[Bug]: Loop Answer
Is there an existing issue for the same bug?
- [x] I have checked the troubleshooting document at https://opendevin.github.io/OpenDevin/modules/usage/troubleshooting
- [x] I have checked the existing issues.
Describe the bug
When I run anything in the interface, a loop start with the Answer...
Current Version
0.40
Installation and Configuration
docker run -e SANDBOX_TYPE=exec -e LLM_API_KEY="my-openai-key" -e WORKSPACE_MOUNT_PATH=workspace -vworkspace:/opt/workspace_base -v /var/run/docker.sock:/var/run/docker.sock -p 3000:3000 --add-host host.docker.internal=host-gateway ghcr.io/opendevin/opendevin:0.4.0
Model and Agent
Model: gpt-3.5-turbo Agent: At every agent
Reproduction Steps
- I give a command in the interface
- Agent give me an Answer
- Agent get stuck in a loop
Logs, Errors, Screenshots, and Additional Context
No response
@zGamerZz can you send a specific example, with a specific agent?
gpt-3.5 should be good enough for simple tasks, and we have loop-detection already...
@zGamerZz can you send a specific example, with a specific agent?
gpt-3.5 should be good enough for simple tasks, and we have loop-detection already...
Sure, I used now the MonologueAgent:
And here the terminal output:
==============
STEP 0
07:58:45 - PLAN
Can we start?
07:58:45 - opendevin:INFO: prompt.py:190
INFO
HINT:
07:58:46 - ACTION
AgentThinkAction(thought='I need to start by marking a task as in_progress.', action='think')
==============
STEP 1
07:58:46 - PLAN
Can we start?
07:58:46 - opendevin:INFO: prompt.py:190
INFO
HINT:
Look at your last thought in the history above. What does it suggest? Don't think anymore--take action.
07:58:47 - ACTION
AddTaskAction(parent='0', goal='Mark a task as in_progress', subtasks=[], action='add_task')
==============
STEP 2
07:58:47 - PLAN
Can we start?
07:58:47 - opendevin:INFO: prompt.py:190
INFO
HINT:
You should think about the next action to take.
07:58:48 - ACTION
AgentThinkAction(thought="I should mark the task 'Mark a task as in_progress' as in_progress.", action='think')
INFO: 172.17.0.1:60580 - "GET /api/plan HTTP/1.1" 200 OK
==============
STEP 3
07:58:48 - PLAN
Can we start?
07:58:48 - opendevin:INFO: prompt.py:190
INFO
HINT:
Look at your last thought in the history above. What does it suggest? Don't think anymore--take action.
07:58:49 - ACTION
ModifyTaskAction(id='0.0', state='in_progress', action='modify_task')
==============
STEP 4
07:58:49 - PLAN
Can we start?
07:58:49 - opendevin:INFO: prompt.py:190
INFO
HINT:
You should think about the next action to take.
07:58:50 - ACTION
AgentThinkAction(thought='I should consider what needs to be done next to move the task forward.', action='think')
INFO: 172.17.0.1:60580 - "GET /api/plan HTTP/1.1" 200 OK
==============
STEP 5
07:58:50 - PLAN
Can we start?
07:58:50 - opendevin:INFO: prompt.py:190
INFO
HINT:
Look at your last thought in the history above. What does it suggest? Don't think anymore--take action.
07:58:51 - ACTION
AgentThinkAction(thought='I should consider what needs to be done next to move the task forward.', action='think')
==============
STEP 6
07:58:51 - PLAN
Can we start?
07:58:51 - opendevin:INFO: prompt.py:190
INFO
HINT:
Look at your last thought in the history above. What does it suggest? Don't think anymore--take action.
07:58:52 - ACTION
AgentThinkAction(thought='I should consider what needs to be done next to move the task forward.', action='think')
issue still there
Looks like the loop got caught at least.
Can you try CodeActAgent?
This issue is stale because it has been open for 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.
This issue was closed because it has been stalled for over 30 days with no activity.