,,
+datarobot_genai/nat/helpers.py,sha256=Q7E3ADZdtFfS8E6OQPyw2wgA6laQ58N3bhLj5CBWwJs,3265
+datarobot_genai-0.2.14.dist-info/METADATA,sha256=MWpePh1Ditr0AsXs5dR8XuQcur7DynbiFvjkBWY9NvM,6301
+datarobot_genai-0.2.14.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+datarobot_genai-0.2.14.dist-info/entry_points.txt,sha256=jEW3WxDZ8XIK9-ISmTyt5DbmBb047rFlzQuhY09rGrM,284
+datarobot_genai-0.2.14.dist-info/licenses/AUTHORS,sha256=isJGUXdjq1U7XZ_B_9AH8Qf0u4eX0XyQifJZ_Sxm4sA,80
+datarobot_genai-0.2.14.dist-info/licenses/LICENSE,sha256=U2_VkLIktQoa60Nf6Tbt7E4RMlfhFSjWjcJJfVC-YCE,11341
+datarobot_genai-0.2.14.dist-info/RECORD,,