diff --git a/docs/concepts/agents.mdx b/docs/concepts/agents.mdx
index 63c609b4..702df75f 100644
--- a/docs/concepts/agents.mdx
+++ b/docs/concepts/agents.mdx
@@ -25,7 +25,7 @@ An agent has the following key properties:
- `name` (str): The name of the agent, which serves as an identifier and is visible to other agents in the workflow.
- `description` (str, optional): A brief description of the agent's role or specialization, which is also visible to other agents.
- `instructions` (str, optional): Specific instructions or guidelines for the agent to follow during task execution. These instructions are private to the agent and not visible to other agents.
-- `tools` (List[ToolType], optional): A list of tools available to the agent. Tools are Python functions that the agent can call to perform specific actions or computations.
+- `tools` (List[Callable], optional): A list of tools available to the agent. Tools are Python functions that the agent can call to perform specific actions or computations.
- `user_access` (bool, default=False): Indicates whether the agent has access to user interactions. If set to `True`, the agent will be provided with the `talk_to_human` tool to communicate with users.
These properties help define the agent's characteristics, behavior, and capabilities within the flow.
diff --git a/docs/guides/deployment.mdx b/docs/guides/deployment.mdx
index 00183869..abb982f3 100644
--- a/docs/guides/deployment.mdx
+++ b/docs/guides/deployment.mdx
@@ -9,9 +9,8 @@ Here are recommendations for production settings:
```bash
-# set log levels to INFO for CF and Marvin
+# set log levels to INFO
CONTROLFLOW_LOGGING_LEVEL=INFO
-MARVIN_LOGGING_LEVEL=INFO
# disable local (terminal) inputs for flows that take user inputs
CONTROLFLOW_ENABLE_LOCAL_INPUT=0
diff --git a/examples/documentation.py b/examples/documentation.py
deleted file mode 100644
index 35394a7c..00000000
--- a/examples/documentation.py
+++ /dev/null
@@ -1,64 +0,0 @@
-import glob as glob_module
-from pathlib import Path
-
-import controlflow
-from controlflow import flow, task
-from marvin.beta.assistants import Assistant, Thread
-from marvin.tools.filesystem import read, write
-
-ROOT = Path(controlflow.__file__).parents[2]
-
-
-def glob(pattern: str) -> list[str]:
- """
- Returns a list of paths matching a valid glob pattern.
- The pattern can include ** for recursive matching, such as
- '~/path/to/root/**/*.py'
- """
- return glob_module.glob(pattern, recursive=True)
-
-
-assistant = Assistant(
- instructions="""
- You are an expert technical writer who writes wonderful documentation for
- open-source tools and believes that documentation is a product unto itself.
- """,
- tools=[read, write, glob],
-)
-
-
-@task(model="gpt-3.5-turbo")
-def examine_source_code(source_dir: Path, extensions: list[str]):
- """
- Load all matching files in the root dir and all subdirectories and
- read them carefully.
- """
-
-
-@task(model="gpt-3.5-turbo")
-def read_docs(docs_dir: Path):
- """
- Read all documentation in the docs dir and subdirectories, if any.
- """
-
-
-@task
-def write_docs(docs_dir: Path, instructions: str = None):
- """
- Write new documentation based on the provided instructions.
- """
-
-
-@flow(assistant=assistant)
-def docs_flow(instructions: str):
- examine_source_code(ROOT / "src", extensions=[".py"])
- # read_docs(ROOT / "docs")
- write_docs(ROOT / "docs", instructions=instructions)
-
-
-if __name__ == "__main__":
- thread = Thread()
- docs_flow(
- _thread=thread,
- instructions="Write documentation for the AI Flow class and save it in docs/flow.md",
- )
diff --git a/pyproject.toml b/pyproject.toml
index 2d2f8904..a2f583db 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -6,12 +6,13 @@ authors = [
{ name = "Jeremiah Lowin", email = "153965+jlowin@users.noreply.github.com" },
]
dependencies = [
- "marvin @ git+https://github.com/prefecthq/marvin@main",
"prefect[dev] @ git+https://github.com/prefecthq/prefect@main",
# can remove when prefect fully migrates to pydantic 2
"pydantic>=2",
"textual>=0.61.1",
"litellm>=1.37.17",
+ "jinja2>=3.1.4",
+ "pydantic-settings>=2.2.1",
]
readme = "README.md"
requires-python = ">= 3.9"
@@ -79,7 +80,7 @@ skip-magic-trailing-comma = false
"__init__.py" = ['I', 'F401', 'E402']
"conftest.py" = ["F401", "F403"]
'tests/fixtures/*.py' = ['F401', 'F403']
-"src/controlflow/utilities/types.py" = ['F401']
+# "src/controlflow/utilities/types.py" = ['F401']
[tool.pytest.ini_options]
timeout = 120
diff --git a/requirements-dev.lock b/requirements-dev.lock
index 860fae07..093abe1a 100644
--- a/requirements-dev.lock
+++ b/requirements-dev.lock
@@ -22,7 +22,6 @@ anyio==3.7.1
# via httpx
# via openai
# via prefect
- # via starlette
# via watchfiles
apprise==1.7.5
# via prefect
@@ -50,7 +49,6 @@ bytecode==0.15.1
# via ddtrace
cachetools==5.3.3
# via google-auth
- # via marvin
# via prefect
cairocffi==1.7.0
# via cairosvg
@@ -121,8 +119,6 @@ execnet==2.1.1
# via pytest-xdist
executing==2.0.1
# via stack-data
-fastapi==0.110.0
- # via marvin
filelock==3.13.3
# via huggingface-hub
# via virtualenv
@@ -154,7 +150,6 @@ httpcore==1.0.5
# via httpx
# via prefect
httpx==0.27.0
- # via marvin
# via openai
# via prefect
# via respx
@@ -187,10 +182,10 @@ itsdangerous==2.1.2
# via prefect
jedi==0.19.1
# via ipython
-jinja2==3.1.3
+jinja2==3.1.4
+ # via controlflow
# via jinja2-humanize-extension
# via litellm
- # via marvin
# via mike
# via mkdocs
# via mkdocs-material
@@ -203,7 +198,6 @@ jmespath==1.0.1
# via boto3
# via botocore
jsonpatch==1.33
- # via marvin
# via prefect
jsonpointer==2.4
# via jsonpatch
@@ -237,8 +231,6 @@ markupsafe==2.1.5
# via mkdocs-autorefs
# via mkdocstrings
# via werkzeug
-marvin @ git+https://github.com/prefecthq/marvin@9ff559318af3dd1185ab7d6a1e85b39967915b81
- # via controlflow
matplotlib-inline==0.1.6
# via ipython
mdit-py-plugins==0.4.1
@@ -289,7 +281,6 @@ oauthlib==3.2.2
# via requests-oauthlib
openai==1.28.1
# via litellm
- # via marvin
opentelemetry-api==1.24.0
# via ddtrace
orjson==3.10.0
@@ -304,8 +295,6 @@ paginate==0.5.6
# via mkdocs-material
parso==0.8.3
# via jedi
-partialjson==0.0.7
- # via marvin
pathspec==0.12.1
# via mkdocs
# via prefect
@@ -325,11 +314,10 @@ pluggy==1.4.0
# via pytest
pre-commit==3.7.0
# via prefect
-prefect @ git+https://github.com/prefecthq/prefect@aad7f63ccfc7767729f65fe20b0defb208b2c451
+prefect @ git+https://github.com/prefecthq/prefect@fa0ffc96b9740b6ab00a81b14b50b17285abdefb
# via controlflow
prompt-toolkit==3.0.43
# via ipython
- # via marvin
protobuf==5.26.1
# via ddtrace
ptyprocess==0.7.0
@@ -347,8 +335,6 @@ pycparser==2.22
# via cffi
pydantic==2.6.4
# via controlflow
- # via fastapi
- # via marvin
# via openai
# via prefect
# via pydantic-settings
@@ -356,7 +342,7 @@ pydantic-core==2.16.3
# via prefect
# via pydantic
pydantic-settings==2.2.1
- # via marvin
+ # via controlflow
pygments==2.17.2
# via ipython
# via mkdocs-material
@@ -458,7 +444,6 @@ respx==0.21.1
rfc3339-validator==0.1.4
# via prefect
rich==13.7.1
- # via marvin
# via prefect
# via textual
# via typer-slim
@@ -502,15 +487,12 @@ sqlparse==0.5.0
# via ddtrace
stack-data==0.6.3
# via ipython
-starlette==0.36.3
- # via fastapi
text-unidecode==1.3
# via python-slugify
textual==0.61.1
# via controlflow
tiktoken==0.6.0
# via litellm
- # via marvin
time-machine==2.14.1
# via pendulum
tinycss2==1.3.0
@@ -527,7 +509,6 @@ traitlets==5.14.2
# via ipython
# via matplotlib-inline
typer==0.12.0
- # via marvin
# via prefect
typer-cli==0.12.0
# via typer
@@ -542,9 +523,7 @@ typing-extensions==4.10.0
# via aiosqlite
# via alembic
# via ddtrace
- # via fastapi
# via huggingface-hub
- # via marvin
# via mypy
# via openai
# via prefect
@@ -554,7 +533,6 @@ typing-extensions==4.10.0
# via textual
# via typer-slim
tzdata==2024.1
- # via marvin
# via pendulum
tzlocal==5.2
# via dateparser
@@ -569,7 +547,6 @@ urllib3==2.2.1
# via requests
# via responses
uvicorn==0.28.1
- # via marvin
# via prefect
vermin==1.6.0
# via prefect
diff --git a/requirements.lock b/requirements.lock
index 2557fb37..eef45d9f 100644
--- a/requirements.lock
+++ b/requirements.lock
@@ -22,7 +22,6 @@ anyio==3.7.1
# via httpx
# via openai
# via prefect
- # via starlette
# via watchfiles
apprise==1.7.5
# via prefect
@@ -50,7 +49,6 @@ bytecode==0.15.1
# via ddtrace
cachetools==5.3.3
# via google-auth
- # via marvin
# via prefect
cairocffi==1.7.0
# via cairosvg
@@ -121,8 +119,6 @@ execnet==2.1.1
# via pytest-xdist
executing==2.0.1
# via stack-data
-fastapi==0.110.0
- # via marvin
filelock==3.14.0
# via huggingface-hub
# via virtualenv
@@ -154,7 +150,6 @@ httpcore==1.0.5
# via httpx
# via prefect
httpx==0.27.0
- # via marvin
# via openai
# via prefect
# via respx
@@ -187,10 +182,10 @@ itsdangerous==2.1.2
# via prefect
jedi==0.19.1
# via ipython
-jinja2==3.1.3
+jinja2==3.1.4
+ # via controlflow
# via jinja2-humanize-extension
# via litellm
- # via marvin
# via mike
# via mkdocs
# via mkdocs-material
@@ -203,7 +198,6 @@ jmespath==1.0.1
# via boto3
# via botocore
jsonpatch==1.33
- # via marvin
# via prefect
jsonpointer==2.4
# via jsonpatch
@@ -237,8 +231,6 @@ markupsafe==2.1.5
# via mkdocs-autorefs
# via mkdocstrings
# via werkzeug
-marvin @ git+https://github.com/prefecthq/marvin@9ff559318af3dd1185ab7d6a1e85b39967915b81
- # via controlflow
matplotlib-inline==0.1.7
# via ipython
mdit-py-plugins==0.4.1
@@ -289,7 +281,6 @@ oauthlib==3.2.2
# via requests-oauthlib
openai==1.28.1
# via litellm
- # via marvin
opentelemetry-api==1.24.0
# via ddtrace
orjson==3.10.0
@@ -304,8 +295,6 @@ paginate==0.5.6
# via mkdocs-material
parso==0.8.4
# via jedi
-partialjson==0.0.7
- # via marvin
pathspec==0.12.1
# via mkdocs
# via prefect
@@ -325,11 +314,10 @@ pluggy==1.5.0
# via pytest
pre-commit==3.7.1
# via prefect
-prefect @ git+https://github.com/prefecthq/prefect@aad7f63ccfc7767729f65fe20b0defb208b2c451
+prefect @ git+https://github.com/prefecthq/prefect@fa0ffc96b9740b6ab00a81b14b50b17285abdefb
# via controlflow
prompt-toolkit==3.0.43
# via ipython
- # via marvin
protobuf==5.26.1
# via ddtrace
ptyprocess==0.7.0
@@ -347,8 +335,6 @@ pycparser==2.22
# via cffi
pydantic==2.6.4
# via controlflow
- # via fastapi
- # via marvin
# via openai
# via prefect
# via pydantic-settings
@@ -356,7 +342,7 @@ pydantic-core==2.16.3
# via prefect
# via pydantic
pydantic-settings==2.2.1
- # via marvin
+ # via controlflow
pygments==2.17.2
# via ipython
# via mkdocs-material
@@ -458,7 +444,6 @@ respx==0.21.1
rfc3339-validator==0.1.4
# via prefect
rich==13.7.1
- # via marvin
# via prefect
# via textual
# via typer-slim
@@ -502,15 +487,12 @@ sqlparse==0.5.0
# via ddtrace
stack-data==0.6.3
# via ipython
-starlette==0.36.3
- # via fastapi
text-unidecode==1.3
# via python-slugify
textual==0.61.1
# via controlflow
tiktoken==0.6.0
# via litellm
- # via marvin
time-machine==2.14.1
# via pendulum
tinycss2==1.3.0
@@ -527,7 +509,6 @@ traitlets==5.14.3
# via ipython
# via matplotlib-inline
typer==0.12.0
- # via marvin
# via prefect
typer-cli==0.12.0
# via typer
@@ -542,9 +523,7 @@ typing-extensions==4.10.0
# via aiosqlite
# via alembic
# via ddtrace
- # via fastapi
# via huggingface-hub
- # via marvin
# via mypy
# via openai
# via prefect
@@ -554,7 +533,6 @@ typing-extensions==4.10.0
# via textual
# via typer-slim
tzdata==2024.1
- # via marvin
# via pendulum
tzlocal==5.2
# via dateparser
@@ -569,7 +547,6 @@ urllib3==2.2.1
# via requests
# via responses
uvicorn==0.28.1
- # via marvin
# via prefect
vermin==1.6.0
# via prefect
diff --git a/src/controlflow/__init__.py b/src/controlflow/__init__.py
index ef7533c1..a9fbaff7 100644
--- a/src/controlflow/__init__.py
+++ b/src/controlflow/__init__.py
@@ -1,5 +1,7 @@
from .settings import settings
+from . import llm
+
from .core.flow import Flow
from .core.task import Task
from .core.agent import Agent
diff --git a/src/controlflow/agents/agents.py b/src/controlflow/agents/agents.py
deleted file mode 100644
index 20c297f3..00000000
--- a/src/controlflow/agents/agents.py
+++ /dev/null
@@ -1,42 +0,0 @@
-import marvin
-
-from controlflow.core.agent import Agent
-from controlflow.instructions import get_instructions
-from controlflow.utilities.context import ctx
-from controlflow.utilities.threads import get_history
-
-
-def choose_agent(
- agents: list[Agent],
- instructions: str = None,
- context: dict = None,
- model: str = None,
-) -> Agent:
- """
- Given a list of potential agents, choose the most qualified assistant to complete the tasks.
- """
-
- instructions = get_instructions()
- history = []
- if (flow := ctx.get("flow")) and flow.thread.id:
- history = get_history(thread_id=flow.thread.id)
-
- info = dict(
- history=history,
- global_instructions=instructions,
- context=context,
- )
-
- agent = marvin.classify(
- info,
- agents,
- instructions="""
- Given the conversation context, choose the AI agent most
- qualified to take the next turn at completing the tasks. Take into
- account the instructions, each agent's own instructions, and the
- tools they have available.
- """,
- model_kwargs=dict(model=model),
- )
-
- return agent
diff --git a/src/controlflow/core/agent.py b/src/controlflow/core/agent.py
index 57ab3364..b26663c0 100644
--- a/src/controlflow/core/agent.py
+++ b/src/controlflow/core/agent.py
@@ -1,17 +1,12 @@
import logging
-from typing import Callable, Optional, Union
+from typing import Callable, Optional
-from marvin.utilities.asyncio import ExposeSyncMethodsMixin, expose_sync_method
from pydantic import Field
import controlflow
-from controlflow.core.flow import Flow, get_flow
-from controlflow.core.task import Task
from controlflow.tools.talk_to_human import talk_to_human
-from controlflow.utilities.prefect import (
- wrap_prefect_tool,
-)
-from controlflow.utilities.types import Assistant, ControlFlowModel, ToolType
+from controlflow.utilities.asyncio import ExposeSyncMethodsMixin
+from controlflow.utilities.types import ControlFlowModel
logger = logging.getLogger(__name__)
@@ -20,42 +15,6 @@ def get_default_agent() -> "Agent":
return controlflow.default_agent
-class AgentOLD(Assistant, ControlFlowModel, ExposeSyncMethodsMixin):
- name: str
- user_access: bool = Field(
- False,
- description="If True, the agent is given tools for interacting with a human user.",
- )
-
- def get_tools(self) -> list[ToolType]:
- tools = super().get_tools()
- if self.user_access:
- tools.append(talk_to_human)
-
- return [wrap_prefect_tool(tool) for tool in tools]
-
- @expose_sync_method("run")
- async def run_async(
- self,
- tasks: Union[list[Task], Task, None] = None,
- flow: Flow = None,
- ):
- from controlflow.core.controller import Controller
-
- if isinstance(tasks, Task):
- tasks = [tasks]
-
- flow = flow or get_flow()
-
- if not flow:
- raise ValueError(
- "Agents must be run within a flow context or with a flow argument."
- )
-
- controller = Controller(agents=[self], tasks=tasks or [], flow=flow)
- return await controller.run_agent_async(agent=self)
-
-
class Agent(ControlFlowModel, ExposeSyncMethodsMixin):
name: str = Field(
...,
@@ -88,34 +47,6 @@ def get_tools(self) -> list[Callable]:
tools.append(talk_to_human)
return tools
- # def say(
- # self, messages: Union[str, dict], thread_id: str = None, history: History = None
- # ) -> Response:
-
- # if thread_id is None:
- # thread_id = self.default_thread_id
- # if history is None:
- # history = get_default_history()
- # if not isinstance(messages, list):
- # raise ValueError("Messages must be provided as a list.")
-
- # messages = [
- # Message(role="user", content=m) if isinstance(m, str) else m
- # for m in messages
- # ]
- # history_messages = history.load_messages(thread_id=thread_id, limit=50)
-
- # response = completion(
- # messages=history_messages + messages,
- # model=self.model,
- # tools=self.tools,
- # )
- # history.save_messages(
- # thread_id=thread_id,
- # messages=messages + history_messages + response.messages,
- # )
- # return response
-
DEFAULT_AGENT = Agent(
name="Marvin",
diff --git a/src/controlflow/core/controller/controller.py b/src/controlflow/core/controller/controller.py
index d5755070..ccb5b4ad 100644
--- a/src/controlflow/core/controller/controller.py
+++ b/src/controlflow/core/controller/controller.py
@@ -3,14 +3,13 @@
from collections import defaultdict
from contextlib import asynccontextmanager
from functools import cached_property
-from typing import Union
+from typing import Callable, Union
-from marvin.utilities.asyncio import ExposeSyncMethodsMixin, expose_sync_method
from pydantic import BaseModel, Field, PrivateAttr, computed_field, model_validator
import controlflow
from controlflow.core.agent import Agent
-from controlflow.core.controller.moderators import marvin_moderator
+from controlflow.core.controller.moderators import classify_moderator
from controlflow.core.flow import Flow, get_flow
from controlflow.core.graph import Graph
from controlflow.core.task import Task
@@ -20,9 +19,9 @@
from controlflow.llm.history import History
from controlflow.llm.messages import AssistantMessage, ControlFlowMessage, SystemMessage
from controlflow.tui.app import TUIApp as TUI
+from controlflow.utilities.asyncio import ExposeSyncMethodsMixin, expose_sync_method
from controlflow.utilities.context import ctx
from controlflow.utilities.tasks import all_complete, any_incomplete
-from controlflow.utilities.types import FunctionTool
logger = logging.getLogger(__name__)
@@ -87,7 +86,7 @@ def _finalize(self):
self.flow.add_task(task)
return self
- def _create_end_turn_tool(self) -> FunctionTool:
+ def _create_end_turn_tool(self) -> Callable:
def end_turn():
"""
Call this tool to skip your turn and let another agent go next. This
@@ -96,13 +95,17 @@ def end_turn():
automatically, so only use it if you are truly stuck and unable to
proceed.
"""
- self._end_run_counts[ctx.get("controller_agent")] += 1
- if self._end_run_counts[ctx.get("controller_agent")] >= 3:
+
+ # the agent's name is used as the key to track the number of times
+ key = getattr(ctx.get("controller_agent", None), "name", None)
+
+ self._end_run_counts[key] += 1
+ if self._end_run_counts[key] >= 3:
self._should_abort = True
- self._end_run_counts[ctx.get("controller_agent")] = 0
+ self._end_run_counts[key] = 0
return (
- f"Ending turn. {3 - self._end_run_counts[ctx.get('controller_agent')]}"
+ f"Ending turn. {3 - self._end_run_counts[key]}"
" more uses will abort the workflow."
)
@@ -151,6 +154,7 @@ async def _run_agent(self, agent: Agent, tasks: list[Task]):
tools=tools,
handlers=handlers,
max_iterations=1,
+ assistant_name=agent.name,
stream=True,
message_preprocessor=add_agent_name_to_message,
):
@@ -173,7 +177,7 @@ async def _run_agent(self, agent: Agent, tasks: list[Task]):
# )
def choose_agent(self, agents: list[Agent], tasks: list[Task]) -> Agent:
- return marvin_moderator(
+ return classify_moderator(
agents=agents,
tasks=tasks,
iteration=self._iteration,
@@ -183,11 +187,13 @@ def choose_agent(self, agents: list[Agent], tasks: list[Task]) -> Agent:
async def tui(self):
if tui := ctx.get("tui"):
yield tui
- else:
+ elif controlflow.settings.enable_tui:
tui = TUI(flow=self.flow)
with ctx(tui=tui):
- async with tui.run_context(run=controlflow.settings.enable_tui):
+ async with tui.run_context():
yield tui
+ else:
+ yield
@expose_sync_method("run_once")
async def run_once_async(self):
@@ -228,7 +234,6 @@ async def run_once_async(self):
elif len(agents) == 1:
agent = agents[0]
else:
- raise NotImplementedError("Need to reimplement multi-agent")
agent = self.choose_agent(agents=agents, tasks=tasks)
with ctx(controller_agent=agent):
diff --git a/src/controlflow/core/controller/instruction_template.py b/src/controlflow/core/controller/instruction_template.py
index 0ced8893..b35d847b 100644
--- a/src/controlflow/core/controller/instruction_template.py
+++ b/src/controlflow/core/controller/instruction_template.py
@@ -19,7 +19,9 @@ def render(self) -> str:
if self.should_render():
render_kwargs = dict(self)
render_kwargs.pop("template")
- return jinja_env.render(inspect.cleandoc(self.template), **render_kwargs)
+ return jinja_env.from_string(inspect.cleandoc(self.template)).render(
+ **render_kwargs
+ )
class AgentTemplate(Template):
@@ -131,10 +133,14 @@ class CommunicationTemplate(Template):
messages unless you need to record information in addition to what you
provide as a task's result, or for the following reasons:
- - You need to post a message or otherwise communicate to complete a task.
+ - You need to post a message or otherwise communicate to complete a
+ task. For example, the task instructs you to write, discuss, or
+ otherwise produce content (and does not accept a result, or the result
+ that meets the objective is different than the instructed actions).
- You need to communicate with other agents to complete a task.
- You want to write your thought process for future reference.
+
Note that You may see other agents post messages; they may have
different instructions than you do, so do not follow their example
automatically.
diff --git a/src/controlflow/core/controller/moderators.py b/src/controlflow/core/controller/moderators.py
index 195b0b0f..ae264e50 100644
--- a/src/controlflow/core/controller/moderators.py
+++ b/src/controlflow/core/controller/moderators.py
@@ -1,9 +1,8 @@
-import marvin
-
from controlflow.core.agent import Agent
from controlflow.core.flow import get_flow_messages
from controlflow.core.task import Task
from controlflow.instructions import get_instructions
+from controlflow.llm.classify import classify
def round_robin(
@@ -15,7 +14,7 @@ def round_robin(
return agents[iteration % len(agents)]
-def marvin_moderator(
+def classify_moderator(
agents: list[Agent],
tasks: list[Task],
context: dict = None,
@@ -26,9 +25,9 @@ def marvin_moderator(
instructions = get_instructions()
context = context or {}
context.update(tasks=tasks, history=history, instructions=instructions)
- agent = marvin.classify(
+ agent = classify(
context,
- agents,
+ labels=agents,
instructions="""
Given the context, choose the AI agent best suited to take the
next turn at completing the tasks in the task graph. Take into account
@@ -37,6 +36,6 @@ def marvin_moderator(
completed before their downstream/parents can be completed. An agent
can only work on a task that it is assigned to.
""",
- model_kwargs=dict(model=model) if model else None,
+ model=model,
)
return agent
diff --git a/src/controlflow/core/flow.py b/src/controlflow/core/flow.py
index c81d120c..c8897d51 100644
--- a/src/controlflow/core/flow.py
+++ b/src/controlflow/core/flow.py
@@ -4,6 +4,7 @@
from pydantic import Field
+from controlflow.llm.history import get_default_history
from controlflow.llm.messages import MessageType
from controlflow.utilities.context import ctx
from controlflow.utilities.logging import get_logger
@@ -82,6 +83,6 @@ def get_flow_messages(limit: int = None) -> list[MessageType]:
"""
flow = get_flow()
if flow:
- return flow.thread.get_messages(limit=limit)
+ get_default_history().load_messages(flow.thread_id, limit=limit)
else:
return []
diff --git a/src/controlflow/core/task.py b/src/controlflow/core/task.py
index 65a6dd1d..1beac2e0 100644
--- a/src/controlflow/core/task.py
+++ b/src/controlflow/core/task.py
@@ -39,7 +39,6 @@
ControlFlowModel,
PandasDataFrame,
PandasSeries,
- ToolType,
)
if TYPE_CHECKING:
@@ -219,7 +218,7 @@ def _serialize_agents(self, agents: Optional[list["Agent"]]):
]
@field_serializer("tools")
- def _serialize_tools(self, tools: list[ToolType]):
+ def _serialize_tools(self, tools: list[Callable]):
tools = controlflow.llm.tools.as_tools(tools)
return [t.model_dump({"name", "description"}) for t in tools]
diff --git a/src/controlflow/decorators.py b/src/controlflow/decorators.py
index b8aa62fa..faa659a6 100644
--- a/src/controlflow/decorators.py
+++ b/src/controlflow/decorators.py
@@ -1,17 +1,17 @@
import functools
import inspect
+from typing import Callable
import prefect
-from marvin.beta.assistants import Thread
import controlflow
from controlflow.core.agent import Agent
from controlflow.core.flow import Flow
from controlflow.core.task import Task
from controlflow.utilities.logging import get_logger
-from controlflow.utilities.marvin import patch_marvin
+
+# from controlflow.utilities.marvin import patch_marvin
from controlflow.utilities.tasks import resolve_tasks
-from controlflow.utilities.types import ToolType
logger = get_logger(__name__)
@@ -19,9 +19,9 @@
def flow(
fn=None,
*,
- thread: Thread = None,
+ thread: str = None,
instructions: str = None,
- tools: list[ToolType] = None,
+ tools: list[Callable] = None,
agents: list["Agent"] = None,
lazy: bool = None,
):
@@ -38,9 +38,9 @@ def flow(
Args:
fn (callable, optional): The function to be wrapped as a flow. If not provided,
the decorator will act as a partial function and return a new flow decorator.
- thread (Thread, optional): The thread to execute the flow on. Defaults to None.
+ thread (str, optional): The thread to execute the flow on. Defaults to None.
instructions (str, optional): Instructions for the flow. Defaults to None.
- tools (list[ToolType], optional): List of tools to be used in the flow. Defaults to None.
+ tools (list[Callable], optional): List of tools to be used in the flow. Defaults to None.
agents (list[Agent], optional): List of agents to be used in the flow. Defaults to None.
lazy (bool, optional): Whether the flow should be run lazily. If not
set, behavior is determined by the global `eager_mode` setting. Lazy execution means
@@ -92,7 +92,7 @@ def wrapper(
# create a function to wrap as a Prefect flow
@prefect.flow
def wrapped_flow(*args, lazy_=None, **kwargs):
- with flow_obj, patch_marvin():
+ with flow_obj:
with controlflow.instructions(instructions):
result = fn(*args, **kwargs)
@@ -126,7 +126,7 @@ def task(
objective: str = None,
instructions: str = None,
agents: list["Agent"] = None,
- tools: list[ToolType] = None,
+ tools: list[Callable] = None,
user_access: bool = None,
lazy: bool = None,
):
@@ -147,7 +147,7 @@ def task(
instructions (str, optional): Instructions for the task. Defaults to None, in which
case the function docstring is used as the instructions.
agents (list[Agent], optional): List of agents to be used in the task. Defaults to None.
- tools (list[ToolType], optional): List of tools to be used in the task. Defaults to None.
+ tools (list[Callable], optional): List of tools to be used in the task. Defaults to None.
user_access (bool, optional): Whether the task requires user access. Defaults to None,
in which case it is set to False.
lazy (bool, optional): Whether the task should be run lazily. If not
diff --git a/src/controlflow/llm/classify.py b/src/controlflow/llm/classify.py
new file mode 100644
index 00000000..a92ba282
--- /dev/null
+++ b/src/controlflow/llm/classify.py
@@ -0,0 +1,90 @@
+from pydantic import TypeAdapter
+
+import controlflow
+from controlflow.llm.messages import AssistantMessage, SystemMessage, UserMessage
+
+
+def classify(
+ data: str,
+ labels: list,
+ instructions: str = None,
+ context: dict = None,
+ model: str = None,
+):
+ try:
+ label_strings = [TypeAdapter(type(t)).dump_json(t).decode() for t in labels]
+ except Exception as exc:
+ raise ValueError(f"Unable to cast labels to strings: {exc}")
+
+ messages = [
+ SystemMessage(
+ """
+ You are an expert classifier that always maintains as much semantic meaning
+ as possible when labeling information. You use inference or deduction whenever
+ necessary to understand missing or omitted data. Classify the provided data,
+ text, or information as one of the provided labels. For boolean labels,
+ consider "truthy" or affirmative inputs to be "true".
+
+ ## Labels
+
+ You must classify the data as one of the following labels, which are
+ numbered (starting from 0) and provide a brief description. Output
+ the label number only.
+
+ {% for label in labels %}
+ - Label #{{ loop.index0 }}: {{ label }}
+ {% endfor %}
+ """
+ ).render(labels=label_strings),
+ UserMessage(
+ """
+ ## Information to classify
+
+ {{ data }}
+
+ {% if instructions -%}
+ ## Additional instructions
+
+ {{ instructions }}
+ {% endif %}
+
+ {% if context -%}
+ ## Additional context
+
+ {% for key, value in context.items() -%}
+ - {{ key }}: {{ value }}
+
+ {% endfor %}
+ {% endif %}
+
+ """
+ ).render(data=data, instructions=instructions, context=context),
+ AssistantMessage("""
+ The best label for the data is Label #
+ """),
+ ]
+
+ result = controlflow.llm.completions.completion(
+ messages=messages,
+ model=model,
+ max_tokens=1,
+ logit_bias={
+ str(encoding): 100
+ for i in range(len(labels))
+ for encoding in _encoder(model)(str(i))
+ },
+ )
+
+ index = int(result[0].content)
+ return labels[index]
+
+
+def _encoder(model: str):
+ import tiktoken
+
+ try:
+ encoding = tiktoken.encoding_for_model(model)
+ except (KeyError, AttributeError):
+ encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")
+
+ return encoding.encode
diff --git a/src/controlflow/llm/completions.py b/src/controlflow/llm/completions.py
index 48f206fc..c30b91ad 100644
--- a/src/controlflow/llm/completions.py
+++ b/src/controlflow/llm/completions.py
@@ -20,10 +20,10 @@
def completion(
messages: list[Union[dict, ControlFlowMessage]],
- model=None,
+ model: str = None,
tools: list[Callable] = None,
assistant_name: str = None,
- max_iterations=None,
+ max_iterations: int = None,
handlers: list[CompletionHandler] = None,
message_preprocessor: Callable[[ControlFlowMessage], ControlFlowMessage] = None,
stream: bool = False,
@@ -124,7 +124,7 @@ def completion(
def _completion_stream(
messages: list[Union[dict, ControlFlowMessage]],
- model=None,
+ model: str = None,
tools: list[Callable] = None,
assistant_name: str = None,
max_iterations: int = None,
@@ -243,10 +243,10 @@ def _completion_stream(
async def completion_async(
messages: list[Union[dict, ControlFlowMessage]],
- model=None,
+ model: str = None,
tools: list[Callable] = None,
assistant_name: str = None,
- max_iterations=None,
+ max_iterations: int = None,
handlers: list[CompletionHandler] = None,
message_preprocessor: Callable[[ControlFlowMessage], ControlFlowMessage] = None,
stream: bool = False,
@@ -344,30 +344,9 @@ async def completion_async(
handler.on_end()
-"""
-Perform asynchronous streaming completion using the LLM model.
-
-Args:
- messages: A list of messages to be used for completion.
- model: The LLM model to be used for completion. If not provided, the default model from controlflow.settings will be used.
- tools: A list of callable tools to be used during completion.
- assistant_name: The name of the assistant, which will be set as the `name` attribute of any messages it generates.
- max_iterations: The maximum number of iterations to perform completion. If not provided, it will continue until completion is done.
- handlers: A list of CompletionHandler objects to handle completion events.
- message_preprocessor: A callable function to preprocess each ControlFlowMessage before completion.
- **kwargs: Additional keyword arguments to be passed to the litellm.acompletion function.
-
-Yields:
- Each ControlFlowMessage generated during completion.
-
-Returns:
- The final completion response as a list of ControlFlowMessage objects.
-"""
-
-
async def _completion_stream_async(
messages: list[Union[dict, ControlFlowMessage]],
- model=None,
+ model: str = None,
tools: list[Callable] = None,
assistant_name: str = None,
max_iterations: int = None,
@@ -383,10 +362,13 @@ async def _completion_stream_async(
model: The LLM model to be used for completion. If not provided, the default model from controlflow.settings will be used.
tools: A list of callable tools to be used during completion.
assistant_name: The name of the assistant, which will be set as the `name` attribute of any messages it generates.
+ max_iterations: The maximum number of iterations to perform completion. If not provided, it will continue until completion is done.
+ handlers: A list of CompletionHandler objects to handle completion events.
+ message_preprocessor: A callable function to preprocess each ControlFlowMessage before completion.
**kwargs: Additional keyword arguments to be passed to the litellm.acompletion function.
Yields:
- Each message
+ Each ControlFlowMessage generated during completion.
Returns:
The final completion response as a list of ControlFlowMessage objects.
@@ -476,5 +458,6 @@ async def _completion_stream_async(
break
except Exception as exc:
handler.on_exception(exc)
+ raise
finally:
handler.on_end()
diff --git a/src/controlflow/llm/formatting.py b/src/controlflow/llm/formatting.py
index e1843c98..dc6d7bea 100644
--- a/src/controlflow/llm/formatting.py
+++ b/src/controlflow/llm/formatting.py
@@ -41,9 +41,14 @@ def format_message(
def format_text_message(message: MessageType) -> Panel:
+ if message.role == "assistant" and message.name:
+ title = f"Agent: {message.name}"
+ else:
+ title = ROLE_NAMES.get(message.role, "Agent")
+
return Panel(
Markdown(message.content or ""),
- title=f"[bold]{ROLE_NAMES.get(message.role, 'Agent')}[/]",
+ title=f"[bold]{title}[/]",
subtitle=f"[italic]{format_timestamp(message.timestamp)}[/]",
title_align="left",
subtitle_align="right",
diff --git a/src/controlflow/llm/messages.py b/src/controlflow/llm/messages.py
index 50cd7c21..91e2928e 100644
--- a/src/controlflow/llm/messages.py
+++ b/src/controlflow/llm/messages.py
@@ -1,4 +1,5 @@
import datetime
+import inspect
import json
import uuid
from typing import Any, List, Literal, Optional, Union
@@ -12,6 +13,7 @@
model_validator,
)
+from controlflow.utilities.jinja import jinja_env
from controlflow.utilities.types import _OpenAIBaseType
# -----------------------------------------------
@@ -67,7 +69,7 @@ def _lowercase_role(cls, v):
@field_validator("timestamp", mode="before")
def _validate_timestamp(cls, v):
if isinstance(v, int):
- v = datetime.datetime.fromtimestamp(v)
+ v = datetime.datetime.fromtimestamp(v, tz=datetime.timezone.utc)
return v
@model_validator(mode="after")
@@ -94,6 +96,20 @@ class SystemMessage(ControlFlowMessage):
_openai_fields = {"role", "content", "name"}
# ---- end openai fields
+ @field_validator("content", mode="before")
+ def _validate_content(cls, v):
+ if isinstance(v, str):
+ v = inspect.cleandoc(v)
+ return v
+
+ def render(self, **kwargs) -> "SystemMessage":
+ """
+ Renders the content as a jinja template with the given keyword arguments
+ and returns a new SystemMessage.
+ """
+ content = jinja_env.from_string(self.content).render(**kwargs)
+ return self.model_copy(update=dict(content=content))
+
class UserMessage(ControlFlowMessage):
# ---- begin openai fields
@@ -106,9 +122,24 @@ class UserMessage(ControlFlowMessage):
@field_validator("content", mode="before")
def _validate_content(cls, v):
if isinstance(v, str):
+ v = inspect.cleandoc(v)
v = [TextContent(text=v)]
return v
+ def render(self, **kwargs) -> "UserMessage":
+ """
+ Renders the content as a jinja template with the given keyword arguments
+ and returns a new SystemMessage.
+ """
+ content = []
+ for c in self.content:
+ if isinstance(c, TextContent):
+ text = jinja_env.from_string(c.text).render(**kwargs)
+ content.append(TextContent(text=text))
+ else:
+ content.append(c)
+ return self.model_copy(update=dict(content=content))
+
class AssistantMessage(ControlFlowMessage):
"""A message from the assistant."""
@@ -126,9 +157,26 @@ class AssistantMessage(ControlFlowMessage):
description="If True, this message is a streamed delta, or chunk, of a full message.",
)
+ @field_validator("content", mode="before")
+ def _validate_content(cls, v):
+ if isinstance(v, str):
+ v = inspect.cleandoc(v)
+ return v
+
def has_tool_calls(self):
return bool(self.tool_calls)
+ def render(self, **kwargs) -> "AssistantMessage":
+ """
+ Renders the content as a jinja template with the given keyword arguments
+ and returns a new AssistantMessage.
+ """
+ if self.content is None:
+ content = self.content
+ else:
+ content = jinja_env.from_string(self.content).render(**kwargs)
+ return self.model_copy(update=dict(content=content))
+
class ToolMessage(ControlFlowMessage):
"""A message for reporting the result of a tool call."""
@@ -144,6 +192,12 @@ class ToolMessage(ControlFlowMessage):
tool_result: Any = Field(None, exclude=True)
tool_metadata: dict = Field(default_factory=dict)
+ @field_validator("content", mode="before")
+ def _validate_content(cls, v):
+ if isinstance(v, str):
+ v = inspect.cleandoc(v)
+ return v
+
MessageType = Union[SystemMessage, UserMessage, AssistantMessage, ToolMessage]
diff --git a/src/controlflow/settings.py b/src/controlflow/settings.py
index cc823936..2ec6bc75 100644
--- a/src/controlflow/settings.py
+++ b/src/controlflow/settings.py
@@ -60,7 +60,6 @@ class Settings(ControlFlowSettings):
"on a task.",
)
prefect: PrefectSettings = Field(default_factory=PrefectSettings)
- openai_api_key: Optional[str] = Field(None, validate_assignment=True)
# ------------ home settings ------------
@@ -119,14 +118,6 @@ def __init__(self, **data):
super().__init__(**data)
self.prefect.apply()
- @field_validator("openai_api_key", mode="after")
- def _apply_api_key(cls, v):
- if v is not None:
- import marvin
-
- marvin.settings.openai.api_key = v
- return v
-
@field_validator("home_path", mode="before")
def _validate_home_path(cls, v):
v = Path(v).expanduser()
diff --git a/src/controlflow/tasks/auto_tasks.py b/src/controlflow/tasks/auto_tasks.py
index 7157598b..5c57797a 100644
--- a/src/controlflow/tasks/auto_tasks.py
+++ b/src/controlflow/tasks/auto_tasks.py
@@ -5,7 +5,7 @@
from controlflow.core.agent import Agent
from controlflow.core.task import Task
-from controlflow.utilities.types import AssistantTool, ControlFlowModel, FunctionTool
+from controlflow.utilities.types import AssistantTool, ControlFlowModel
ToolLiteral = TypeVar("ToolLiteral", bound=str)
@@ -176,12 +176,7 @@ def auto_tasks(
) -> list[Task]:
tool_names = []
for tool in available_tools or []:
- if isinstance(tool, FunctionTool):
- tool_names.append(tool.function.name)
- elif isinstance(tool, AssistantTool):
- tool_names.append(tool.type)
- else:
- tool_names.append(tool.__name__)
+ tool_names.append(tool.__name__)
if tool_names:
literal_tool_names = Literal[*tool_names] # type: ignore
diff --git a/src/controlflow/tools/talk_to_human.py b/src/controlflow/tools/talk_to_human.py
index 4455137c..cd58a367 100644
--- a/src/controlflow/tools/talk_to_human.py
+++ b/src/controlflow/tools/talk_to_human.py
@@ -2,11 +2,11 @@
import contextlib
from typing import TYPE_CHECKING
-from marvin.utilities.tools import tool_from_function
from prefect.context import FlowRunContext
from prefect.input.run_input import receive_input
import controlflow
+from controlflow.llm.tools import tool
from controlflow.utilities.context import ctx
if TYPE_CHECKING:
@@ -36,7 +36,7 @@ async def listen_for_response():
return response
-@tool_from_function
+@tool
async def talk_to_human(message: str, get_response: bool = True) -> str:
"""
Send a message to the human user and optionally wait for a response.
diff --git a/src/controlflow/tui/test2.py b/src/controlflow/tui/test2.py
index f78ae9fa..378af6b8 100644
--- a/src/controlflow/tui/test2.py
+++ b/src/controlflow/tui/test2.py
@@ -1,12 +1,9 @@
import asyncio
-from marvin.utilities.asyncio import run_sync
-
from controlflow import Task
from controlflow.core.flow import Flow
from controlflow.tui.app import TUIApp
-run_sync
asyncio
with Flow() as flow:
t = Task("get the user name", user_access=True)
diff --git a/src/controlflow/utilities/asyncio.py b/src/controlflow/utilities/asyncio.py
new file mode 100644
index 00000000..63d7e623
--- /dev/null
+++ b/src/controlflow/utilities/asyncio.py
@@ -0,0 +1,94 @@
+import asyncio
+import functools
+from typing import Any, Callable, Coroutine, TypeVar, cast
+
+from prefect.utilities.asyncutils import run_sync
+
+T = TypeVar("T")
+
+BACKGROUND_TASKS = set()
+
+
+def create_task(coro):
+ """
+ Creates async background tasks in a way that is safe from garbage
+ collection.
+
+ See
+ https://textual.textualize.io/blog/2023/02/11/the-heisenbug-lurking-in-your-async-code/
+
+ Example:
+
+ async def my_coro(x: int) -> int:
+ return x + 1
+
+ # safely submits my_coro for background execution
+ create_task(my_coro(1))
+ """ # noqa: E501
+ task = asyncio.create_task(coro)
+ BACKGROUND_TASKS.add(task)
+ task.add_done_callback(BACKGROUND_TASKS.discard)
+ return task
+
+
+class ExposeSyncMethodsMixin:
+ """
+ A mixin that can take functions decorated with `expose_sync_method`
+ and automatically create synchronous versions.
+ """
+
+ def __init_subclass__(cls, **kwargs: Any) -> None:
+ super().__init_subclass__(**kwargs)
+ for method in list(cls.__dict__.values()):
+ if callable(method) and hasattr(method, "_sync_name"):
+ sync_method_name = method._sync_name
+ setattr(cls, sync_method_name, method._sync_wrapper)
+
+
+def expose_sync_method(name: str) -> Callable[..., Any]:
+ """
+ Decorator that automatically exposes synchronous versions of async methods.
+ Note it doesn't work with classmethods.
+
+ Args:
+ name: The name of the synchronous method.
+
+ Returns:
+ The decorated function.
+
+ Example:
+ Basic usage:
+ ```python
+ class MyClass(ExposeSyncMethodsMixin):
+
+ @expose_sync_method("my_method")
+ async def my_method_async(self):
+ return 42
+
+ my_instance = MyClass()
+ await my_instance.my_method_async() # returns 42
+ my_instance.my_method() # returns 42
+ ```
+ """
+
+ def decorator(
+ async_method: Callable[..., Coroutine[Any, Any, T]],
+ ) -> Callable[..., Coroutine[Any, Any, T]]:
+ @functools.wraps(async_method)
+ def sync_wrapper(*args: Any, **kwargs: Any) -> T:
+ coro = async_method(*args, **kwargs)
+ return run_sync(coro)
+
+ # Cast the sync_wrapper to the same type as the async_method to give the
+ # type checker the needed information.
+ casted_sync_wrapper = cast(Callable[..., T], sync_wrapper)
+
+ # Attach attributes to the async wrapper
+ setattr(async_method, "_sync_wrapper", casted_sync_wrapper)
+ setattr(async_method, "_sync_name", name)
+
+ # return the original async method; the sync wrapper will be added to
+ # the class by the init hook
+ return async_method
+
+ return decorator
diff --git a/src/controlflow/utilities/context.py b/src/controlflow/utilities/context.py
index 007f78b1..4640002b 100644
--- a/src/controlflow/utilities/context.py
+++ b/src/controlflow/utilities/context.py
@@ -1,4 +1,73 @@
-from marvin.utilities.context import ScopedContext
+"""Module for defining context utilities."""
+
+import contextvars
+from contextlib import contextmanager
+from typing import Any, Generator
+
+
+class ScopedContext:
+ """
+ `ScopedContext` provides a context management mechanism using `contextvars`.
+
+ This class allows setting and retrieving key-value pairs in a scoped context,
+ which is preserved across asynchronous tasks and threads within the same context.
+
+ Attributes:
+ _context_storage (ContextVar): A context variable to store the context data.
+
+ Example:
+ Basic Usage of ScopedContext
+ ```python
+ context = ScopedContext()
+ with context(key="value"):
+ assert context.get("key") == "value"
+ # Outside the context, the value is no longer available.
+ assert context.get("key") is None
+ ```
+ """
+
+ def __init__(self, initial_value: dict = None):
+ """Initializes the ScopedContext with an initial valuedictionary."""
+ self._context_storage = contextvars.ContextVar(
+ "scoped_context_storage", default=initial_value or {}
+ )
+
+ def get(self, key: str, default: Any = None) -> Any:
+ return self._context_storage.get().get(key, default)
+
+ def __getitem__(self, key: str) -> Any:
+ notfound = object()
+ result = self.get(key, default=notfound)
+ if result == notfound:
+ raise KeyError(key)
+ return result
+
+ def set(self, **kwargs: Any) -> None:
+ ctx = self._context_storage.get()
+ updated_ctx = {**ctx, **kwargs}
+ token = self._context_storage.set(updated_ctx)
+ return token
+
+ @contextmanager
+ def __call__(self, **kwargs: Any) -> Generator[None, None, Any]:
+ current_context_copy = self._context_storage.get().copy()
+ token = self.set(**kwargs)
+ try:
+ yield
+ finally:
+ try:
+ self._context_storage.reset(token)
+ except ValueError as exc:
+ if "was created in a different context" in str(exc).lower():
+ # the only way we can reach this line is if the setup and
+ # teardown of this context are run in different frames or
+ # threads (which happens with pytest fixtures!), in which case
+ # the token is considered invalid. This catch serves as a
+ # "manual" reset of the context values
+ self._context_storage.set(current_context_copy)
+ else:
+ raise
+
ctx = ScopedContext(
dict(
diff --git a/src/controlflow/utilities/jinja.py b/src/controlflow/utilities/jinja.py
index 92d4ca71..30649c59 100644
--- a/src/controlflow/utilities/jinja.py
+++ b/src/controlflow/utilities/jinja.py
@@ -1,13 +1,23 @@
import inspect
+import os
from datetime import datetime
from zoneinfo import ZoneInfo
-from marvin.utilities.jinja import BaseEnvironment
+from jinja2 import Environment as JinjaEnvironment
+from jinja2 import StrictUndefined, select_autoescape
-jinja_env = BaseEnvironment(
- globals={
+jinja_env = JinjaEnvironment(
+ autoescape=select_autoescape(default_for_string=False),
+ trim_blocks=True,
+ lstrip_blocks=True,
+ auto_reload=True,
+ undefined=StrictUndefined,
+)
+
+jinja_env.globals.update(
+ {
"now": lambda: datetime.now(ZoneInfo("UTC")),
"inspect": inspect,
- "id": id,
+ "getcwd": os.getcwd,
}
)
diff --git a/src/controlflow/utilities/logging.py b/src/controlflow/utilities/logging.py
index c04b4c79..04909d31 100644
--- a/src/controlflow/utilities/logging.py
+++ b/src/controlflow/utilities/logging.py
@@ -2,8 +2,6 @@
from functools import lru_cache
from typing import Optional
-from marvin.utilities.logging import add_logging_methods
-
@lru_cache()
def get_logger(name: Optional[str] = None) -> logging.Logger:
@@ -40,5 +38,4 @@ def get_logger(name: Optional[str] = None) -> logging.Logger:
else:
logger = parent_logger
- add_logging_methods(logger)
return logger
diff --git a/src/controlflow/utilities/marvin.py b/src/controlflow/utilities/marvin.py
index 5cf60426..9de99a46 100644
--- a/src/controlflow/utilities/marvin.py
+++ b/src/controlflow/utilities/marvin.py
@@ -1,93 +1,93 @@
-import inspect
-from contextlib import contextmanager
-from typing import Any, Callable
+# import inspect
+# from contextlib import contextmanager
+# from typing import Any, Callable
-import marvin.ai.text
-from marvin.client.openai import AsyncMarvinClient
-from marvin.settings import temporary_settings as temporary_marvin_settings
-from openai.types.chat import ChatCompletion
-from prefect import task as prefect_task
+# import marvin.ai.text
+# from marvin.client.openai import AsyncMarvinClient
+# from marvin.settings import temporary_settings as temporary_marvin_settings
+# from openai.types.chat import ChatCompletion
+# from prefect import task as prefect_task
-from controlflow.utilities.prefect import (
- create_json_artifact,
-)
+# from controlflow.utilities.prefect import (
+# create_json_artifact,
+# )
-original_classify_async = marvin.classify_async
-original_cast_async = marvin.cast_async
-original_extract_async = marvin.extract_async
-original_generate_async = marvin.generate_async
-original_paint_async = marvin.paint_async
-original_speak_async = marvin.speak_async
-original_transcribe_async = marvin.transcribe_async
+# original_classify_async = marvin.classify_async
+# original_cast_async = marvin.cast_async
+# original_extract_async = marvin.extract_async
+# original_generate_async = marvin.generate_async
+# original_paint_async = marvin.paint_async
+# original_speak_async = marvin.speak_async
+# original_transcribe_async = marvin.transcribe_async
-class AsyncControlFlowClient(AsyncMarvinClient):
- async def generate_chat(self, **kwargs: Any) -> "ChatCompletion":
- super_method = super().generate_chat
+# class AsyncControlFlowClient(AsyncMarvinClient):
+# async def generate_chat(self, **kwargs: Any) -> "ChatCompletion":
+# super_method = super().generate_chat
- @prefect_task(task_run_name="Generate OpenAI chat completion")
- async def _generate_chat(**kwargs):
- messages = kwargs.get("messages", [])
- create_json_artifact(key="prompt", data=messages)
- response = await super_method(**kwargs)
- create_json_artifact(key="response", data=response)
- return response
+# @prefect_task(task_run_name="Generate OpenAI chat completion")
+# async def _generate_chat(**kwargs):
+# messages = kwargs.get("messages", [])
+# create_json_artifact(key="prompt", data=messages)
+# response = await super_method(**kwargs)
+# create_json_artifact(key="response", data=response)
+# return response
- return await _generate_chat(**kwargs)
+# return await _generate_chat(**kwargs)
-def generate_task(name: str, original_fn: Callable):
- if inspect.iscoroutinefunction(original_fn):
+# def generate_task(name: str, original_fn: Callable):
+# if inspect.iscoroutinefunction(original_fn):
- @prefect_task(name=name)
- async def wrapper(*args, **kwargs):
- create_json_artifact(key="args", data=[args, kwargs])
- result = await original_fn(*args, **kwargs)
- create_json_artifact(key="result", data=result)
- return result
- else:
+# @prefect_task(name=name)
+# async def wrapper(*args, **kwargs):
+# create_json_artifact(key="args", data=[args, kwargs])
+# result = await original_fn(*args, **kwargs)
+# create_json_artifact(key="result", data=result)
+# return result
+# else:
- @prefect_task(name=name)
- def wrapper(*args, **kwargs):
- create_json_artifact(key="args", data=[args, kwargs])
- result = original_fn(*args, **kwargs)
- create_json_artifact(key="result", data=result)
- return result
+# @prefect_task(name=name)
+# def wrapper(*args, **kwargs):
+# create_json_artifact(key="args", data=[args, kwargs])
+# result = original_fn(*args, **kwargs)
+# create_json_artifact(key="result", data=result)
+# return result
- return wrapper
+# return wrapper
-@contextmanager
-def patch_marvin():
- with temporary_marvin_settings(default_async_client_cls=AsyncControlFlowClient):
- try:
- marvin.ai.text.classify_async = generate_task(
- "marvin.classify", original_classify_async
- )
- marvin.ai.text.cast_async = generate_task(
- "marvin.cast", original_cast_async
- )
- marvin.ai.text.extract_async = generate_task(
- "marvin.extract", original_extract_async
- )
- marvin.ai.text.generate_async = generate_task(
- "marvin.generate", original_generate_async
- )
- marvin.ai.images.paint_async = generate_task(
- "marvin.paint", original_paint_async
- )
- marvin.ai.audio.speak_async = generate_task(
- "marvin.speak", original_speak_async
- )
- marvin.ai.audio.transcribe_async = generate_task(
- "marvin.transcribe", original_transcribe_async
- )
- yield
- finally:
- marvin.ai.text.classify_async = original_classify_async
- marvin.ai.text.cast_async = original_cast_async
- marvin.ai.text.extract_async = original_extract_async
- marvin.ai.text.generate_async = original_generate_async
- marvin.ai.images.paint_async = original_paint_async
- marvin.ai.audio.speak_async = original_speak_async
- marvin.ai.audio.transcribe_async = original_transcribe_async
+# @contextmanager
+# def patch_marvin():
+# with temporary_marvin_settings(default_async_client_cls=AsyncControlFlowClient):
+# try:
+# marvin.ai.text.classify_async = generate_task(
+# "marvin.classify", original_classify_async
+# )
+# marvin.ai.text.cast_async = generate_task(
+# "marvin.cast", original_cast_async
+# )
+# marvin.ai.text.extract_async = generate_task(
+# "marvin.extract", original_extract_async
+# )
+# marvin.ai.text.generate_async = generate_task(
+# "marvin.generate", original_generate_async
+# )
+# marvin.ai.images.paint_async = generate_task(
+# "marvin.paint", original_paint_async
+# )
+# marvin.ai.audio.speak_async = generate_task(
+# "marvin.speak", original_speak_async
+# )
+# marvin.ai.audio.transcribe_async = generate_task(
+# "marvin.transcribe", original_transcribe_async
+# )
+# yield
+# finally:
+# marvin.ai.text.classify_async = original_classify_async
+# marvin.ai.text.cast_async = original_cast_async
+# marvin.ai.text.extract_async = original_extract_async
+# marvin.ai.text.generate_async = original_generate_async
+# marvin.ai.images.paint_async = original_paint_async
+# marvin.ai.audio.speak_async = original_speak_async
+# marvin.ai.audio.transcribe_async = original_transcribe_async
diff --git a/src/controlflow/utilities/prefect.py b/src/controlflow/utilities/prefect.py
index 305d4b0a..493ad130 100644
--- a/src/controlflow/utilities/prefect.py
+++ b/src/controlflow/utilities/prefect.py
@@ -1,20 +1,13 @@
import inspect
-import json
-from typing import Any, Callable
+from typing import Any
from uuid import UUID
-import prefect
-from marvin.types import FunctionTool
-from marvin.utilities.asyncio import run_sync
-from marvin.utilities.tools import tool_from_function
from prefect import get_client as get_prefect_client
-from prefect import task as prefect_task
from prefect.artifacts import ArtifactRequest
from prefect.context import FlowRunContext, TaskRunContext
+from prefect.utilities.asyncutils import run_sync
from pydantic import TypeAdapter
-from controlflow.utilities.types import AssistantTool, ToolType
-
def create_markdown_artifact(
key: str,
@@ -117,66 +110,49 @@ def create_python_artifact(
)
-def safe_isinstance(obj, type_) -> bool:
- # FunctionTool objects are typed generics, and
- # Python 3.9 will raise an error if you try to isinstance a typed generic...
- try:
- return isinstance(obj, type_)
- except TypeError:
- try:
- return issubclass(type(obj), type_)
- except TypeError:
- return False
-
-
-def wrap_prefect_tool(tool: ToolType) -> AssistantTool:
- """
- Wraps a Marvin tool in a prefect task
- """
- if not (
- safe_isinstance(tool, AssistantTool) or safe_isinstance(tool, FunctionTool)
- ):
- tool = tool_from_function(tool)
-
- if safe_isinstance(tool, FunctionTool):
- # for functions, we modify the function to become a Prefect task and
- # publish an artifact that contains details about the function call
-
- if isinstance(tool.function._python_fn, prefect.tasks.Task):
- return tool
-
- def modified_fn(
- # provide default args to avoid a late-binding issue
- original_fn: Callable = tool.function._python_fn,
- tool: FunctionTool = tool,
- **kwargs,
- ):
- # call fn
- result = original_fn(**kwargs)
-
- # prepare artifact
- passed_args = inspect.signature(original_fn).bind(**kwargs).arguments
- try:
- passed_args = json.dumps(passed_args, indent=2)
- except Exception:
- pass
- create_markdown_artifact(
- markdown=TOOL_CALL_FUNCTION_RESULT_TEMPLATE.format(
- name=tool.function.name,
- description=tool.function.description or "(none provided)",
- args=passed_args,
- result=result,
- ),
- key="result",
- )
-
- # return result
- return result
-
- # replace the function with the modified version
- tool.function._python_fn = prefect_task(
- modified_fn,
- task_run_name=f"Tool call: {tool.function.name}",
- )
-
- return tool
+# def wrap_prefect_tool(tool: ToolType) -> AssistantTool:
+# if not (isinstance(tool, AssistantTool) or isinstance(tool, ToolFunction)):
+# tool = tool(tool)
+
+# if isinstance(tool, ToolFunction):
+# # for functions, we modify the function to become a Prefect task and
+# # publish an artifact that contains details about the function call
+
+# if isinstance(tool.function._python_fn, prefect.tasks.Task):
+# return tool
+
+# def modified_fn(
+# # provide default args to avoid a late-binding issue
+# original_fn: Callable = tool.function._python_fn,
+# tool: ToolFunction = tool,
+# **kwargs,
+# ):
+# # call fn
+# result = original_fn(**kwargs)
+
+# # prepare artifact
+# passed_args = inspect.signature(original_fn).bind(**kwargs).arguments
+# try:
+# passed_args = json.dumps(passed_args, indent=2)
+# except Exception:
+# pass
+# create_markdown_artifact(
+# markdown=TOOL_CALL_FUNCTION_RESULT_TEMPLATE.format(
+# name=tool.function.name,
+# description=tool.function.description or "(none provided)",
+# args=passed_args,
+# result=result,
+# ),
+# key="result",
+# )
+
+# # return result
+# return result
+
+# # replace the function with the modified version
+# tool.function._python_fn = prefect_task(
+# modified_fn,
+# task_run_name=f"Tool call: {tool.function.name}",
+# )
+
+# return tool
diff --git a/src/controlflow/utilities/threads.py b/src/controlflow/utilities/threads.py
deleted file mode 100644
index c707a9f8..00000000
--- a/src/controlflow/utilities/threads.py
+++ /dev/null
@@ -1,27 +0,0 @@
-from marvin.beta.assistants.threads import Message, Thread
-
-THREAD_REGISTRY = {}
-
-
-def save_thread(name: str, thread: Thread):
- """
- Save an OpenAI thread to the thread registry under a known name
- """
- THREAD_REGISTRY[name] = thread
-
-
-def load_thread(name: str):
- """
- Load an OpenAI thread from the thread registry by name
- """
- if name not in THREAD_REGISTRY:
- thread = Thread()
- save_thread(name, thread)
- return THREAD_REGISTRY[name]
-
-
-def get_history(thread_id: str, limit: int = None) -> list[Message]:
- """
- Get the history of a thread
- """
- return Thread(id=thread_id).get_messages(limit=limit)
diff --git a/src/controlflow/utilities/types.py b/src/controlflow/utilities/types.py
index 722cd7cb..43d878ab 100644
--- a/src/controlflow/utilities/types.py
+++ b/src/controlflow/utilities/types.py
@@ -1,23 +1,13 @@
-from enum import Enum
-from functools import partial, update_wrapper
-from typing import Callable, Optional, Union
-
-from marvin.beta.assistants import Assistant, Thread
-from marvin.beta.assistants.assistants import AssistantTool
-from marvin.types import FunctionTool
-from marvin.utilities.asyncio import ExposeSyncMethodsMixin
+from typing import Union
+
from pydantic import (
BaseModel,
- computed_field,
)
# flag for unset defaults
NOTSET = "__NOTSET__"
-ToolType = Union[FunctionTool, AssistantTool, Callable]
-
-
class ControlFlowModel(BaseModel):
model_config = dict(validate_assignment=True, extra="forbid")
diff --git a/tests/core/test_controller.py b/tests/core/test_controller.py
index a9953828..6422dc6c 100644
--- a/tests/core/test_controller.py
+++ b/tests/core/test_controller.py
@@ -67,9 +67,9 @@ def test_controller_agent_selection(self, flow, monkeypatch):
agent2 = Agent(name="Agent 2")
task = Task(objective="Test Task", agents=[agent1, agent2])
controller = Controller(flow=flow, tasks=[task], agents=[agent1, agent2])
- mocked_marvin_moderator = AsyncMock(return_value=agent1)
+ mocked_classify_moderator = AsyncMock(return_value=agent1)
monkeypatch.setattr(
- "controlflow.core.controller.moderators.marvin_moderator",
- mocked_marvin_moderator,
+ "controlflow.core.controller.moderators.classify_moderator",
+ mocked_classify_moderator,
)
assert controller.agents == [agent1, agent2]
diff --git a/tests/core/test_flows.py b/tests/core/test_flows.py
index f159b278..180a08f9 100644
--- a/tests/core/test_flows.py
+++ b/tests/core/test_flows.py
@@ -6,7 +6,7 @@
class TestFlow:
def test_flow_initialization(self):
flow = Flow()
- assert flow.thread is not None
+ assert flow.thread_id is not None
assert len(flow.tools) == 0
assert len(flow.agents) == 0
assert len(flow.context) == 0
diff --git a/tests/fixtures/mocks.py b/tests/fixtures/mocks.py
index 4009476b..a4cbf656 100644
--- a/tests/fixtures/mocks.py
+++ b/tests/fixtures/mocks.py
@@ -6,7 +6,7 @@
from controlflow.core.agent import Agent
from controlflow.core.task import Task, TaskStatus
from controlflow.llm.completions import Response
-from marvin.settings import temporary_settings as temporary_marvin_settings
+from controlflow.settings import temporary_settings
def new_chunk():
@@ -18,35 +18,10 @@ def new_chunk():
@pytest.fixture
def prevent_openai_calls():
"""Prevent any calls to the OpenAI API from being made."""
- with temporary_marvin_settings(openai__api_key="unset"):
+ with temporary_settings(llm_api_key="unset"):
yield
-@pytest.fixture
-def mock_run(monkeypatch, prevent_openai_calls):
- """
- This fixture mocks the calls to the OpenAI Assistants API. Use it in a test
- and assign any desired side effects (like completing a task) to the mock
- object's `.side_effect` attribute.
-
- For example:
-
- def test_example(mock_run):
- task = Task(objective="Say hello")
-
- def side_effect():
- task.mark_complete()
-
- mock_run.side_effect = side_effect
-
- task.run()
-
- """
- MockRun = AsyncMock()
- monkeypatch.setattr("controlflow.core.controller.controller.Run.run_async", MockRun)
- yield MockRun
-
-
@pytest.fixture
def mock_controller_run_agent(monkeypatch, prevent_openai_calls):
MockRunAgent = AsyncMock()
@@ -68,9 +43,6 @@ def get_messages(*args, **kwargs):
monkeypatch.setattr(
"controlflow.core.controller.controller.Controller._run_agent", MockRunAgent
)
- monkeypatch.setattr(
- "marvin.beta.assistants.Thread.get_messages", MockThreadGetMessages
- )
yield MockRunAgent