From baea83cb3d7c26d8607da803175d65b4d8c2370f Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 08:41:59 -0700
Subject: [PATCH 01/12] fix: use agent's configured model for eval simulations
 instead of gpt-4.1-mini default

The LLM mocker and input mocker for eval simulations were falling back to
the chat_completions default model (gpt-4.1-mini) when no simulation-specific
model was configured. This caused 417 errors in EU regions where gpt-4.1-mini
isn't provisioned for the agentsplayground product.

Now threads the agent's own model through MockingContext so simulations use
the same model the agent is configured with. The model is read from agent.json
(server/debug path) or from the runtime schema (CLI eval path).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/_cli/cli_eval.py    |  6 +++++-
 .../src/uipath/eval/mocks/_input_mocker.py     |  5 +++++
 .../src/uipath/eval/mocks/_llm_mocker.py       |  4 ++++
 .../src/uipath/eval/mocks/_mock_runtime.py     | 18 ++++++++++++++++++
 .../uipath/src/uipath/eval/mocks/_types.py     |  1 +
 .../uipath/src/uipath/eval/runtime/context.py  |  1 +
 .../uipath/src/uipath/eval/runtime/runtime.py  |  2 ++
 7 files changed, 36 insertions(+), 1 deletion(-)

diff --git a/packages/uipath/src/uipath/_cli/cli_eval.py b/packages/uipath/src/uipath/_cli/cli_eval.py
index ef1edb200..e82c33fa9 100644
--- a/packages/uipath/src/uipath/_cli/cli_eval.py
+++ b/packages/uipath/src/uipath/_cli/cli_eval.py
@@ -428,10 +428,14 @@ async def execute_eval():
 
                         eval_context.runtime_schema = await runtime.get_schema()
 
+                        eval_context.agent_model = _get_agent_model(
+                            eval_context.runtime_schema
+                        )
+
                         eval_context.evaluators = await EvalHelpers.load_evaluators(
                             resolved_eval_set_path,
                             eval_context.evaluation_set,
-                            _get_agent_model(eval_context.runtime_schema),
+                            eval_context.agent_model,
                         )
 
                         # Runtime is not required anymore.
diff --git a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
index f1d253ba8..89a57a091 100644
--- a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
@@ -65,6 +65,7 @@ async def generate_llm_input(
     input_schema: dict[str, Any],
     expected_behavior: str,
     expected_output: dict[str, Any],
+    agent_model: str | None = None,
 ) -> dict[str, Any]:
     """Generate synthetic input using an LLM based on the evaluation context."""
     # Set custom span attributes to match agents repo pattern
@@ -117,6 +118,10 @@ async def generate_llm_input(
             else {}
         )
 
+        # Use the agent's configured model when no simulation-specific model is set
+        if "model" not in completion_kwargs and agent_model:
+            completion_kwargs["model"] = agent_model
+
         if cache_manager is not None:
             cache_key_data = {
                 "response_format": response_format,
diff --git a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
index 194aa6c09..1c5ac5009 100644
--- a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
@@ -182,6 +182,10 @@ async def response(
                     else {}
                 )
 
+                # Use the agent's configured model when no simulation-specific model is set
+                if "model" not in completion_kwargs and self.context.agent_model:
+                    completion_kwargs["model"] = self.context.agent_model
+
                 formatted_prompt = PROMPT.format(**prompt_generation_args)
 
                 cache_key_data = {
diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index 512d8d6ee..5ba404650 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -34,6 +34,20 @@
 logger = logging.getLogger(__name__)
 
 
+def _read_agent_model() -> str | None:
+    """Read the agent's configured model from agent.json in current directory."""
+    agent_path = Path.cwd() / "agent.json"
+    if not agent_path.exists():
+        return None
+    try:
+        with open(agent_path, "r", encoding="utf-8") as f:
+            agent_data = json.load(f)
+        return agent_data.get("settings", {}).get("model")
+    except Exception as e:
+        logger.warning(f"Failed to read agent model from agent.json: {e}")
+        return None
+
+
 def load_simulation_config() -> MockingContext | None:
     """Load simulation.json from current directory and convert to MockingContext.
 
@@ -70,11 +84,15 @@ def load_simulation_config() -> MockingContext | None:
             tools_to_simulate=tools_to_simulate,
         )
 
+        # Read agent model so simulations use the same model as the agent
+        agent_model = _read_agent_model()
+
         # Create MockingContext for debugging
         mocking_context = MockingContext(
             strategy=mocking_strategy,
             name="debug-simulation",
             inputs={},
+            agent_model=agent_model,
         )
 
         logger.info(f"Loaded simulation config for {len(tools_to_simulate)} tool(s)")
diff --git a/packages/uipath/src/uipath/eval/mocks/_types.py b/packages/uipath/src/uipath/eval/mocks/_types.py
index 827569879..043613c44 100644
--- a/packages/uipath/src/uipath/eval/mocks/_types.py
+++ b/packages/uipath/src/uipath/eval/mocks/_types.py
@@ -127,6 +127,7 @@ class MockingContext(BaseModel):
     strategy: MockingStrategy | None
     inputs: dict[str, Any] = Field(default_factory=lambda: {})
     name: str = Field(default="debug")
+    agent_model: str | None = Field(default=None)
 
 
 class ExampleCall(BaseModel):
diff --git a/packages/uipath/src/uipath/eval/runtime/context.py b/packages/uipath/src/uipath/eval/runtime/context.py
index b8224718c..81263c54d 100644
--- a/packages/uipath/src/uipath/eval/runtime/context.py
+++ b/packages/uipath/src/uipath/eval/runtime/context.py
@@ -27,3 +27,4 @@ class UiPathEvalContext:
     input_overrides: dict[str, Any] | None = None
     resume: bool = False
     job_id: str | None = None
+    agent_model: str | None = None
diff --git a/packages/uipath/src/uipath/eval/runtime/runtime.py b/packages/uipath/src/uipath/eval/runtime/runtime.py
index 1c32b9efe..ffba867b4 100644
--- a/packages/uipath/src/uipath/eval/runtime/runtime.py
+++ b/packages/uipath/src/uipath/eval/runtime/runtime.py
@@ -534,6 +534,7 @@ async def _execute_eval(
                             strategy=eval_item.mocking_strategy,
                             name=eval_item.name,
                             inputs=eval_item.inputs,
+                            agent_model=self.context.agent_model,
                         ),
                         eval_set_run_id=self.context.eval_set_run_id,
                     )
@@ -816,6 +817,7 @@ async def _generate_input_for_eval(
             (await self.get_schema()).input,
             expected_behavior=eval_item.expected_agent_behavior or "",
             expected_output=expected_output,
+            agent_model=self.context.agent_model,
         )
         updated_eval_item = eval_item.model_copy(update={"inputs": generated_input})
         return updated_eval_item

From b336544c1349f3f0a7c28b0e247559a8958557cd Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 08:49:53 -0700
Subject: [PATCH 02/12] refactor: pass agent_model from runtime schema instead
 of reading agent.json
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Remove _read_agent_model() file-reading hack. The model is already
available from the runtime schema — pass it through UiPathMockRuntime
from the callers (cli_debug.py and eval runtime) who have access to it.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/_cli/cli_debug.py  |  7 +++++
 .../src/uipath/eval/mocks/_mock_runtime.py    | 29 +++++++------------
 .../uipath/src/uipath/eval/runtime/runtime.py |  1 +
 3 files changed, 18 insertions(+), 19 deletions(-)

diff --git a/packages/uipath/src/uipath/_cli/cli_debug.py b/packages/uipath/src/uipath/_cli/cli_debug.py
index 44415dd26..eb2ebc160 100644
--- a/packages/uipath/src/uipath/_cli/cli_debug.py
+++ b/packages/uipath/src/uipath/_cli/cli_debug.py
@@ -163,8 +163,15 @@ async def execute_debug_runtime():
                                 trigger_poll_interval=trigger_poll_interval,
                             )
 
+                            # Get agent model from runtime schema for simulations
+                            schema = await runtime.get_schema()
+                            agent_model = None
+                            if schema.metadata and "settings" in schema.metadata:
+                                agent_model = schema.metadata["settings"].get("model")
+
                             mock_runtime = UiPathMockRuntime(
                                 delegate=debug_runtime,
+                                agent_model=agent_model,
                             )
 
                             try:
diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index 5ba404650..63814b13d 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -34,21 +34,7 @@
 logger = logging.getLogger(__name__)
 
 
-def _read_agent_model() -> str | None:
-    """Read the agent's configured model from agent.json in current directory."""
-    agent_path = Path.cwd() / "agent.json"
-    if not agent_path.exists():
-        return None
-    try:
-        with open(agent_path, "r", encoding="utf-8") as f:
-            agent_data = json.load(f)
-        return agent_data.get("settings", {}).get("model")
-    except Exception as e:
-        logger.warning(f"Failed to read agent model from agent.json: {e}")
-        return None
-
-
-def load_simulation_config() -> MockingContext | None:
+def load_simulation_config(agent_model: str | None = None) -> MockingContext | None:
     """Load simulation.json from current directory and convert to MockingContext.
 
     Returns:
@@ -84,9 +70,6 @@ def load_simulation_config() -> MockingContext | None:
             tools_to_simulate=tools_to_simulate,
         )
 
-        # Read agent model so simulations use the same model as the agent
-        agent_model = _read_agent_model()
-
         # Create MockingContext for debugging
         mocking_context = MockingContext(
             strategy=mocking_strategy,
@@ -151,9 +134,17 @@ def __init__(
         span_collector: ExecutionSpanCollector | None = None,
         execution_id: str | None = None,
         eval_set_run_id: str | None = None,
+        agent_model: str | None = None,
     ):
         self.delegate = delegate
-        self._mocking_context = mocking_context or load_simulation_config()
+        self._mocking_context = mocking_context or load_simulation_config(
+            agent_model=agent_model
+        )
+        # If mocking_context was passed without agent_model, inject it
+        if self._mocking_context and not self._mocking_context.agent_model and agent_model:
+            self._mocking_context = self._mocking_context.model_copy(
+                update={"agent_model": agent_model}
+            )
         self._span_collector = span_collector or ExecutionSpanCollector()
         self._execution_id = execution_id or str(uuid.uuid4())
         self._eval_set_run_id = eval_set_run_id
diff --git a/packages/uipath/src/uipath/eval/runtime/runtime.py b/packages/uipath/src/uipath/eval/runtime/runtime.py
index ffba867b4..8b25b01eb 100644
--- a/packages/uipath/src/uipath/eval/runtime/runtime.py
+++ b/packages/uipath/src/uipath/eval/runtime/runtime.py
@@ -870,6 +870,7 @@ async def execute_runtime(
                     span_collector=self.span_collector,
                     execution_id=execution_id,
                     eval_set_run_id=eval_set_run_id,
+                    agent_model=self.context.agent_model,
                 )
 
             execution_runtime = UiPathExecutionRuntime(

From cd5aeacd01babb2602cdbce52ca0bf294f12968f Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 09:08:16 -0700
Subject: [PATCH 03/12] fix: ruff format and update test mocks for get_schema
 call

- Fix ruff formatting in _mock_runtime.py
- Add get_schema AsyncMock to mock_runtime in debug simulation tests,
  since cli_debug.py now calls runtime.get_schema() to extract the
  agent model before creating UiPathMockRuntime

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/eval/mocks/_mock_runtime.py | 6 +++++-
 packages/uipath/tests/cli/test_debug_simulation.py     | 6 ++++++
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index 63814b13d..bf2dae649 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -141,7 +141,11 @@ def __init__(
             agent_model=agent_model
         )
         # If mocking_context was passed without agent_model, inject it
-        if self._mocking_context and not self._mocking_context.agent_model and agent_model:
+        if (
+            self._mocking_context
+            and not self._mocking_context.agent_model
+            and agent_model
+        ):
             self._mocking_context = self._mocking_context.model_copy(
                 update={"agent_model": agent_model}
             )
diff --git a/packages/uipath/tests/cli/test_debug_simulation.py b/packages/uipath/tests/cli/test_debug_simulation.py
index b2d795c79..d9266327b 100644
--- a/packages/uipath/tests/cli/test_debug_simulation.py
+++ b/packages/uipath/tests/cli/test_debug_simulation.py
@@ -241,6 +241,9 @@ def test_debug_always_wraps_with_mock_runtime(
                     ) as mock_factory_get:
                         mock_runtime = Mock()
                         mock_runtime.dispose = AsyncMock()
+                        mock_runtime.get_schema = AsyncMock(
+                            return_value=Mock(metadata=None)
+                        )
 
                         mock_factory = Mock()
                         mock_factory.new_runtime = AsyncMock(return_value=mock_runtime)
@@ -305,6 +308,9 @@ def test_debug_wraps_with_mock_runtime_on_error(
                     ) as mock_factory_get:
                         mock_runtime = Mock()
                         mock_runtime.dispose = AsyncMock()
+                        mock_runtime.get_schema = AsyncMock(
+                            return_value=Mock(metadata=None)
+                        )
 
                         mock_factory = Mock()
                         mock_factory.new_runtime = AsyncMock(return_value=mock_runtime)

From b093568d817b75c9b52285be6b554dfa0e957c7b Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 09:11:55 -0700
Subject: [PATCH 04/12] chore: bump uipath version to 2.10.47

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/pyproject.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/packages/uipath/pyproject.toml b/packages/uipath/pyproject.toml
index 7eb42ce59..25c704503 100644
--- a/packages/uipath/pyproject.toml
+++ b/packages/uipath/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "uipath"
-version = "2.10.46"
+version = "2.10.47"
 description = "Python SDK and CLI for UiPath Platform, enabling programmatic interaction with automation services, process management, and deployment tools."
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.11"

From 5eaea72b3e2979906473175622324eb67378dab2 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 09:19:10 -0700
Subject: [PATCH 05/12] fix: update uv.lock and add simulation model logging

- Regenerate uv.lock after version bump to 2.10.47
- Log the model being used when running simulation mocking

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/eval/mocks/_input_mocker.py | 6 ++++++
 packages/uipath/src/uipath/eval/mocks/_llm_mocker.py   | 5 +++++
 packages/uipath/uv.lock                                | 2 +-
 3 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
index 89a57a091..2bf6e2efb 100644
--- a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
@@ -1,6 +1,7 @@
 """LLM Input Mocker implementation."""
 
 import json
+import logging
 from datetime import datetime
 from typing import Any
 
@@ -17,6 +18,8 @@
     InputMockingStrategy,
 )
 
+logger = logging.getLogger(__name__)
+
 
 def get_input_mocking_prompt(
     input_schema: str,
@@ -122,6 +125,9 @@ async def generate_llm_input(
         if "model" not in completion_kwargs and agent_model:
             completion_kwargs["model"] = agent_model
 
+        simulation_model = completion_kwargs.get("model", "default")
+        logger.info(f"Simulating input generation using model: {simulation_model}")
+
         if cache_manager is not None:
             cache_key_data = {
                 "response_format": response_format,
diff --git a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
index 1c5ac5009..e9e797f1c 100644
--- a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
@@ -186,6 +186,11 @@ async def response(
                 if "model" not in completion_kwargs and self.context.agent_model:
                     completion_kwargs["model"] = self.context.agent_model
 
+                simulation_model = completion_kwargs.get("model", "default")
+                logger.info(
+                    f"Simulating tool '{function_name}' using model: {simulation_model}"
+                )
+
                 formatted_prompt = PROMPT.format(**prompt_generation_args)
 
                 cache_key_data = {
diff --git a/packages/uipath/uv.lock b/packages/uipath/uv.lock
index 3c06d2318..f956b3d82 100644
--- a/packages/uipath/uv.lock
+++ b/packages/uipath/uv.lock
@@ -2543,7 +2543,7 @@ wheels = [
 
 [[package]]
 name = "uipath"
-version = "2.10.46"
+version = "2.10.47"
 source = { editable = "." }
 dependencies = [
     { name = "applicationinsights" },

From 075b14a4c9f7b93c0389a3db5d4818ccb59060e3 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 09:53:12 -0700
Subject: [PATCH 06/12] fix: log actual default model name instead of 'default'

Show gpt-4.1-mini-2025-04-14 in logs when no agent model is set,
so it's clear which model the simulation falls back to.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/eval/mocks/_input_mocker.py | 5 ++++-
 packages/uipath/src/uipath/eval/mocks/_llm_mocker.py   | 6 ++++--
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
index 2bf6e2efb..dc9fb74b4 100644
--- a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
@@ -10,6 +10,7 @@
 from uipath.core.tracing import traced
 from uipath.platform import UiPath
 from uipath.platform.chat import UiPathLlmChatService
+from uipath.platform.chat._llm_gateway_service import ChatModels
 
 from .._execution_context import eval_set_run_id_context
 from ._mock_context import cache_manager_context
@@ -125,7 +126,9 @@ async def generate_llm_input(
         if "model" not in completion_kwargs and agent_model:
             completion_kwargs["model"] = agent_model
 
-        simulation_model = completion_kwargs.get("model", "default")
+        simulation_model = completion_kwargs.get(
+            "model", ChatModels.gpt_4_1_mini_2025_04_14
+        )
         logger.info(f"Simulating input generation using model: {simulation_model}")
 
         if cache_manager is not None:
diff --git a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
index e9e797f1c..83e3498fb 100644
--- a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
@@ -10,7 +10,7 @@
 from uipath.core.tracing import traced
 from uipath.platform import UiPath
 from uipath.platform.chat import UiPathLlmChatService
-from uipath.platform.chat._llm_gateway_service import _cleanup_schema
+from uipath.platform.chat._llm_gateway_service import ChatModels, _cleanup_schema
 
 from .._execution_context import (
     eval_set_run_id_context,
@@ -186,7 +186,9 @@ async def response(
                 if "model" not in completion_kwargs and self.context.agent_model:
                     completion_kwargs["model"] = self.context.agent_model
 
-                simulation_model = completion_kwargs.get("model", "default")
+                simulation_model = completion_kwargs.get(
+                    "model", ChatModels.gpt_4_1_mini_2025_04_14
+                )
                 logger.info(
                     f"Simulating tool '{function_name}' using model: {simulation_model}"
                 )

From 77d456582bd052c232fcdab22c02095af34e0b84 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 10:05:02 -0700
Subject: [PATCH 07/12] refactor: set model on strategy instead of threading
 agent_model through context

Per code review feedback: the cleanest approach is to set the correct
model on LLMMockingStrategy/InputMockingStrategy before they reach
the mocker, rather than adding agent_model to MockingContext.

- runtime.py: set strategy.model from agent_model when strategy.model is None
- _mock_runtime.py: pass agent_model to ModelSettings on strategy in load_simulation_config
- Revert agent_model from MockingContext, LLMMocker fallback, and generate_llm_input param
- Keep logging that shows the actual model being used

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 .../src/uipath/eval/mocks/_input_mocker.py    |  5 ----
 .../src/uipath/eval/mocks/_llm_mocker.py      |  4 ---
 .../src/uipath/eval/mocks/_mock_runtime.py    | 14 ++-------
 .../uipath/src/uipath/eval/mocks/_types.py    |  1 -
 .../uipath/src/uipath/eval/runtime/runtime.py | 30 +++++++++++++++----
 5 files changed, 27 insertions(+), 27 deletions(-)

diff --git a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
index dc9fb74b4..57a727ec1 100644
--- a/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_input_mocker.py
@@ -69,7 +69,6 @@ async def generate_llm_input(
     input_schema: dict[str, Any],
     expected_behavior: str,
     expected_output: dict[str, Any],
-    agent_model: str | None = None,
 ) -> dict[str, Any]:
     """Generate synthetic input using an LLM based on the evaluation context."""
     # Set custom span attributes to match agents repo pattern
@@ -122,10 +121,6 @@ async def generate_llm_input(
             else {}
         )
 
-        # Use the agent's configured model when no simulation-specific model is set
-        if "model" not in completion_kwargs and agent_model:
-            completion_kwargs["model"] = agent_model
-
         simulation_model = completion_kwargs.get(
             "model", ChatModels.gpt_4_1_mini_2025_04_14
         )
diff --git a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
index 83e3498fb..3715ac226 100644
--- a/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
+++ b/packages/uipath/src/uipath/eval/mocks/_llm_mocker.py
@@ -182,10 +182,6 @@ async def response(
                     else {}
                 )
 
-                # Use the agent's configured model when no simulation-specific model is set
-                if "model" not in completion_kwargs and self.context.agent_model:
-                    completion_kwargs["model"] = self.context.agent_model
-
                 simulation_model = completion_kwargs.get(
                     "model", ChatModels.gpt_4_1_mini_2025_04_14
                 )
diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index bf2dae649..03c2c8aa7 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -28,6 +28,7 @@
     LLMMockingStrategy,
     MockingContext,
     MockingStrategyType,
+    ModelSettings,
     ToolSimulation,
 )
 
@@ -63,11 +64,12 @@ def load_simulation_config(agent_model: str | None = None) -> MockingContext | N
         if not tools_to_simulate:
             return None
 
-        # Create LLM mocking strategy
+        # Create LLM mocking strategy with the agent's model
         mocking_strategy = LLMMockingStrategy(
             type=MockingStrategyType.LLM,
             prompt=simulation_data.get("instructions", ""),
             tools_to_simulate=tools_to_simulate,
+            model=ModelSettings(model=agent_model) if agent_model else None,
         )
 
         # Create MockingContext for debugging
@@ -75,7 +77,6 @@ def load_simulation_config(agent_model: str | None = None) -> MockingContext | N
             strategy=mocking_strategy,
             name="debug-simulation",
             inputs={},
-            agent_model=agent_model,
         )
 
         logger.info(f"Loaded simulation config for {len(tools_to_simulate)} tool(s)")
@@ -140,15 +141,6 @@ def __init__(
         self._mocking_context = mocking_context or load_simulation_config(
             agent_model=agent_model
         )
-        # If mocking_context was passed without agent_model, inject it
-        if (
-            self._mocking_context
-            and not self._mocking_context.agent_model
-            and agent_model
-        ):
-            self._mocking_context = self._mocking_context.model_copy(
-                update={"agent_model": agent_model}
-            )
         self._span_collector = span_collector or ExecutionSpanCollector()
         self._execution_id = execution_id or str(uuid.uuid4())
         self._eval_set_run_id = eval_set_run_id
diff --git a/packages/uipath/src/uipath/eval/mocks/_types.py b/packages/uipath/src/uipath/eval/mocks/_types.py
index 043613c44..827569879 100644
--- a/packages/uipath/src/uipath/eval/mocks/_types.py
+++ b/packages/uipath/src/uipath/eval/mocks/_types.py
@@ -127,7 +127,6 @@ class MockingContext(BaseModel):
     strategy: MockingStrategy | None
     inputs: dict[str, Any] = Field(default_factory=lambda: {})
     name: str = Field(default="debug")
-    agent_model: str | None = Field(default=None)
 
 
 class ExampleCall(BaseModel):
diff --git a/packages/uipath/src/uipath/eval/runtime/runtime.py b/packages/uipath/src/uipath/eval/runtime/runtime.py
index 8b25b01eb..73385c141 100644
--- a/packages/uipath/src/uipath/eval/runtime/runtime.py
+++ b/packages/uipath/src/uipath/eval/runtime/runtime.py
@@ -53,7 +53,7 @@
 )
 from ..mocks._mock_context import cache_manager_context
 from ..mocks._mock_runtime import UiPathMockRuntime
-from ..mocks._types import MockingContext
+from ..mocks._types import LLMMockingStrategy, MockingContext, ModelSettings
 from ..models import EvaluationResult
 from ..models.evaluation_set import (
     EvaluationItem,
@@ -526,15 +526,28 @@ async def _execute_eval(
                                 eval_item=eval_item,
                             ),
                         )
+                    # Set agent model on the mocking strategy if not already set
+                    mocking_strategy = eval_item.mocking_strategy
+                    if (
+                        mocking_strategy
+                        and isinstance(mocking_strategy, LLMMockingStrategy)
+                        and not mocking_strategy.model
+                        and self.context.agent_model
+                    ):
+                        mocking_strategy = mocking_strategy.model_copy(
+                            update={
+                                "model": ModelSettings(model=self.context.agent_model)
+                            }
+                        )
+
                     agent_execution_output = await self.execute_runtime(
                         eval_item,
                         execution_id,
                         input_overrides=self.context.input_overrides,
                         mocking_context=MockingContext(
-                            strategy=eval_item.mocking_strategy,
+                            strategy=mocking_strategy,
                             name=eval_item.name,
                             inputs=eval_item.inputs,
-                            agent_model=self.context.agent_model,
                         ),
                         eval_set_run_id=self.context.eval_set_run_id,
                     )
@@ -812,12 +825,18 @@ async def _generate_input_for_eval(
             or getattr(eval_item, "expected_output", None)
             or {}
         )
+        # Set agent model on the input mocking strategy if not already set
+        input_strategy = eval_item.input_mocking_strategy
+        if input_strategy and not input_strategy.model and self.context.agent_model:
+            input_strategy = input_strategy.model_copy(
+                update={"model": ModelSettings(model=self.context.agent_model)}
+            )
+
         generated_input = await generate_llm_input(
-            eval_item.input_mocking_strategy,
+            input_strategy,
             (await self.get_schema()).input,
             expected_behavior=eval_item.expected_agent_behavior or "",
             expected_output=expected_output,
-            agent_model=self.context.agent_model,
         )
         updated_eval_item = eval_item.model_copy(update={"inputs": generated_input})
         return updated_eval_item
@@ -870,7 +889,6 @@ async def execute_runtime(
                     span_collector=self.span_collector,
                     execution_id=execution_id,
                     eval_set_run_id=eval_set_run_id,
-                    agent_model=self.context.agent_model,
                 )
 
             execution_runtime = UiPathExecutionRuntime(

From 81c9858cac5c54e226cc01dbb2953af11a2fe83a Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 10:51:04 -0700
Subject: [PATCH 08/12] =?UTF-8?q?refactor:=20address=20PR=20review=20?=
 =?UTF-8?q?=E2=80=94=20build=20mocking=5Fcontext=20in=20cli=5Fdebug,=20kee?=
 =?UTF-8?q?p=20cli=5Feval=20minimal?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Per @akshaylive review:
- cli_debug.py: build MockingContext via load_simulation_config(agent_model)
  and pass it directly, instead of relying on UiPathMockRuntime's fallback
- cli_eval.py: keep _get_agent_model() call for load_evaluators unchanged,
  only store result on eval_context for downstream use

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/_cli/cli_debug.py | 9 +++++++--
 packages/uipath/src/uipath/_cli/cli_eval.py  | 7 +++----
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/packages/uipath/src/uipath/_cli/cli_debug.py b/packages/uipath/src/uipath/_cli/cli_debug.py
index eb2ebc160..d2e08353b 100644
--- a/packages/uipath/src/uipath/_cli/cli_debug.py
+++ b/packages/uipath/src/uipath/_cli/cli_debug.py
@@ -9,6 +9,7 @@
 from uipath._cli._utils._studio_project import StudioClient
 from uipath.core.tracing import UiPathTraceManager
 from uipath.eval.mocks import UiPathMockRuntime
+from uipath.eval.mocks._mock_runtime import load_simulation_config
 from uipath.platform.common import ResourceOverwritesContext, UiPathConfig
 from uipath.runtime import (
     UiPathExecuteOptions,
@@ -163,15 +164,19 @@ async def execute_debug_runtime():
                                 trigger_poll_interval=trigger_poll_interval,
                             )
 
-                            # Get agent model from runtime schema for simulations
+                            # Build mocking context with agent model for simulations
                             schema = await runtime.get_schema()
                             agent_model = None
                             if schema.metadata and "settings" in schema.metadata:
                                 agent_model = schema.metadata["settings"].get("model")
 
+                            mocking_context = load_simulation_config(
+                                agent_model=agent_model
+                            )
+
                             mock_runtime = UiPathMockRuntime(
                                 delegate=debug_runtime,
-                                agent_model=agent_model,
+                                mocking_context=mocking_context,
                             )
 
                             try:
diff --git a/packages/uipath/src/uipath/_cli/cli_eval.py b/packages/uipath/src/uipath/_cli/cli_eval.py
index e82c33fa9..b1e6ad495 100644
--- a/packages/uipath/src/uipath/_cli/cli_eval.py
+++ b/packages/uipath/src/uipath/_cli/cli_eval.py
@@ -428,14 +428,13 @@ async def execute_eval():
 
                         eval_context.runtime_schema = await runtime.get_schema()
 
-                        eval_context.agent_model = _get_agent_model(
-                            eval_context.runtime_schema
-                        )
+                        agent_model = _get_agent_model(eval_context.runtime_schema)
+                        eval_context.agent_model = agent_model
 
                         eval_context.evaluators = await EvalHelpers.load_evaluators(
                             resolved_eval_set_path,
                             eval_context.evaluation_set,
-                            eval_context.agent_model,
+                            agent_model,
                         )
 
                         # Runtime is not required anymore.

From ae6f134dba00acfd395e0281cc5a85a73ac93908 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 10:55:48 -0700
Subject: [PATCH 09/12] fix: honor UI-specified model in simulation.json over
 agent model

load_simulation_config now checks simulation.json for a model field
first. If the UI specified a model, it's honored. Otherwise falls back
to the agent's configured model.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 .../uipath/src/uipath/eval/mocks/_mock_runtime.py   | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index 03c2c8aa7..30a56126a 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -64,12 +64,21 @@ def load_simulation_config(agent_model: str | None = None) -> MockingContext | N
         if not tools_to_simulate:
             return None
 
-        # Create LLM mocking strategy with the agent's model
+        # Honor model from simulation config if specified, otherwise use the agent model
+        simulation_model = simulation_data.get("model")
+        model = (
+            ModelSettings(model=simulation_model)
+            if simulation_model
+            else ModelSettings(model=agent_model)
+            if agent_model
+            else None
+        )
+
         mocking_strategy = LLMMockingStrategy(
             type=MockingStrategyType.LLM,
             prompt=simulation_data.get("instructions", ""),
             tools_to_simulate=tools_to_simulate,
-            model=ModelSettings(model=agent_model) if agent_model else None,
+            model=model,
         )
 
         # Create MockingContext for debugging

From 9bad552a18a886231f740b35438f97c2deafc4ea Mon Sep 17 00:00:00 2001
From: Akshaya Shanbhogue <akshaya.shanbhogue@uipath.com>
Date: Thu, 9 Apr 2026 11:34:56 -0700
Subject: [PATCH 10/12] fix: resolve circular import between cli_eval and
 eval.runtime
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Move _get_agent_model import from top-level in runtime.py to local
imports at the two call sites, breaking the cycle:
  eval.runtime.__init__ → _evaluate → runtime → cli_eval → eval.runtime

Also removes agent_model from UiPathEvalContext since the model is now
extracted lazily from runtime_schema where needed.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/_cli/cli_eval.py   |  5 +---
 .../uipath/src/uipath/eval/runtime/context.py |  1 -
 .../uipath/src/uipath/eval/runtime/runtime.py | 26 ++++++++++++-------
 3 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/packages/uipath/src/uipath/_cli/cli_eval.py b/packages/uipath/src/uipath/_cli/cli_eval.py
index b1e6ad495..ef1edb200 100644
--- a/packages/uipath/src/uipath/_cli/cli_eval.py
+++ b/packages/uipath/src/uipath/_cli/cli_eval.py
@@ -428,13 +428,10 @@ async def execute_eval():
 
                         eval_context.runtime_schema = await runtime.get_schema()
 
-                        agent_model = _get_agent_model(eval_context.runtime_schema)
-                        eval_context.agent_model = agent_model
-
                         eval_context.evaluators = await EvalHelpers.load_evaluators(
                             resolved_eval_set_path,
                             eval_context.evaluation_set,
-                            agent_model,
+                            _get_agent_model(eval_context.runtime_schema),
                         )
 
                         # Runtime is not required anymore.
diff --git a/packages/uipath/src/uipath/eval/runtime/context.py b/packages/uipath/src/uipath/eval/runtime/context.py
index 81263c54d..b8224718c 100644
--- a/packages/uipath/src/uipath/eval/runtime/context.py
+++ b/packages/uipath/src/uipath/eval/runtime/context.py
@@ -27,4 +27,3 @@ class UiPathEvalContext:
     input_overrides: dict[str, Any] | None = None
     resume: bool = False
     job_id: str | None = None
-    agent_model: str | None = None
diff --git a/packages/uipath/src/uipath/eval/runtime/runtime.py b/packages/uipath/src/uipath/eval/runtime/runtime.py
index 73385c141..ceb44d14b 100644
--- a/packages/uipath/src/uipath/eval/runtime/runtime.py
+++ b/packages/uipath/src/uipath/eval/runtime/runtime.py
@@ -532,13 +532,14 @@ async def _execute_eval(
                         mocking_strategy
                         and isinstance(mocking_strategy, LLMMockingStrategy)
                         and not mocking_strategy.model
-                        and self.context.agent_model
                     ):
-                        mocking_strategy = mocking_strategy.model_copy(
-                            update={
-                                "model": ModelSettings(model=self.context.agent_model)
-                            }
-                        )
+                        from ..._cli.cli_eval import _get_agent_model
+
+                        mocking_model = _get_agent_model(self.context.runtime_schema)
+                        if mocking_model:
+                            mocking_strategy = mocking_strategy.model_copy(
+                                update={"model": ModelSettings(model=mocking_model)}
+                            )
 
                     agent_execution_output = await self.execute_runtime(
                         eval_item,
@@ -827,10 +828,15 @@ async def _generate_input_for_eval(
         )
         # Set agent model on the input mocking strategy if not already set
         input_strategy = eval_item.input_mocking_strategy
-        if input_strategy and not input_strategy.model and self.context.agent_model:
-            input_strategy = input_strategy.model_copy(
-                update={"model": ModelSettings(model=self.context.agent_model)}
-            )
+        # If input strategy does not specify a model, extract it
+        if input_strategy and not input_strategy.model:
+            from ..._cli.cli_eval import _get_agent_model
+
+            input_generation_model = _get_agent_model(self.context.runtime_schema)
+            if input_generation_model:
+                input_strategy = input_strategy.model_copy(
+                    update={"model": ModelSettings(model=input_generation_model)}
+                )
 
         generated_input = await generate_llm_input(
             input_strategy,

From 8d0532dfc7a50ec05ccd40fc1458571271ee82e6 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 12:05:50 -0700
Subject: [PATCH 11/12] refactor: move get_agent_model to eval/helpers to avoid
 circular import

Move _get_agent_model from cli_eval.py to eval/helpers.py as
get_agent_model. This eliminates the inline imports in runtime.py
that were needed to avoid the circular dependency between
eval/runtime/runtime.py and _cli/cli_eval.py.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 packages/uipath/src/uipath/_cli/cli_eval.py   | 23 -------------------
 packages/uipath/src/uipath/eval/helpers.py    | 23 +++++++++++++++++++
 .../src/uipath/eval/mocks/_mock_runtime.py    |  5 +---
 .../uipath/src/uipath/eval/runtime/runtime.py |  9 +++-----
 4 files changed, 27 insertions(+), 33 deletions(-)

diff --git a/packages/uipath/src/uipath/_cli/cli_eval.py b/packages/uipath/src/uipath/_cli/cli_eval.py
index ef1edb200..9ebe275dc 100644
--- a/packages/uipath/src/uipath/_cli/cli_eval.py
+++ b/packages/uipath/src/uipath/_cli/cli_eval.py
@@ -25,7 +25,6 @@
 from uipath.runtime import (
     UiPathRuntimeContext,
     UiPathRuntimeFactoryRegistry,
-    UiPathRuntimeSchema,
 )
 from uipath.telemetry._track import flush_events
 from uipath.tracing import (
@@ -65,27 +64,6 @@ def setup_reporting_prereq(no_report: bool) -> bool:
     return True
 
 
-def _get_agent_model(schema: UiPathRuntimeSchema) -> str | None:
-    """Get agent model from the runtime schema metadata.
-
-    The model is read from schema.metadata["settings"]["model"] which is
-    populated by the low-code agents runtime from agent.json.
-
-    Returns:
-        The model name from agent settings, or None if not found.
-    """
-    try:
-        if schema.metadata and "settings" in schema.metadata:
-            settings = schema.metadata["settings"]
-            model = settings.get("model")
-            if model:
-                logger.debug(f"Got agent model from schema.metadata: {model}")
-                return model
-        return None
-    except Exception:
-        return None
-
-
 def _resolve_model_settings_override(
     model_settings_id: str, evaluation_set: EvaluationSet
 ) -> dict[str, Any] | None:
@@ -431,7 +409,6 @@ async def execute_eval():
                         eval_context.evaluators = await EvalHelpers.load_evaluators(
                             resolved_eval_set_path,
                             eval_context.evaluation_set,
-                            _get_agent_model(eval_context.runtime_schema),
                         )
 
                         # Runtime is not required anymore.
diff --git a/packages/uipath/src/uipath/eval/helpers.py b/packages/uipath/src/uipath/eval/helpers.py
index 0a0a0ca7f..8405e4a7a 100644
--- a/packages/uipath/src/uipath/eval/helpers.py
+++ b/packages/uipath/src/uipath/eval/helpers.py
@@ -7,6 +7,8 @@
 
 from pydantic import ValidationError
 
+from uipath.runtime.schema import UiPathRuntimeSchema
+
 from .evaluators.base_evaluator import GenericBaseEvaluator
 from .evaluators.evaluator_factory import EvaluatorFactory
 from .mocks._types import InputMockingStrategy, LLMMockingStrategy
@@ -277,3 +279,24 @@ async def load_evaluators(
             )
 
         return evaluators
+
+
+def get_agent_model(schema: UiPathRuntimeSchema) -> str | None:
+    """Get agent model from the runtime schema metadata.
+
+    The model is read from schema.metadata["settings"]["model"] which is
+    populated by the low-code agents runtime from agent.json.
+
+    Returns:
+        The model name from agent settings, or None if not found.
+    """
+    try:
+        if schema.metadata and "settings" in schema.metadata:
+            settings = schema.metadata["settings"]
+            model = settings.get("model")
+            if model:
+                logger.debug(f"Got agent model from schema.metadata: {model}")
+                return model
+        return None
+    except Exception:
+        return None
diff --git a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
index 30a56126a..df41dadeb 100644
--- a/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
+++ b/packages/uipath/src/uipath/eval/mocks/_mock_runtime.py
@@ -144,12 +144,9 @@ def __init__(
         span_collector: ExecutionSpanCollector | None = None,
         execution_id: str | None = None,
         eval_set_run_id: str | None = None,
-        agent_model: str | None = None,
     ):
         self.delegate = delegate
-        self._mocking_context = mocking_context or load_simulation_config(
-            agent_model=agent_model
-        )
+        self._mocking_context = mocking_context or load_simulation_config()
         self._span_collector = span_collector or ExecutionSpanCollector()
         self._execution_id = execution_id or str(uuid.uuid4())
         self._eval_set_run_id = eval_set_run_id
diff --git a/packages/uipath/src/uipath/eval/runtime/runtime.py b/packages/uipath/src/uipath/eval/runtime/runtime.py
index ceb44d14b..7f7614446 100644
--- a/packages/uipath/src/uipath/eval/runtime/runtime.py
+++ b/packages/uipath/src/uipath/eval/runtime/runtime.py
@@ -47,6 +47,7 @@
 from .._execution_context import ExecutionSpanCollector
 from ..evaluators.base_evaluator import GenericBaseEvaluator
 from ..evaluators.output_evaluator import OutputEvaluationCriteria
+from ..helpers import get_agent_model
 from ..mocks._cache_manager import CacheManager
 from ..mocks._input_mocker import (
     generate_llm_input,
@@ -533,9 +534,7 @@ async def _execute_eval(
                         and isinstance(mocking_strategy, LLMMockingStrategy)
                         and not mocking_strategy.model
                     ):
-                        from ..._cli.cli_eval import _get_agent_model
-
-                        mocking_model = _get_agent_model(self.context.runtime_schema)
+                        mocking_model = get_agent_model(self.context.runtime_schema)
                         if mocking_model:
                             mocking_strategy = mocking_strategy.model_copy(
                                 update={"model": ModelSettings(model=mocking_model)}
@@ -830,9 +829,7 @@ async def _generate_input_for_eval(
         input_strategy = eval_item.input_mocking_strategy
         # If input strategy does not specify a model, extract it
         if input_strategy and not input_strategy.model:
-            from ..._cli.cli_eval import _get_agent_model
-
-            input_generation_model = _get_agent_model(self.context.runtime_schema)
+            input_generation_model = get_agent_model(self.context.runtime_schema)
             if input_generation_model:
                 input_strategy = input_strategy.model_copy(
                     update={"model": ModelSettings(model=input_generation_model)}

From 3bdf2fc4dd2aee375cacbe62470d95574810c217 Mon Sep 17 00:00:00 2001
From: Chibi Vikram <chibicha@live.com>
Date: Thu, 9 Apr 2026 13:24:28 -0700
Subject: [PATCH 12/12] fix: update test to import get_agent_model from helpers

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 .../cli/eval/test_eval_runtime_metadata.py    | 22 +++++++++----------
 1 file changed, 10 insertions(+), 12 deletions(-)

diff --git a/packages/uipath/tests/cli/eval/test_eval_runtime_metadata.py b/packages/uipath/tests/cli/eval/test_eval_runtime_metadata.py
index 112f8774b..07042cc12 100644
--- a/packages/uipath/tests/cli/eval/test_eval_runtime_metadata.py
+++ b/packages/uipath/tests/cli/eval/test_eval_runtime_metadata.py
@@ -1,7 +1,7 @@
 """Tests for UiPathEvalRuntime metadata loading functionality.
 
 This module tests:
-- _get_agent_model() - cached agent model retrieval
+- get_agent_model() - cached agent model retrieval
 - get_schema() - cached schema retrieval
 """
 
@@ -10,11 +10,9 @@
 
 import pytest
 
-from uipath._cli.cli_eval import (
-    _get_agent_model,
-)
 from uipath.core.events import EventBus
 from uipath.core.tracing import UiPathTraceManager
+from uipath.eval.helpers import get_agent_model
 from uipath.eval.runtime import UiPathEvalContext, UiPathEvalRuntime
 from uipath.runtime import (
     UiPathExecuteOptions,
@@ -119,34 +117,34 @@ async def dispose(self) -> None:
 
 
 class TestGetAgentModel:
-    """Tests for _get_agent_model function."""
+    """Tests for get_agent_model function."""
 
     @pytest.mark.asyncio
     async def test_returns_agent_model(self):
-        """Test that _get_agent_model returns the correct model from schema."""
+        """Test that get_agent_model returns the correct model from schema."""
         schema = MockRuntimeSchema()
         schema.metadata = {"settings": {"model": "gpt-4o-2024-11-20"}}
 
-        model = _get_agent_model(schema)
+        model = get_agent_model(schema)
         assert model == "gpt-4o-2024-11-20"
 
     @pytest.mark.asyncio
     async def test_returns_none_when_no_model(self):
-        """Test that _get_agent_model returns None when runtime has no model."""
+        """Test that get_agent_model returns None when runtime has no model."""
         schema = MockRuntimeSchema()
 
-        model = _get_agent_model(schema)
+        model = get_agent_model(schema)
         assert model is None
 
     @pytest.mark.asyncio
     async def test_returns_model_consistently(self):
-        """Test that _get_agent_model returns consistent results."""
+        """Test that get_agent_model returns consistent results."""
         schema = MockRuntimeSchema()
         schema.metadata = {"settings": {"model": "consistent-model"}}
 
         # Multiple calls should return the same value
-        model1 = _get_agent_model(schema)
-        model2 = _get_agent_model(schema)
+        model1 = get_agent_model(schema)
+        model2 = get_agent_model(schema)
 
         assert model1 == model2 == "consistent-model"