fix: move LLMError into the llm package (#1096)

packages/railtracks/src/railtracks/built_nodes/concrete/_llm_base.py

@@ -15,7 +15,7 @@
 from pydantic import BaseModel
 from typing_extensions import Self
 
-from railtracks.exceptions.errors import LLMError, NodeInvocationError
+from railtracks.exceptions.errors import NodeInvocationError
 from railtracks.exceptions.messages.exception_messages import get_message
 from railtracks.llm import (
     Message,
@@ -26,6 +26,7 @@
     SystemMessage,
     UserMessage,
 )
+from railtracks.llm.errors import LLMError
 from railtracks.llm.response import Response
 from railtracks.nodes.nodes import Node
 from railtracks.prompts.prompt import inject_context

packages/railtracks/src/railtracks/built_nodes/concrete/_tool_call_base.py

@@ -15,7 +15,7 @@
 )
 
 from railtracks.built_nodes.concrete.response import LLMResponse
-from railtracks.exceptions import LLMError, NodeCreationError
+from railtracks.exceptions import NodeCreationError
 from railtracks.interaction._call import call
 from railtracks.llm import (
     AssistantMessage,
@@ -28,6 +28,7 @@
     UserMessage,
 )
 from railtracks.llm.content import Content
+from railtracks.llm.errors import LLMError
 from railtracks.llm.message import Role
 from railtracks.llm.providers import ModelProvider
 from railtracks.llm.response import Response

packages/railtracks/src/railtracks/built_nodes/concrete/chat_tool_call_llm.py

@@ -1,7 +1,6 @@
 import asyncio
 from abc import ABC
 
-from railtracks.exceptions import LLMError
 from railtracks.interaction import call
 from railtracks.llm import (
     AssistantMessage,
@@ -10,6 +9,7 @@
     ToolResponse,
     UserMessage,
 )
+from railtracks.llm.errors import LLMError
 from railtracks.llm.message import Role
 
 from ._llm_base import StringOutputMixIn

packages/railtracks/src/railtracks/built_nodes/concrete/structured_llm_base.py

@@ -4,8 +4,8 @@
 
 from pydantic import BaseModel
 
-from railtracks.exceptions.errors import LLMError
 from railtracks.llm import Message, MessageHistory, ModelBase, UserMessage
+from railtracks.llm.errors import LLMError
 from railtracks.validation.node_creation.validation import (
     check_classmethod,
     check_schema,

packages/railtracks/src/railtracks/built_nodes/concrete/structured_tool_call_llm_base.py

@@ -3,7 +3,6 @@
 
 from pydantic import BaseModel
 
-from railtracks.exceptions.errors import LLMError
 from railtracks.interaction import call
 from railtracks.llm import (
     AssistantMessage,
@@ -12,6 +11,7 @@
     ModelBase,
     UserMessage,
 )
+from railtracks.llm.errors import LLMError
 
 from ._llm_base import StructuredOutputMixIn
 from ._tool_call_base import OutputLessToolCallLLM

packages/railtracks/src/railtracks/built_nodes/concrete/terminal_llm_base.py

@@ -4,7 +4,7 @@
 from abc import ABC
 from typing import Generator, Generic, Literal, TypeVar
 
-from railtracks.exceptions import LLMError
+from railtracks.llm.errors import LLMError
 from railtracks.llm.response import Response
 
 from ._llm_base import LLMBase, StringOutputMixIn

packages/railtracks/src/railtracks/exceptions/__init__.py

@@ -2,7 +2,6 @@
     ContextError,
     FatalError,
     GlobalTimeOutError,
-    LLMError,
     NodeCreationError,
     NodeInvocationError,
 )
@@ -13,7 +12,6 @@
     "NodeCreationError",
     "NodeInvocationError",
     "GlobalTimeOutError",
-    "LLMError",
     "ContextError",
     "VisualExtraRequiredError",
 ]

packages/railtracks/src/railtracks/exceptions/errors.py

@@ -1,10 +1,5 @@
-from typing import TYPE_CHECKING
-
 from ._base import RTError
 
-if TYPE_CHECKING:
-    from railtracks.llm.history import MessageHistory
-
 
 class NodeInvocationError(RTError):
     """
@@ -55,44 +50,6 @@ def __str__(self):
         return self._color(base, self.RED)
 
 
-class LLMError(RTError):
-    """
-    Raised when an error occurs during LLM invocation or completion.
-    """
-
-    def __init__(
-        self,
-        reason: str,
-        message_history: "MessageHistory" = None,
-    ):
-        self.reason = reason
-        self.message_history = message_history
-
-        message = f"{self._color('LLM Error: ', self.BOLD_RED)}{self._color(reason, self.RED)}"
-        super().__init__(message)
-
-    def __str__(self):
-        base = super().__str__()
-        details = []
-        if self.message_history:
-            mh_str = str(self.message_history)
-            indented_mh = "\n".join(
-                "    " + line for line in mh_str.splitlines()
-            )  # 2 indents (2-spaces) per indent
-            details.append(
-                self._color("Message History:\n", self.BOLD_GREEN)
-                + self._color(indented_mh, self.GREEN)
-            )
-        if details:
-            notes_str = (
-                "\n"
-                + self._color("Details:\n", self.BOLD_GREEN)
-                + "\n".join(f"  {d}" for d in details)
-            )
-            return f"\n{self._color(base, self.RED)}{notes_str}"
-        return self._color(base, self.RED)
-
-
 class GlobalTimeOutError(RTError):
     """
     Raised on global timeout for whole execution.

packages/railtracks/src/railtracks/llm/__init__.py

@@ -1,5 +1,6 @@
 from . import retries
 from .content import ToolCall, ToolResponse
+from .errors import LLMError
 from .history import MessageHistory
 from .message import AssistantMessage, Message, SystemMessage, ToolMessage, UserMessage
 from .model import ModelBase
@@ -27,6 +28,7 @@
 
 __all__ = [
     "ModelBase",
+    "LLMError",
     "ToolCall",
     "ToolResponse",
     "UserMessage",

packages/railtracks/src/railtracks/llm/errors.py

@@ -0,0 +1,46 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from ._exceptions import RTLLMError
+
+if TYPE_CHECKING:
+    from .history import MessageHistory
+
+
+class LLMError(RTLLMError):
+    """
+    Raised when an error occurs during LLM invocation or completion.
+    """
+
+    def __init__(
+        self,
+        reason: str,
+        message_history: "MessageHistory" = None,
+    ):
+        self.reason = reason
+        self.message_history = message_history
+
+        message = f"{self._color('LLM Error: ', self.BOLD_RED)}{self._color(reason, self.RED)}"
+        super().__init__(message)
+
+    def __str__(self):
+        base = super().__str__()
+        details = []
+        if self.message_history:
+            mh_str = str(self.message_history)
+            indented_mh = "\n".join(
+                "    " + line for line in mh_str.splitlines()
+            )  # 2 indents (2-spaces) per indent
+            details.append(
+                self._color("Message History:\n", self.BOLD_GREEN)
+                + self._color(indented_mh, self.GREEN)
+            )
+        if details:
+            notes_str = (
+                "\n"
+                + self._color("Details:\n", self.BOLD_GREEN)
+                + "\n".join(f"  {d}" for d in details)
+            )
+            return f"\n{self._color(base, self.RED)}{notes_str}"
+        return self._color(base, self.RED)

packages/railtracks/src/railtracks/llm/models/_litellm_wrapper.py

@@ -26,15 +26,16 @@
 from litellm.types.utils import ModelResponse
 from pydantic import BaseModel, Field
 
-from ...exceptions.errors import LLMError, NodeInvocationError
 from ..content import ToolCall
+from ..errors import LLMError
 from ..history import MessageHistory
 from ..message import AssistantMessage, Message, ToolMessage, UserMessage
 from ..model import ModelBase
 from ..response import MessageInfo, Response
 from ..retries import RetryApproach
 from ..tools import Tool
 from ..tools.parameters import Parameter
+from ._model_exception_base import ModelError
 
 _TBaseModel = TypeVar("_TBaseModel", bound=BaseModel)
 
@@ -96,12 +97,8 @@ def _parameters_to_json_schema(
     ):
         return _handle_set_of_parameters(list(parameters))
 
-    raise NodeInvocationError(
-        message=f"Unable to parse Tool.parameters. It was {parameters}",
-        fatal=True,
-        notes=[
-            "Tool.parameters must be a set of Parameter objects",
-        ],
+    raise ModelError(
+        reason=f"Unable to parse Tool.parameters. It was {parameters}. Tool.parameters must be a set of Parameter objects.",
     )
 
 

packages/railtracks/tests/integration_tests/library/test_retry_agent_node.py

@@ -3,9 +3,8 @@
 
 import litellm
 import pytest
-
 import railtracks as rt
-from railtracks.exceptions import LLMError
+from railtracks.llm.errors import LLMError
 from railtracks.llm.retries import ExponentialRetry, FixedRetry
 
 

packages/railtracks/tests/unit_tests/llm/models/test_litellm_wrapper.py

@@ -1,20 +1,21 @@
+import json
+from json import JSONDecodeError
 from typing import Generator, Literal
 from unittest.mock import patch
+
+import litellm
 import pytest
+from pydantic import BaseModel
+from railtracks.llm import AssistantMessage, UserMessage
+from railtracks.llm.errors import LLMError
 from railtracks.llm.models._litellm_wrapper import (
     LiteLLMWrapper,
     _parameters_to_json_schema,
     _to_litellm_tool,
 )
-from railtracks.exceptions import NodeInvocationError, LLMError
-from railtracks.llm import AssistantMessage, UserMessage
+from railtracks.llm.models._model_exception_base import ModelError
 from railtracks.llm.providers import ModelProvider
-from pydantic import BaseModel
 from railtracks.llm.response import Response
-from json import JSONDecodeError
-import litellm
-from railtracks.llm.content import Stream
-import json
 
 
 class _ConcreteLiteLLMWrapperForTest(LiteLLMWrapper[Literal[False]]):
@@ -54,7 +55,7 @@ def test_parameters_to_json_schema_invalid_input(self):
         """
         Test _parameters_to_json_schema with invalid input.
         """
-        with pytest.raises(NodeInvocationError):
+        with pytest.raises(ModelError):
             _parameters_to_json_schema(123)  # type: ignore
 
     # =================================== END _parameters_to_json_schema Tests ====================================
@@ -252,9 +253,9 @@ class Schema(BaseModel):
             wrapper = mock_litellm_wrapper(content="Invalid JSON")
             method = getattr(wrapper, method_name)
             if is_async:
-                result = await method(message_history, schema=Schema)
+                await method(message_history, schema=Schema)
             else:
-                result = method(message_history, schema=Schema)
+                method(message_history, schema=Schema)
 
     @pytest.mark.asyncio
     @pytest.mark.parametrize("method_name,is_async", [
@@ -270,9 +271,9 @@ class Schema(BaseModel):
             wrapper = mock_litellm_wrapper(content='{"field": "VAL", "invalid": "json"}')
             method = getattr(wrapper, method_name)
             if is_async:
-                result = await method(message_history, schema=Schema)
+                await method(message_history, schema=Schema)
             else:
-                result = method(message_history, schema=Schema)
+                method(message_history, schema=Schema)
 
     @pytest.mark.parametrize("method_name,is_async,stream", [
         ("_chat_with_tools", False, False),
@@ -324,7 +325,7 @@ async def test_chat_with_tools(
                         assert calls[0]["name"] == "tool_x"
                         assert calls[0]["arguments"] == {"foo": 1}
                         assert calls[0]["identifier"] == "id123"
-                    except Exception as e:
+                    except Exception:
                         pytest.fail("Structured response did not match schema")
                 elif not isinstance(chunk, str):
                     pytest.fail("Stream yielded non-string, non-Response chunk")

packages/railtracks/tests/unit_tests/nodes/library/tool_calling_llms/test_structured_tool_call_llm.py

@@ -1,14 +1,14 @@
 import pytest
 import railtracks as rt
 from pydantic import BaseModel, Field
-
-from railtracks.llm.response import Response
-from railtracks.built_nodes.easy_usage_wrappers.helpers import structured_tool_call_llm, structured_llm
-from railtracks.exceptions import NodeCreationError, LLMError
-from railtracks.llm import MessageHistory, SystemMessage, UserMessage, AssistantMessage
-
 from railtracks.built_nodes.concrete import StructuredToolCallLLM
-
+from railtracks.built_nodes.easy_usage_wrappers.helpers import (
+    structured_llm,
+    structured_tool_call_llm,
+)
+from railtracks.exceptions import NodeCreationError
+from railtracks.llm import AssistantMessage, MessageHistory, SystemMessage, UserMessage
+from railtracks.llm.errors import LLMError
 
 # =========================== Basic functionality ==========================
 

packages/railtracks/tests/unit_tests/test_exceptions.py

@@ -1,13 +1,13 @@
 import pytest
+from railtracks.exceptions._base import RTError
 from railtracks.exceptions.errors import (
-    NodeInvocationError,
-    NodeCreationError,
-    LLMError,
-    GlobalTimeOutError,
     ContextError,
     FatalError,
+    GlobalTimeOutError,
+    NodeCreationError,
+    NodeInvocationError,
 )
-from railtracks.exceptions._base import RTError
+from railtracks.llm.errors import LLMError
 
 # NOTE: This file contains very basic tests to ensure that the exception classes are working as expected.