Fix issues with file reading and writing with Python code

[erik-megarad]

Background

Several times I've had Auto-GPT fail to recognize "does not exist" error for files, and it will attempt other fixes such as analyzing the code (for a file that does not exist). I've also had issues with python code not being executed in the working directory, which means that code written by Auto-GPT that reads and writes files will usually not work (since they use relative paths).

These issues are related:
#4439
#4413
#4366
#4360
#3583

Changes

• Change the error messages for file not found errors so that they're easier to recognize as such
• Set cwd when executing python code so that relative paths work as expected

Documentation

• In-code comments

Test Plan

• Wrote a new test for reading a file that does not exist (was previously untested)
• Wrote a new test for executing a python file that does not exist (was previously untested)

PR Quality Checklist

• My pull request is atomic and focuses on a single change.
• I have thoroughly tested my changes with multiple different prompts.
• I have considered potential risks and mitigations for my changes.
• I have documented my changes clearly and comprehensively.
• I have not snuck in any "extra" small tweaks changes.
• I have run the following commands against my code to ensure it passes our linters:

  black .
  isort .
  mypy
  autoflake --remove-all-unused-imports --recursive --ignore-init-module-imports autogpt tests --in-place

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 11 hours (more than 100, code: "api-deployments-free-per-day").

[erik-megarad]

Example of the new error being recognized:

WEATHERGPT THOUGHTS:  I have found two reliable sources of weather data for Paris, Texas on October 10th, 1991. I will cross-reference the data from both sources to ensure accuracy and reliability.
REASONING:  Cross-referencing multiple sources of weather data will help ensure the accuracy and reliability of the information.
PLAN:
-  Retrieve the weather data from both sources.
-  Cross-reference the data to ensure accuracy and reliability.
-  Present the weather information in a clear and concise manner.
CRITICISM:  I need to be more efficient in cross-referencing multiple sources of weather data.
SPEAK:  I have found two reliable sources of weather data for Lilburn, Georgia on October 12th, 1991. Please wait a moment while I cross-reference the data to ensure accuracy and reliability.
NEXT ACTION:  COMMAND = execute_python_file ARGUMENTS = {'filename': '/app/autogpt/auto_gpt_workspace/cross_reference_weather_data.py'}
  Enter 'y' to authorise command, 'y -N' to run N continuous commands, 's' to run self-feedback commands, 'n' to exit program, or enter feedback for WeatherGPT...
  Asking user via keyboard...
Input:y
-=-=-=-=-=-=-= COMMAND AUTHORISED BY USER -=-=-=-=-=-=-=
  Executing python file '/app/autogpt/auto_gpt_workspace/cross_reference_weather_data.py' in working directory '/app/autogpt/auto_gpt_workspace'
SYSTEM:  Command execute_python_file returned: python: can't open file '/app/autogpt/auto_gpt_workspace/cross_reference_weather_data.py': [Errno 2] No such file or directory
WEATHERGPT THOUGHTS:  I need to create a Python script to cross-reference the weather data from both sources. I will create a new file and write the necessary code to cross-reference the data.
REASONING:  Creating a Python script will allow me to automate the process of cross-referencing the weather data from both sources.
PLAN:
-  Create a new Python file.
-  Write the necessary code to cross-reference the weather data from both sources.
-  Save the file.
CRITICISM:  I need to ensure that the code I write is efficient and accurate.
SPEAK:  I need to create a Python script to cross-reference the weather data from both sources. Please wait a moment while I create the necessary file and write the code.

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 8 hours (more than 100, code: "api-deployments-free-per-day").

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Comments	Updated (UTC)
docs	⬜️ Ignored (Inspect)			Jun 6, 2023 7:28pm

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.14 🎉

Comparison is base (20a4922) 69.57% compared to head (743fa25) 69.71%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4567      +/-   ##
==========================================
+ Coverage   69.57%   69.71%   +0.14%     
==========================================
  Files          72       72              
  Lines        3556     3560       +4     
  Branches      569      569              
==========================================
+ Hits         2474     2482       +8     
+ Misses        892      889       -3     
+ Partials      190      189       -1     

Impacted Files	Coverage Δ
autogpt/commands/execute_code.py	65.55% <100.00%> (+1.60%)	⬆️
autogpt/commands/file_operations_utils.py	92.85% <100.00%> (+2.04%)	⬆️

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[Boostrix]

Several times I've had Auto-GPT fail to recognize "does not exist" error for files,

that's true, I ended up adding commands to check if a directory/file exists and made this available so that the LLM can use it.
I also changed the prompt to specifically state "do not try to read or execute files that do not exist", and I told it "do not overwrite existing files"
So, now it needs to check first if a file exists before reading/executing it.

it will attempt other fixes such as analyzing the code (for a file that does not exist).

analyze_code seems to be called for all sorts of stuff where it wants to "analyze" a problem (not just code).
But if you prime it to check first if a file/folder exists before accessing it, that basically solves the problem. Especially in combination with a file_exists command:

+@command("file_exists", "Check if File or Directory Exists", '"path": "<path>"')
+def file_exists(path: str) -> int:
+    """Check if a file or directory exists.
+
+    Args:
+        path (str): The path of the file or directory to check.
+
+    Returns:
+        int: 1 if the file or directory exists, 0 otherwise.
+    """
+    full_path = os.path.join(CFG.workspace_path, path)
+    if os.path.exists(full_path):
+        return 1
+    else:
+        return 0

That is one of the reasons why I suggested elsewhere that we should be able to chain commands together, as in:

file_exists foo.py && read_file foo.py

note that this would make it possible to run these chains locally, without having to go through the LLM at all: #4157

[Pwuts]

Thanks for picking this up!

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 43 minutes (more than 100, code: "api-deployments-free-per-day").

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 29 minutes (more than 100, code: "api-deployments-free-per-day").

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 22 minutes (more than 100, code: "api-deployments-free-per-day").

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 20 minutes (more than 100, code: "api-deployments-free-per-day").

[vercel]

Deployment failed with the following error:

Resource is limited - try again in 11 minutes (more than 100, code: "api-deployments-free-per-day").

[maxfield-allison]

Thank you for your service!