send messages separately, adjust summarize prompt

agenthub/codeact_agent/codeact_agent.py

@@ -53,7 +53,7 @@ def action_to_str(action: Action) -> str:
     elif isinstance(action, MessageAction):
         return action.content
     elif isinstance(action, AgentSummarizeAction):
-        return action.summary
+        return action.summarized_actions
     return ''
 
 
@@ -97,6 +97,8 @@ def get_observation_message(obs) -> dict[str, str] | None:
     elif isinstance(obs, AgentDelegateObservation):
         content = 'OBSERVATION:\n' + truncate_content(str(obs.outputs))
         return {'role': 'user', 'content': content}
+    elif isinstance(obs, AgentSummarizeAction):
+        return {'role': 'user', 'content': obs.summarized_observations}
     return None
 
 
@@ -262,14 +264,29 @@ class CodeActAgent(Agent):
         ]
 
         for event in state.history.get_events():
+            # split summarize message into action and observation
+            if isinstance(event, AgentSummarizeAction):
+                action_message = get_action_message(event)
+                if action_message:
+                    messages.append(action_message)
+                observation_message = get_observation_message(event)
+                if observation_message:
+                    messages.append(observation_message)
+                continue
+
+            # find regular message
             message = (
                 get_action_message(event)
                 if isinstance(event, Action)
                 else get_observation_message(event)
             )
+
+            # add regular message
             if message:
                 messages.append(message)
 
+        # the latest user message is important:
+        # we want to remind the agent of the environment constraints
         latest_user_message = next(
             (m for m in reversed(messages) if m['role'] == 'user'), None
         )
@@ -277,11 +294,11 @@ class CodeActAgent(Agent):
         if latest_user_message:
             if state.almost_stuck == 1:
                 latest_user_message['content'] += (
-                    '\n\nENVIRONMENT REMINDER: You are almost stuck repeating the same action. You have only 1 iteration left and you must change your approach. Now.'
+                    '\n\nENVIRONMENT REMINDER: You are almost stuck, repeating the same action. You have only 1 chance left before you fail. You MUST change your approach. Re-think the problem. Do NOT repeat the same action again, you will fail.'
                 )
             elif state.almost_stuck == 2:
                 latest_user_message['content'] += (
-                    '\n\nENVIRONMENT REMINDER: You are almost stuck repeating the same action. You have only 2 iterations left and you must change your approach.'
+                    '\n\nENVIRONMENT REMINDER: You are almost stuck, repeating the same action. You have only 2 iterations left before you fail. Do NOT repeat the same action again, you will fail the same as before. Re-think the problem.'
                 )
             else:
                 latest_user_message['content'] += (

opendevin/core/main.py

@@ -85,7 +85,7 @@ async def main(
     AgentCls: Type[Agent] = Agent.get_cls(args.agent_cls)
     agent = AgentCls(llm=llm)
 
-    event_stream = EventStream('main' + (f'_{sid if sid else ""}'))
+    event_stream = EventStream('main' + ('_' + sid if sid else ''))
 
     # restore main session if enabled
     initial_state = None

opendevin/events/action/agent.py

@@ -37,12 +37,12 @@ class AgentSummarizeAction(Action):
     Action to summarize a list of events.
 
     Attributes:
-    - actions: A comma-separated list of all the action names from the summarized actions.
-    - summary: A single sentence summarizing all the observations.
+    - summarized_actions: A sentence summarizing all the actions.
+    - summarized_observations: A few sentences summarizing all the observations.
     """
 
     summarized_actions: str = ''
-    summary: str = ''
+    summarized_observations: str = ''
     action: str = ActionType.SUMMARIZE
     _chunk_start: int = -1
     _chunk_end: int = -1
@@ -50,11 +50,12 @@ class AgentSummarizeAction(Action):
 
     @property
     def message(self) -> str:
-        return self.summary
+        return self.summarized_observations
 
     def __str__(self) -> str:
         ret = '**AgentSummarizeAction**\n'
-        ret += f'SUMMARY: {self.summary}'
+        ret += f'SUMMARIZED ACTIONS: {self.summarized_actions}\n'
+        ret += f'SUMMARIZED OBSERVATIONS: {self.summarized_observations}\n'
         return ret
 
 

opendevin/memory/prompts.py

@@ -6,14 +6,25 @@ from opendevin.core.exceptions import (
 from opendevin.core.logger import opendevin_logger as logger
 from opendevin.core.utils import json
 from opendevin.events.action.agent import AgentSummarizeAction
+from opendevin.events.event import EventSource
 from opendevin.events.serialization.action import action_from_dict
 
 SUMMARY_PROMPT = """
 Given the following actions and observations, create a JSON response with:
     - "action": "summarize"
     - args:
-      - "summarized_actions": A comma-separated list of unique action names from the provided actions
-      - "summary": A single sentence summarizing all the provided observations
+      - "summarized_actions": A sentence summarizing all the provided actions, at first person
+      - "summarized observations": A few sentences summarizing all the provided observations, at third person
+
+Example:
+        {
+            "action": "summarize",
+            "args": {
+                "summarized_actions": "I opened the uml file.",
+                "summarized observations": "The agent ran a python script to open the uml.pdf file."
+            }
+        }
+Make sure to include in observations any relevant information that the agent should remember.
 %(events)s
 """
 
@@ -46,6 +57,7 @@ def parse_summary_response(response: str) -> AgentSummarizeAction:
             error_message = f'Expected a summarize action, but the response got {str(type(action)) if action else None}'
             logger.error(error_message)
             raise InvalidSummaryResponseError(error_message)
+        action._source = EventSource.AGENT  # type: ignore
     except (LLMResponseError, LLMMalformedActionError) as e:
         logger.error(f'Failed to parse summary response: {e}')
         raise InvalidSummaryResponseError(