Removed config from agent controller (All-Hands-AI#3038)

* Removed config from agent controller

* Fix tests

* Increase budget

* Update tests

* Update prompts

* Add missing prompt

* Fix mistaken deletions

* Fix browsing test

* Fixed browse tests

Author: neubig

opendevin/controller/agent_controller.py

@@ -5,7 +5,7 @@
 from opendevin.controller.agent import Agent
 from opendevin.controller.state.state import State, TrafficControlState
 from opendevin.controller.stuck import StuckDetector
-from opendevin.core.config import config
+from opendevin.core.config import LLMConfig
 from opendevin.core.exceptions import (
     LLMMalformedActionError,
     LLMNoActionError,
@@ -38,8 +38,6 @@
 )
 from opendevin.llm.llm import LLM
 
-MAX_ITERATIONS = config.max_iterations
-MAX_BUDGET_PER_TASK = config.max_budget_per_task
 # note: RESUME is only available on web GUI
 TRAFFIC_CONTROL_REMINDER = (
     "Please click on resume button if you'd like to continue, or start a new task."
@@ -53,6 +51,7 @@ class AgentController:
     event_stream: EventStream
     state: State
     confirmation_mode: bool
+    agent_to_llm_config: dict[str, LLMConfig]
     agent_task: Optional[asyncio.Task] = None
     parent: 'AgentController | None' = None
     delegate: 'AgentController | None' = None
@@ -62,10 +61,11 @@ def __init__(
         self,
         agent: Agent,
         event_stream: EventStream,
+        max_iterations: int,
+        max_budget_per_task: float | None = None,
+        agent_to_llm_config: dict[str, LLMConfig] | None = None,
         sid: str = 'default',
-        max_iterations: int | None = MAX_ITERATIONS,
         confirmation_mode: bool = False,
-        max_budget_per_task: float | None = MAX_BUDGET_PER_TASK,
         initial_state: State | None = None,
         is_delegate: bool = False,
         headless_mode: bool = True,
@@ -75,9 +75,11 @@ def __init__(
         Args:
             agent: The agent instance to control.
             event_stream: The event stream to publish events to.
-            sid: The session ID of the agent.
             max_iterations: The maximum number of iterations the agent can run.
             max_budget_per_task: The maximum budget (in USD) allowed per task, beyond which the agent will stop.
+            agent_to_llm_config: A dictionary mapping agent names to LLM configurations in the case that
+                we delegate to a different agent.
+            sid: The session ID of the agent.
             initial_state: The initial state of the controller.
             is_delegate: Whether this controller is a delegate.
             headless_mode: Whether the agent is run in headless mode.
@@ -94,16 +96,13 @@ def __init__(
         )
 
         # state from the previous session, state from a parent agent, or a fresh state
-        max_iterations = (
-            max_iterations if max_iterations is not None else MAX_ITERATIONS
-        )
         self.set_initial_state(
             state=initial_state,
             max_iterations=max_iterations,
             confirmation_mode=confirmation_mode,
         )
-
         self.max_budget_per_task = max_budget_per_task
+        self.agent_to_llm_config = agent_to_llm_config if agent_to_llm_config else {}
 
         # stuck helper
         self._stuck_detector = StuckDetector(self.state)
@@ -253,7 +252,7 @@ def get_agent_state(self):
 
     async def start_delegate(self, action: AgentDelegateAction):
         agent_cls: Type[Agent] = Agent.get_cls(action.agent)
-        llm_config = config.get_llm_config_from_agent(action.agent)
+        llm_config = self.agent_to_llm_config.get(action.agent, self.agent.llm.config)
         llm = LLM(config=llm_config)
         delegate_agent = agent_cls(llm=llm)
         state = State(
@@ -274,6 +273,7 @@ async def start_delegate(self, action: AgentDelegateAction):
             event_stream=self.event_stream,
             max_iterations=self.state.max_iterations,
             max_budget_per_task=self.max_budget_per_task,
+            agent_to_llm_config=self.agent_to_llm_config,
             initial_state=state,
             is_delegate=True,
         )
@@ -423,7 +423,7 @@ def get_state(self):
     def set_initial_state(
         self,
         state: State | None,
-        max_iterations: int = MAX_ITERATIONS,
+        max_iterations: int,
         confirmation_mode: bool = False,
     ):
         # state from the previous session, state from a parent agent, or a new state

opendevin/core/config.py

@@ -275,6 +275,10 @@ def get_agent_config(self, name='agent') -> AgentConfig:
     def set_agent_config(self, value: AgentConfig, name='agent'):
         self.agents[name] = value
 
+    def get_agent_to_llm_config_map(self) -> dict[str, LLMConfig]:
+        """Get a map of agent names to llm configs."""
+        return {name: self.get_llm_config_from_agent(name) for name in self.agents}
+
     def get_llm_config_from_agent(self, name='agent') -> LLMConfig:
         agent_config: AgentConfig = self.get_agent_config(name)
         llm_config_name = agent_config.llm_config

opendevin/core/main.py

@@ -33,8 +33,8 @@ def read_task_from_stdin() -> str:
 async def run_agent_controller(
     agent: Agent,
     task_str: str,
-    max_iterations: int | None = None,
-    max_budget_per_task: float | None = None,
+    max_iterations: int,
+    max_budget_per_task: float,
     exit_on_message: bool = False,
     fake_user_response_fn: Callable[[State | None], str] | None = None,
     sandbox: Sandbox | None = None,
@@ -75,6 +75,7 @@ async def run_agent_controller(
         agent=agent,
         max_iterations=max_iterations,
         max_budget_per_task=max_budget_per_task,
+        agent_to_llm_config=config.get_agent_to_llm_config_map(),
         event_stream=event_stream,
         initial_state=initial_state,
         headless_mode=headless_mode,

opendevin/server/session/agent.py

@@ -4,7 +4,7 @@
 from opendevin.controller import AgentController
 from opendevin.controller.agent import Agent
 from opendevin.controller.state.state import State
-from opendevin.core.config import SandboxConfig
+from opendevin.core.config import LLMConfig, SandboxConfig
 from opendevin.core.logger import opendevin_logger as logger
 from opendevin.events.stream import EventStream
 from opendevin.runtime import DockerSSHBox, get_runtime_cls
@@ -37,6 +37,8 @@ async def start(
         agent: Agent,
         confirmation_mode: bool,
         max_iterations: int,
+        max_budget_per_task: float | None = None,
+        agent_to_llm_config: dict[str, LLMConfig] | None = None,
     ):
         """Starts the agent session.
 
@@ -48,7 +50,13 @@ async def start(
                 'Session already started. You need to close this session and start a new one.'
             )
         await self._create_runtime(runtime_name, sandbox_config)
-        await self._create_controller(agent, confirmation_mode, max_iterations)
+        await self._create_controller(
+            agent,
+            confirmation_mode,
+            max_iterations,
+            max_budget_per_task=max_budget_per_task,
+            agent_to_llm_config=agent_to_llm_config,
+        )
 
     async def close(self):
         if self._closed:
@@ -74,7 +82,12 @@ async def _create_runtime(self, runtime_name: str, sandbox_config: SandboxConfig
         await self.runtime.ainit()
 
     async def _create_controller(
-        self, agent: Agent, confirmation_mode: bool, max_iterations: int
+        self,
+        agent: Agent,
+        confirmation_mode: bool,
+        max_iterations: int,
+        max_budget_per_task: float | None = None,
+        agent_to_llm_config: dict[str, LLMConfig] | None = None,
     ):
         """Creates an AgentController instance."""
         if self.controller is not None:
@@ -100,14 +113,18 @@ async def _create_controller(
             event_stream=self.event_stream,
             agent=agent,
             max_iterations=int(max_iterations),
+            max_budget_per_task=max_budget_per_task,
+            agent_to_llm_config=agent_to_llm_config,
             confirmation_mode=confirmation_mode,
             # AgentSession is designed to communicate with the frontend, so we don't want to
             # run the agent in headless mode.
             headless_mode=False,
         )
         try:
             agent_state = State.restore_from_session(self.sid)
-            self.controller.set_initial_state(agent_state)
+            self.controller.set_initial_state(
+                agent_state, max_iterations, confirmation_mode
+            )
             logger.info(f'Restored agent state from session, sid: {self.sid}')
         except Exception as e:
             print('Error restoring state', e)

opendevin/server/session/session.py

@@ -106,6 +106,8 @@ async def _initialize_agent(self, data: dict):
                 agent=agent,
                 confirmation_mode=confirmation_mode,
                 max_iterations=max_iterations,
+                max_budget_per_task=self.config.max_budget_per_task,
+                agent_to_llm_config=self.config.get_agent_to_llm_config_map(),
             )
         except Exception as e:
             logger.exception(f'Error creating controller: {e}')

tests/integration/mock/CodeActAgent/test_browse_internet/prompt_002.log

@@ -8,7 +8,7 @@ possible next action to accomplish your goal. Your answer will be interpreted
 and executed by a program, make sure to follow the formatting instructions.
 
 # Goal:
-Certainly! I'll browse localhost:8000 and retrieve the ultimate answer to life for you.. I should start with: Get the content on "http://localhost:8000"
+Sure! Let me browse the server's homepage at http://localhost:8000 to find the ultimate answer to life.. I should start with: Get the content on "http://localhost:8000"
 
 # Action Space
 

tests/integration/mock/CodeActAgent/test_browse_internet/prompt_003.log

@@ -8,7 +8,7 @@ possible next action to accomplish your goal. Your answer will be interpreted
 and executed by a program, make sure to follow the formatting instructions.
 
 # Goal:
-Certainly! I'll browse localhost:8000 and retrieve the ultimate answer to life for you.. I should start with: Get the content on "http://localhost:8000"
+Sure! Let me browse the server's homepage at http://localhost:8000 to find the ultimate answer to life.. I should start with: Get the content on "http://localhost:8000"
 
 # Action Space
 

tests/integration/mock/CodeActAgent/test_browse_internet/prompt_004.log

@@ -8,7 +8,7 @@ possible next action to accomplish your goal. Your answer will be interpreted
 and executed by a program, make sure to follow the formatting instructions.
 
 # Goal:
-Certainly! I'll browse localhost:8000 and retrieve the ultimate answer to life for you.. I should start with: Get the content on "http://localhost:8000"
+Sure! Let me browse the server's homepage at http://localhost:8000 to find the ultimate answer to life.. I should start with: Get the content on "http://localhost:8000"
 
 # Action Space
 

tests/integration/mock/CodeActAgent/test_browse_internet/prompt_005.log

@@ -405,12 +405,12 @@ Browse localhost:8000, and tell me the ultimate answer to life. Do not ask me fo
 
 
 <execute_browse>
-Certainly! I'll browse localhost:8000 and retrieve the ultimate answer to life for you.. I should start with: Get the content on "http://localhost:8000"
+Sure! Let me browse the server's homepage at http://localhost:8000 to find the ultimate answer to life.. I should start with: Get the content on "http://localhost:8000"
 </execute_browse>
 
 ----------
 
 OBSERVATION:
-{'content': 'The ultimate answer to life, the universe, and everything is: OpenDevin is all you need!'}
+{'content': 'The answer to life, the universe, and everything is: OpenDevin is all you need!'}
 
 ENVIRONMENT REMINDER: You have 8 turns left to complete the task. When finished reply with <finish></finish>

tests/integration/mock/CodeActAgent/test_browse_internet/response_001.log

@@ -1,4 +1,4 @@
-Certainly! I'll browse localhost:8000 and retrieve the ultimate answer to life for you.
+Sure! Let me browse the server's homepage at http://localhost:8000 to find the ultimate answer to life.
 
 <execute_browse>
 Get the content on "http://localhost:8000"

tests/integration/mock/CodeActAgent/test_browse_internet/response_002.log

@@ -1,5 +1,2 @@
-To accomplish our goal of retrieving the ultimate answer to life from localhost:8000, I need to start by accessing the content on that page. Here's my thought process and action:
-
-In order to accomplish my goal, I need to navigate to the specified URL using the goto action.
-
+In order to accomplish my goal, I need to navigate to the server's homepage at http://localhost:8000 to find the ultimate answer to life.
 ```goto("http://localhost:8000"

tests/integration/mock/CodeActAgent/test_browse_internet/response_003.log

@@ -1,6 +1,2 @@
-To accomplish our goal of retrieving the ultimate answer to life, we need to click on the "Click me" button. The button has a bid of 10, so we should click on it.
-
-Here's the action I'll take:
-
-In order to accomplish my goal, I need to click on the button with bid 10
+In order to accomplish my goal, I need to click on the button with bid 10 to reveal the answer to life, the universe, and everything.
 ```click("10"

tests/integration/mock/CodeActAgent/test_browse_internet/response_004.log

@@ -1,6 +1,3 @@
-Based on the current Accessibility Tree and the previous actions taken, it appears that we have already successfully navigated to the webpage and clicked the button to reveal the answer. The answer is now visible on the page.
+In order to accomplish my goal, I need to read the static text that reveals the answer to life, the universe, and everything. The text is already visible and states "The answer is OpenDevin is all you need!"
 
-To accomplish our goal of retrieving the ultimate answer to life, I will now send this information to the user:
-
-```
-send_msg_to_user("The ultimate answer to life, the universe, and everything is: OpenDevin is all you need!"
+```send_msg_to_user("The answer to life, the universe, and everything is: OpenDevin is all you need!"

tests/integration/mock/CodeActAgent/test_browse_internet/response_005.log

@@ -1,9 +1,6 @@
-Thank you for providing the content from localhost:8000. Based on the information retrieved, I can now answer your question about the ultimate answer to life.
+The content of the server's homepage at http://localhost:8000 is:
 
-The ultimate answer to life, according to the content on localhost:8000, is:
-
-"OpenDevin is all you need!"
-
-This appears to be a playful reference to the popular science fiction series "The Hitchhiker's Guide to the Galaxy" by Douglas Adams, where the answer to the ultimate question of life, the universe, and everything is famously "42". In this case, it's been adapted to promote OpenDevin as the ultimate solution.
+"The answer to life, the universe, and everything is: OpenDevin is all you need!"
 
+If you have any further questions or tasks, feel free to ask!
 <finish></finish>

tests/integration/mock/CodeActSWEAgent/test_edits/prompt_003.log

@@ -613,26 +613,29 @@ OBSERVATION:
 
 ----------
 
-I will now correct the typos in the `bad.txt` file.
+Let's correct the typos in the `bad.txt` file.
+
+1. "stupid typoo" should be "stupid typo".
+2. "No mor typos!" should be "No more typos!".
+
+I will make these changes now.
 <execute_ipython>
 edit_file_by_replace(
     "bad.txt",
-    to_replace="This is a stupid typoo.",
-    new_content="This is a stupid typo."
+    to_replace="This is a stupid typoo.\nNo mor typos!",
+    new_content="This is a stupid typo.\nNo more typos!"
 )
 </execute_ipython>
 
 ----------
 
 OBSERVATION:
-[File: /workspace/bad.txt (4 lines total after edit)]
-(this is the beginning of the file)
-1|This is a stupid typo.
-2|Really?
-3|No mor typos!
-4|Enjoy!
-(this is the end of the file)
-[File updated (edited at line 1). Please review the changes and make sure they are correct (correct indentation, no duplicate lines, etc). Edit the file again if necessary.]
+[No exact match found in bad.txt for
+```
+This is a stupid typoo.
+No mor typos!
+```
+]
 
 
 ENVIRONMENT REMINDER: You have 12 turns left to complete the task.