from __future__ import annotations
import asyncio
import json
from json import JSONDecodeError
from time import sleep
from typing import (
TYPE_CHECKING,
Any,
Callable,
Dict,
List,
Optional,
Sequence,
Tuple,
Type,
Union,
)
from langchain_core.agents import AgentAction, AgentFinish
from langchain_core.callbacks import CallbackManager
from langchain_core.load import dumpd
from langchain_core.runnables import RunnableConfig, RunnableSerializable, ensure_config
from langchain_core.tools import BaseTool
from langchain_core.utils.function_calling import convert_to_openai_tool
from pydantic import BaseModel, Field, model_validator
from typing_extensions import Self
if TYPE_CHECKING:
import openai
from openai.types.beta.threads import ThreadMessage
from openai.types.beta.threads.required_action_function_tool_call import (
RequiredActionFunctionToolCall,
)
[docs]
class OpenAIAssistantFinish(AgentFinish):
"""AgentFinish with run and thread metadata.
Parameters:
run_id: Run id.
thread_id: Thread id.
"""
run_id: str
thread_id: str
@classmethod
def is_lc_serializable(cls) -> bool:
"""Check if the class is serializable by LangChain.
Returns:
False
"""
return False
[docs]
class OpenAIAssistantAction(AgentAction):
"""AgentAction with info needed to submit custom tool output to existing run.
Parameters:
tool_call_id: Tool call id.
run_id: Run id.
thread_id: Thread id
"""
tool_call_id: str
run_id: str
thread_id: str
@classmethod
def is_lc_serializable(cls) -> bool:
"""Check if the class is serializable by LangChain.
Returns:
False
"""
return False
def _get_openai_client() -> openai.OpenAI:
try:
import openai
return openai.OpenAI()
except ImportError as e:
raise ImportError(
"Unable to import openai, please install with `pip install openai`."
) from e
except AttributeError as e:
raise AttributeError(
"Please make sure you are using a v1.1-compatible version of openai. You "
'can install with `pip install "openai>=1.1"`.'
) from e
def _get_openai_async_client() -> openai.AsyncOpenAI:
try:
import openai
return openai.AsyncOpenAI()
except ImportError as e:
raise ImportError(
"Unable to import openai, please install with `pip install openai`."
) from e
except AttributeError as e:
raise AttributeError(
"Please make sure you are using a v1.1-compatible version of openai. You "
'can install with `pip install "openai>=1.1"`.'
) from e
def _is_assistants_builtin_tool(
tool: Union[Dict[str, Any], Type[BaseModel], Callable, BaseTool],
) -> bool:
"""Determine if tool corresponds to OpenAI Assistants built-in."""
assistants_builtin_tools = ("code_interpreter", "retrieval")
return (
isinstance(tool, dict)
and ("type" in tool)
and (tool["type"] in assistants_builtin_tools)
)
def _get_assistants_tool(
tool: Union[Dict[str, Any], Type[BaseModel], Callable, BaseTool],
) -> Dict[str, Any]:
"""Convert a raw function/class to an OpenAI tool.
Note that OpenAI assistants supports several built-in tools,
such as "code_interpreter" and "retrieval."
"""
if _is_assistants_builtin_tool(tool):
return tool # type: ignore
else:
return convert_to_openai_tool(tool)
OutputType = Union[
List[OpenAIAssistantAction],
OpenAIAssistantFinish,
List["ThreadMessage"],
List["RequiredActionFunctionToolCall"],
]
[docs]
class OpenAIAssistantRunnable(RunnableSerializable[Dict, OutputType]):
"""Run an OpenAI Assistant.
Example using OpenAI tools:
.. code-block:: python
from langchain_experimental.openai_assistant import OpenAIAssistantRunnable
interpreter_assistant = OpenAIAssistantRunnable.create_assistant(
name="langchain assistant",
instructions="You are a personal math tutor. Write and run code to answer math questions.",
tools=[{"type": "code_interpreter"}],
model="gpt-4-1106-preview"
)
output = interpreter_assistant.invoke({"content": "What's 10 - 4 raised to the 2.7"})
Example using custom tools and AgentExecutor:
.. code-block:: python
from langchain_experimental.openai_assistant import OpenAIAssistantRunnable
from langchain.agents import AgentExecutor
from langchain.tools import E2BDataAnalysisTool
tools = [E2BDataAnalysisTool(api_key="...")]
agent = OpenAIAssistantRunnable.create_assistant(
name="langchain assistant e2b tool",
instructions="You are a personal math tutor. Write and run code to answer math questions.",
tools=tools,
model="gpt-4-1106-preview",
as_agent=True
)
agent_executor = AgentExecutor(agent=agent, tools=tools)
agent_executor.invoke({"content": "What's 10 - 4 raised to the 2.7"})
Example using custom tools and custom execution:
.. code-block:: python
from langchain_experimental.openai_assistant import OpenAIAssistantRunnable
from langchain.agents import AgentExecutor
from langchain_core.agents import AgentFinish
from langchain.tools import E2BDataAnalysisTool
tools = [E2BDataAnalysisTool(api_key="...")]
agent = OpenAIAssistantRunnable.create_assistant(
name="langchain assistant e2b tool",
instructions="You are a personal math tutor. Write and run code to answer math questions.",
tools=tools,
model="gpt-4-1106-preview",
as_agent=True
)
def execute_agent(agent, tools, input):
tool_map = {tool.name: tool for tool in tools}
response = agent.invoke(input)
while not isinstance(response, AgentFinish):
tool_outputs = []
for action in response:
tool_output = tool_map[action.tool].invoke(action.tool_input)
tool_outputs.append({"output": tool_output, "tool_call_id": action.tool_call_id})
response = agent.invoke(
{
"tool_outputs": tool_outputs,
"run_id": action.run_id,
"thread_id": action.thread_id
}
)
return response
response = execute_agent(agent, tools, {"content": "What's 10 - 4 raised to the 2.7"})
next_response = execute_agent(agent, tools, {"content": "now add 17.241", "thread_id": response.thread_id})
""" # noqa: E501
client: Any = Field(default_factory=_get_openai_client)
"""OpenAI or AzureOpenAI client."""
async_client: Any = None
"""OpenAI or AzureOpenAI async client."""
assistant_id: str
"""OpenAI assistant id."""
check_every_ms: float = 1_000.0
"""Frequency with which to check run progress in ms."""
as_agent: bool = False
"""Use as a LangChain agent, compatible with the AgentExecutor."""
@model_validator(mode="after")
def validate_async_client(self) -> Self:
if self.async_client is None:
import openai
api_key = self.client.api_key
self.async_client = openai.AsyncOpenAI(api_key=api_key)
return self
[docs]
@classmethod
def create_assistant(
cls,
name: str,
instructions: str,
tools: Sequence[Union[BaseTool, dict]],
model: str,
*,
client: Optional[Union[openai.OpenAI, openai.AzureOpenAI]] = None,
**kwargs: Any,
) -> OpenAIAssistantRunnable:
"""Create an OpenAI Assistant and instantiate the Runnable.
Args:
name: Assistant name.
instructions: Assistant instructions.
tools: Assistant tools. Can be passed in OpenAI format or as BaseTools.
model: Assistant model to use.
client: OpenAI or AzureOpenAI client.
Will create a default OpenAI client if not specified.
kwargs: Additional arguments.
Returns:
OpenAIAssistantRunnable configured to run using the created assistant.
"""
client = client or _get_openai_client()
assistant = client.beta.assistants.create(
name=name,
instructions=instructions,
tools=[_get_assistants_tool(tool) for tool in tools], # type: ignore
model=model,
)
return cls(assistant_id=assistant.id, client=client, **kwargs)
[docs]
def invoke(
self, input: dict, config: Optional[RunnableConfig] = None, **kwargs: Any
) -> OutputType:
"""Invoke assistant.
Args:
input: Runnable input dict that can have:
content: User message when starting a new run.
thread_id: Existing thread to use.
run_id: Existing run to use. Should only be supplied when providing
the tool output for a required action after an initial invocation.
message_metadata: Metadata to associate with new message.
thread_metadata: Metadata to associate with new thread. Only relevant
when new thread being created.
instructions: Additional run instructions.
model: Override Assistant model for this run.
tools: Override Assistant tools for this run.
run_metadata: Metadata to associate with new run.
config: Runnable config. Defaults to None.
Return:
If self.as_agent, will return
Union[List[OpenAIAssistantAction], OpenAIAssistantFinish].
Otherwise, will return OpenAI types
Union[List[ThreadMessage], List[RequiredActionFunctionToolCall]].
"""
config = ensure_config(config)
callback_manager = CallbackManager.configure(
inheritable_callbacks=config.get("callbacks"),
inheritable_tags=config.get("tags"),
inheritable_metadata=config.get("metadata"),
)
run_manager = callback_manager.on_chain_start(
dumpd(self), input, name=config.get("run_name") or self.get_name()
)
try:
# Being run within AgentExecutor and there are tool outputs to submit.
if self.as_agent and input.get("intermediate_steps"):
tool_outputs = self._parse_intermediate_steps(
input["intermediate_steps"]
)
run = self.client.beta.threads.runs.submit_tool_outputs(**tool_outputs)
# Starting a new thread and a new run.
elif "thread_id" not in input:
thread = {
"messages": [
{
"role": "user",
"content": input["content"],
"metadata": input.get("message_metadata"),
}
],
"metadata": input.get("thread_metadata"),
}
run = self._create_thread_and_run(input, thread)
# Starting a new run in an existing thread.
elif "run_id" not in input:
_ = self.client.beta.threads.messages.create(
input["thread_id"],
content=input["content"],
role="user",
metadata=input.get("message_metadata"),
)
run = self._create_run(input)
# Submitting tool outputs to an existing run, outside the AgentExecutor
# framework.
else:
run = self.client.beta.threads.runs.submit_tool_outputs(**input)
run = self._wait_for_run(run.id, run.thread_id)
except BaseException as e:
run_manager.on_chain_error(e)
raise e
try:
response = self._get_response(run)
except BaseException as e:
run_manager.on_chain_error(e, metadata=run.dict())
raise e
else:
run_manager.on_chain_end(response)
return response
[docs]
@classmethod
async def acreate_assistant(
cls,
name: str,
instructions: str,
tools: Sequence[Union[BaseTool, dict]],
model: str,
*,
async_client: Optional[
Union[openai.AsyncOpenAI, openai.AsyncAzureOpenAI]
] = None,
**kwargs: Any,
) -> OpenAIAssistantRunnable:
"""Async create an AsyncOpenAI Assistant and instantiate the Runnable.
Args:
name: Assistant name.
instructions: Assistant instructions.
tools: Assistant tools. Can be passed in OpenAI format or as BaseTools.
model: Assistant model to use.
async_client: AsyncOpenAI client.
Will create default async_client if not specified.
Returns:
AsyncOpenAIAssistantRunnable configured to run using the created assistant.
"""
async_client = async_client or _get_openai_async_client()
openai_tools = [_get_assistants_tool(tool) for tool in tools]
assistant = await async_client.beta.assistants.create(
name=name,
instructions=instructions,
tools=openai_tools, # type: ignore
model=model,
)
return cls(assistant_id=assistant.id, async_client=async_client, **kwargs)
[docs]
async def ainvoke(
self, input: dict, config: Optional[RunnableConfig] = None, **kwargs: Any
) -> OutputType:
"""Async invoke assistant.
Args:
input: Runnable input dict that can have:
content: User message when starting a new run.
thread_id: Existing thread to use.
run_id: Existing run to use. Should only be supplied when providing
the tool output for a required action after an initial invocation.
message_metadata: Metadata to associate with a new message.
thread_metadata: Metadata to associate with new thread. Only relevant
when a new thread is created.
instructions: Additional run instructions.
model: Override Assistant model for this run.
tools: Override Assistant tools for this run.
run_metadata: Metadata to associate with new run.
config: Runnable config. Defaults to None.
kwargs: Additional arguments.
Return:
If self.as_agent, will return
Union[List[OpenAIAssistantAction], OpenAIAssistantFinish].
Otherwise, will return OpenAI types
Union[List[ThreadMessage], List[RequiredActionFunctionToolCall]].
"""
config = config or {}
callback_manager = CallbackManager.configure(
inheritable_callbacks=config.get("callbacks"),
inheritable_tags=config.get("tags"),
inheritable_metadata=config.get("metadata"),
)
run_manager = callback_manager.on_chain_start(
dumpd(self), input, name=config.get("run_name") or self.get_name()
)
try:
# Being run within AgentExecutor and there are tool outputs to submit.
if self.as_agent and input.get("intermediate_steps"):
tool_outputs = await self._aparse_intermediate_steps(
input["intermediate_steps"]
)
run = await self.async_client.beta.threads.runs.submit_tool_outputs(
**tool_outputs
)
# Starting a new thread and a new run.
elif "thread_id" not in input:
thread = {
"messages": [
{
"role": "user",
"content": input["content"],
"metadata": input.get("message_metadata"),
}
],
"metadata": input.get("thread_metadata"),
}
run = await self._acreate_thread_and_run(input, thread)
# Starting a new run in an existing thread.
elif "run_id" not in input:
_ = await self.async_client.beta.threads.messages.create(
input["thread_id"],
content=input["content"],
role="user",
metadata=input.get("message_metadata"),
)
run = await self._acreate_run(input)
# Submitting tool outputs to an existing run, outside the AgentExecutor
# framework.
else:
run = await self.async_client.beta.threads.runs.submit_tool_outputs(
**input
)
run = await self._await_for_run(run.id, run.thread_id)
except BaseException as e:
run_manager.on_chain_error(e)
raise e
try:
response = self._get_response(run)
except BaseException as e:
run_manager.on_chain_error(e, metadata=run.dict())
raise e
else:
run_manager.on_chain_end(response)
return response
def _parse_intermediate_steps(
self, intermediate_steps: List[Tuple[OpenAIAssistantAction, str]]
) -> dict:
last_action, last_output = intermediate_steps[-1]
run = self._wait_for_run(last_action.run_id, last_action.thread_id)
required_tool_call_ids = set()
if run.required_action:
required_tool_call_ids = {
tc.id for tc in run.required_action.submit_tool_outputs.tool_calls
}
tool_outputs = [
{"output": str(output), "tool_call_id": action.tool_call_id}
for action, output in intermediate_steps
if action.tool_call_id in required_tool_call_ids
]
submit_tool_outputs = {
"tool_outputs": tool_outputs,
"run_id": last_action.run_id,
"thread_id": last_action.thread_id,
}
return submit_tool_outputs
def _create_run(self, input: dict) -> Any:
params = {
k: v
for k, v in input.items()
if k in ("instructions", "model", "tools", "run_metadata")
}
return self.client.beta.threads.runs.create(
input["thread_id"],
assistant_id=self.assistant_id,
**params,
)
def _create_thread_and_run(self, input: dict, thread: dict) -> Any:
params = {
k: v
for k, v in input.items()
if k in ("instructions", "model", "tools", "run_metadata")
}
run = self.client.beta.threads.create_and_run(
assistant_id=self.assistant_id,
thread=thread,
**params,
)
return run
def _get_response(self, run: Any) -> Any:
# TODO: Pagination
if run.status == "completed":
import openai
major_version = int(openai.version.VERSION.split(".")[0])
minor_version = int(openai.version.VERSION.split(".")[1])
version_gte_1_14 = (major_version > 1) or (
major_version == 1 and minor_version >= 14
)
messages = self.client.beta.threads.messages.list(
run.thread_id, order="asc"
)
new_messages = [msg for msg in messages if msg.run_id == run.id]
if not self.as_agent:
return new_messages
answer: Any = [
msg_content for msg in new_messages for msg_content in msg.content
]
if all(
(
isinstance(content, openai.types.beta.threads.TextContentBlock)
if version_gte_1_14
else isinstance(
content, openai.types.beta.threads.MessageContentText
)
)
for content in answer
):
answer = "\n".join(content.text.value for content in answer)
return OpenAIAssistantFinish(
return_values={
"output": answer,
"thread_id": run.thread_id,
"run_id": run.id,
},
log="",
run_id=run.id,
thread_id=run.thread_id,
)
elif run.status == "requires_action":
if not self.as_agent:
return run.required_action.submit_tool_outputs.tool_calls
actions = []
for tool_call in run.required_action.submit_tool_outputs.tool_calls:
function = tool_call.function
try:
args = json.loads(function.arguments, strict=False)
except JSONDecodeError as e:
raise ValueError(
f"Received invalid JSON function arguments: "
f"{function.arguments} for function {function.name}"
) from e
if len(args) == 1 and "__arg1" in args:
args = args["__arg1"]
actions.append(
OpenAIAssistantAction(
tool=function.name,
tool_input=args,
tool_call_id=tool_call.id,
log="",
run_id=run.id,
thread_id=run.thread_id,
)
)
return actions
else:
run_info = json.dumps(run.dict(), indent=2)
raise ValueError(
f"Unexpected run status: {run.status}. Full run info:\n\n{run_info})"
)
def _wait_for_run(self, run_id: str, thread_id: str) -> Any:
in_progress = True
while in_progress:
run = self.client.beta.threads.runs.retrieve(run_id, thread_id=thread_id)
in_progress = run.status in ("in_progress", "queued")
if in_progress:
sleep(self.check_every_ms / 1000)
return run
async def _aparse_intermediate_steps(
self, intermediate_steps: List[Tuple[OpenAIAssistantAction, str]]
) -> dict:
last_action, last_output = intermediate_steps[-1]
run = await self._wait_for_run(last_action.run_id, last_action.thread_id)
required_tool_call_ids = set()
if run.required_action:
required_tool_call_ids = {
tc.id for tc in run.required_action.submit_tool_outputs.tool_calls
}
tool_outputs = [
{"output": str(output), "tool_call_id": action.tool_call_id}
for action, output in intermediate_steps
if action.tool_call_id in required_tool_call_ids
]
submit_tool_outputs = {
"tool_outputs": tool_outputs,
"run_id": last_action.run_id,
"thread_id": last_action.thread_id,
}
return submit_tool_outputs
async def _acreate_run(self, input: dict) -> Any:
params = {
k: v
for k, v in input.items()
if k in ("instructions", "model", "tools", "run_metadata")
}
return await self.async_client.beta.threads.runs.create(
input["thread_id"],
assistant_id=self.assistant_id,
**params,
)
async def _acreate_thread_and_run(self, input: dict, thread: dict) -> Any:
params = {
k: v
for k, v in input.items()
if k in ("instructions", "model", "tools", "run_metadata")
}
run = await self.async_client.beta.threads.create_and_run(
assistant_id=self.assistant_id,
thread=thread,
**params,
)
return run
async def _aget_response(self, run: Any) -> Any:
# TODO: Pagination
if run.status == "completed":
import openai
major_version = int(openai.version.VERSION.split(".")[0])
minor_version = int(openai.version.VERSION.split(".")[1])
version_gte_1_14 = (major_version > 1) or (
major_version == 1 and minor_version >= 14
)
messages = await self.async_client.beta.threads.messages.list(
run.thread_id, order="asc"
)
new_messages = [msg for msg in messages if msg.run_id == run.id]
if not self.as_agent:
return new_messages
answer: Any = [
msg_content for msg in new_messages for msg_content in msg.content
]
if all(
(
isinstance(content, openai.types.beta.threads.TextContentBlock)
if version_gte_1_14
else isinstance(
content, openai.types.beta.threads.MessageContentText
)
)
for content in answer
):
answer = "\n".join(content.text.value for content in answer)
return OpenAIAssistantFinish(
return_values={
"output": answer,
"thread_id": run.thread_id,
"run_id": run.id,
},
log="",
run_id=run.id,
thread_id=run.thread_id,
)
elif run.status == "requires_action":
if not self.as_agent:
return run.required_action.submit_tool_outputs.tool_calls
actions = []
for tool_call in run.required_action.submit_tool_outputs.tool_calls:
function = tool_call.function
try:
args = json.loads(function.arguments, strict=False)
except JSONDecodeError as e:
raise ValueError(
f"Received invalid JSON function arguments: "
f"{function.arguments} for function {function.name}"
) from e
if len(args) == 1 and "__arg1" in args:
args = args["__arg1"]
actions.append(
OpenAIAssistantAction(
tool=function.name,
tool_input=args,
tool_call_id=tool_call.id,
log="",
run_id=run.id,
thread_id=run.thread_id,
)
)
return actions
else:
run_info = json.dumps(run.dict(), indent=2)
raise ValueError(
f"Unexpected run status: {run.status}. Full run info:\n\n{run_info})"
)
async def _await_for_run(self, run_id: str, thread_id: str) -> Any:
in_progress = True
while in_progress:
run = await self.async_client.beta.threads.runs.retrieve(
run_id, thread_id=thread_id
)
in_progress = run.status in ("in_progress", "queued")
if in_progress:
await asyncio.sleep(self.check_every_ms / 1000)
return run
