search

stitionai / devika

Devika is an Agentic AI Software Engineer that can understand high-level human instructions, break them down into steps, research relevant information, and write code to achieve the given objective. Devika aims to be a competitive open-source alternative to Devin by Cognition AI.
MIT License
18.34k stars 2.38k forks source link

[ISSUE] Devika the does not search in Google #500

Open Sinicyn78 opened 5 months ago

Sinicyn78 commented 5 months ago

Describe your issue

Devika does not search on Google. all apis are registered

How To Reproduce

  1. Tsk: write the code for a simple telegram bot in Python
  2. Wrote a plan
  3. Invalid response from the model

logs: 24.04.27 13:31:01: root: INFO : SOCKET server-message MESSAGE: {'messages': {'from_devika': True, 'message': '{\n "1": "Install required libraries by running pip install python-telegram-bot in the terminal.",\n "2": "Create a new Python file (e.g., bot.py) and import the necessary modules, including telegram-bot and logging.",\n "3": "Define the bot\'s token and API endpoint using the Telegram BotFather tool or an existing bot token.",\n "4": "Implement basic bot functionality by creating a simple command-handling mechanism that responds to user input (e.g., greetings, goodbyes).",\n "5": "Test the bot by sending messages and verifying responses. "\n}', 'timestamp': '2024-04-27 13:31:01'}}

24.04.27 13:31:01: root: INFO : SOCKET server-message MESSAGE: {'messages': {'from_devika': True, 'message': "I'm excited to create a simple Telegram bot in Python, let's get started!", 'timestamp': '2024-04-27 13:31:01'}}

24.04.27 13:31:01: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 6637}

24.04.27 13:30:50: root: INFO : SOCKET inference MESSAGE: {'type': 'time', 'elapsed_time': '0.00'}

24.04.27 13:30:50: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 6457}

24.04.27 13:30:50: root: INFO : SOCKET agent-state MESSAGE: [{'internal_monologue': "I'm starting the work...", 'browser_session': {'url': None, 'screenshot': None}, 'terminal_session': {'command': None, 'output': None, 'title': None}, 'step': 1, 'message': None, 'completed': False, 'agent_is_active': True, 'token_usage': 0, 'timestamp': '2024-04-27 13:30:50'}]

24.04.27 13:30:49: root: INFO : SOCKET server-message MESSAGE: {'messages': {'from_devika': False, 'message': 'write the code for a simple telegram bot in Python', 'timestamp': '2024-04-27 13:30:49'}}

24.04.27 13:30:49: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': "previous agent doesn't completed it's task."}

24.04.27 13:30:46: root: INFO : SOCKET socket_response MESSAGE: {'data': 'Server Connected'}

24.04.27 13:25:03: root: INFO : SOCKET info MESSAGE: {'type': 'error', 'message': 'Maximum attempts reached. model keeps failing.'}

24.04.27 13:25:01: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': 'Invalid response from the model, trying again...'}

24.04.27 13:25:01: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 5914}

24.04.27 13:24:57: root: INFO : SOCKET inference MESSAGE: {'type': 'time', 'elapsed_time': '0.00'}

24.04.27 13:24:57: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 6402}

24.04.27 13:24:55: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': 'Invalid response from the model, trying again...'}

24.04.27 13:24:55: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 4865}

24.04.27 13:24:50: root: INFO : SOCKET inference MESSAGE: {'type': 'time', 'elapsed_time': '0.00'}

24.04.27 13:24:50: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 5269}

24.04.27 13:24:48: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': 'Invalid response from the model, trying again...'}

24.04.27 13:24:48: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 3779}

24.04.27 13:24:42: root: INFO : SOCKET inference MESSAGE: {'type': 'time', 'elapsed_time': '0.00'}

24.04.27 13:24:42: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 4089}

24.04.27 13:24:40: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': 'Invalid response from the model, trying again...'}

24.04.27 13:24:40: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 2601}

24.04.27 13:24:34: root: INFO : SOCKET inference MESSAGE: {'type': 'time', 'elapsed_time': '0.00'}

24.04.27 13:24:34: root: INFO : SOCKET tokens MESSAGE: {'token_usage': 2907}

24.04.27 13:24:31: root: INFO : SOCKET info MESSAGE: {'type': 'warning', 'message': 'Invalid response from the model, trying again...'}

Configuration

- OS: [Windows 10]
- Python version: [e.g. 3.10]
- Node version: [20.12.0]
- bun version: [1.1.5]
- search engine: [google]
- Model: [llama 3]

image

obliviousz commented 5 months ago

This is due to the not-so-good research prompt. I have faced the same issue in llama3. You can check here the issue and how I fixed it in my youtube channel.

https://youtu.be/Cn_39Cb8WbQ?si=1qr_CsLeRD5kd1eW