Updated Files | as of April 12, 2023, edit** EST 3:28pm | Some Q/A

[ChessScholar]

Hi everyone,

After meticulously going through all the forums, issues, and suggested file updates, and noticing everyone having similar issues that are being reposted umpteen times, I created a .zip of all the [currently working] updates and suggestions to have a version of Auto-GPT without errors (json, invalid, yada yada)...

They are still working on adding the edits to the master. As one guy said somewhere in the discussions, give the guy (Torantulino) time to sleep! This .zip took me several hours/days to make sure it comes with no errors, my girlfriend was upset that I was ignoring her :')

Anyway, if there are any other errors on your end, please post the issues as you normally would either in the forums or in this post. Understand that I cannot answer them as the others who are actually working on the project are far more experienced in debugging. I simply took their suggestions and edits and put them in this .zip to help reduce the repeated issue posts. I will continue to edit the .zip as shown below when I see the improvements.

A couple notes:

• If it looks like the program is going through a "do_nothing" binge, give it a few more "y" commands and it will eventually come up with a command on its own without the need for additional prompts. Mine took as many as 16 prompts before giving itself a command.
• If it looks like it's going through an infinite "start_agent" binge, don't worry. It should correct itself by watching which agents its using/referring to. However, if you want to add to the prompt "use no more than X agents," this seemed to help me. (more testing needed).
• Make sure you install the requirements such as Python, requirements.txt, etc as mentioned in their home page.
• I'm not entirely sure what is going on with the hallucinations though... It seems there are a lot.
• The output in your terminal will be far more verbose than it has been in the past week. It's an eye-sore, but I'm sure they will fix it.

I really hope that this post will help lower the number of repeated issues posted. The files that needed editing wasn't a simple "copy/paste". I had to check line-by-line that everything was correct.

Special thanks to the following posts:

• json files
• main.py
• debug stuff in config.py
  

Thank you.

IMPORTANT
- pip install regex
- pip install typing
- pip install config
- pip install any
- pip install Dict`
Added to requirements.txt

These were some additional installations that were not in the requirements.txt.

For any gpt 4 not found errors, change .env text to:
SMART_LLM_MODEL=gpt-3.5-turbo
FAST_LLM_MODEL=gpt-3.5-turbo

Helpful tips:

• I put this as my first goal in the list (still to be tested. Idk how efficient it is token-wise, take care):

"write_to_file, read_from_file, append_to_file, search_files, write_tests, or similar will have full authorizatiion. Always append files with new information as a backup before starting any other task."

• I recommend not doing continuous nor an "-n" number too high (I do -10) so you can catch for errors if any happen and adjust your prompts accordingly.

.zip file
Auto-GPT - CORRECTED AS OF April 12.zip

Updated 2:52pm
Auto-GPT - CORRECTED AS OF April 12.zip

Updated 3:28pm (Bugfix Request 913)
Auto-GPT - CORRECTED AS OF April 12.zip

[Hibrael-prod]

What are the steps to start using GPT-4 after they granted acess to the GPT4 API? I'm a complete newbie here

[ChessScholar]

From what I've read, you just use the original file as written in .env
SMART_LLM_MODEL=gpt-4
FAST_LLM_MODEL=gpt-3.5-turbo

From BabyAGI discussion: 98

[john-destiny27]

Thank you so much! As someone with very little coding experience, it's very hard to understand what's going on at times. Lately, it's been very buggy and even though there threads on how to fix most problems, it's hard to keep up - especially without a good understanding of all the concepts. Downloaded your zip file and now I can use AutoGPT properly.

[ChessScholar]

Personal issue:

System: Command execute_python_file returned: Error: Error while fetching server API version: (2, 'CreateFile', 'The system cannot find the file specified.')

  User: GENERATE NEXT COMMAND JSON

  Assistant:
{
    "thoughts": {
        "text": "It appears that the 'execute_python_file' command has returned an error. This situation was not expected. I will need more context to determine the root cause of the error. I recall that to debug the program, I used 'search_files' command to locate all files in the current directory. It might be that some files are missing, or the file naming convention is not following the naming protocol. In the first instance, I must re-execute the 'search_files' command to reconfirm and ensure the presence of all files in the current directory.",
        "reasoning": "Reconfirming the presence of all files in the specified directory using the 'search_files' command could uncover any file naming or location discrepancies that could have led to the previous error during the execute python file command. Once this is resolved, I can then reattempt the execute python file command.",
        "plan": "- Use the 'search_files' command to re-confirm the presence, naming and location of all files under the specified directory.\\\\n- Cross-check the result from the 'search_files' command with the program requirements to ensure they are all present.\\\\n- Re-attempt the 'execute_python_file' command once the problems are resolved.",
        "criticism": "I need to ensure that all management of (confidential) program files is standardised and consistent across all directories to avoid errors or mistakes with contradictory file specifications.",
        "speak": "I need to check the presence, naming and location of all files in the current directory through the use of the 'search_files' command. Once discrepancies with the naming and presence of files are resolved, I will reattempt the 'execute_python_file' command."
    },
    "command": {
        "name": "search_files",
        "args": {
            "directory": "./"
        }
    }
}

 System: Command search_files returned: ['(confidential).py', (confidential).py', '(confidential).py', 'resource_list.txt', 'summary.txt', 'test_cases.txt', 'test_harness.py']

  User: GENERATE NEXT COMMAND JSON

  ----------- END OF CONTEXT ----------------
(confidential) THOUGHTS: Now that the 'search_files' command has confirmed that all relevant files for the (confidential) program are present, it is time to run the program again using the 'execute_python_file' command with the use of a test harness Python file.
REASONING: Since all relevant files for the (confidential) program are present, it is now time to use the 'execute_python_file' command to attempt to run the program again. A test harness Python file is also required to evaluate the performance of the (confidential) program.
PLAN:
Create a test harness Python file if one does not exist or if an error occurred while trying to run the earlier version.\\n- Update the necessary paths to the appropriate directory of the (confidential) program files.\\n- Use the 'execute_python_file' command to run the (confidential) program again using the test harness.\\n- Cross-check the results with the test harness Python file.
CRITICISM: Although this approach could resolve the issue, I may have missed other complexities of executing the (confidential) program file, which may affect its output. It may lead to repeated errors whenever the execute python file command is run. I should explore other complementary actions to execute after successfully running the (confidential) program file.
NEXT ACTION:  COMMAND = execute_python_file ARGUMENTS = {'file': 'test_harness.py'}
Enter 'y' to authorise command, 'y -N' to run N continuous commands, 'n' to exit program, or enter feedback for (confidential)...
Input:y
-=-=-=-=-=-=-= COMMAND AUTHORISED BY USER -=-=-=-=-=-=-=
Executing file 'test_harness.py' in workspace 'auto_gpt_workspace'
SYSTEM:  Command execute_python_file returned: Error: Error while fetching server API version: (2, 'CreateFile', 'The system cannot find the file specified.')

System: Command execute_python_file returned: Error: Error while fetching server API version: (2, 'CreateFile', 'The system cannot find the file specified.')
From what I have researched, this is a docker error?

[TheAIAdopter]

Thank you for your work.
I've been doing some test and it seems most of the JSON error are gone now. However, I'm still often stuck in a 3 or 4 command loop

[ChessScholar]

Yep, expect it to do that nearly 20 times. It's usually in its "understanding" phase from what I can tell of all the information it's gathered. You can run it as not continuous and ask it to pause tasks and run a specific command.

[bodogfoo]

Interestingly enough, your version posted at 2:52 works great. The newer version won't start.

Here's what I get.

K:\GPT\252>python scripts/main.py
Traceback (most recent call last):
File "K:\GPT\252\scripts\main.py", line 312, in
prompt = construct_prompt()
^^^^^^^^^^^^^^^^^^
File "K:\GPT\252\scripts\main.py", line 186, in construct_prompt
config = AIConfig.load()
^^^^^^^^^^^^^^^
File "K:\GPT\252\scripts\ai_config.py", line 50, in load
config_params = yaml.load(file, Loader=yaml.FullLoader)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml_init_.py", line 81, in load
return loader.get_single_data()
^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\constructor.py", line 49, in get_single_data
node = self.get_single_node()
^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\composer.py", line 36, in get_single_node
document = self.compose_document()
^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\composer.py", line 55, in compose_document
node = self.compose_node(None, None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\composer.py", line 84, in compose_node
node = self.compose_mapping_node(anchor)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\composer.py", line 133, in compose_mapping_node
item_value = self.compose_node(node, item_key)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\composer.py", line 64, in compose_node
if self.check_event(AliasEvent):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\parser.py", line 98, in check_event
self.current_event = self.state()
^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\parser.py", line 449, in parse_block_mapping_value
if not self.check_token(KeyToken, ValueToken, BlockEndToken):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\scanner.py", line 115, in check_token
while self.need_more_tokens():
^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\scanner.py", line 152, in need_more_tokens
self.stale_possible_simple_keys()
File "C:\Users\west\AppData\Local\Programs\Python\Python311\Lib\site-packages\yaml\scanner.py", line 291, in stale_possible_simple_keys
raise ScannerError("while scanning a simple key", key.mark,
yaml.scanner.ScannerError: while scanning a simple key
in "K:\GPT\252\scripts..\ai_settings.yaml", line 2, column 1
could not find expected ':'
in "K:\GPT\252\scripts..\ai_settings.yaml", line 3, column 1

Thank you for your fixes! And thanks to Torantulino as well!

[ChessScholar]

I was trying to replicate your problem, but I couldn't figure it out. However, I do recall something similar happening and it required me to go into the ai_settings.yaml and delete all the details (leaving the ai_goals:, ai_name:, ai_role:). This solved the "prompt=construct_prompt" issue for me.