fix: stream and capture ipython output

[ErikBjare]

Was needed to make the browser tool able to update the user on progress via logging/console messages. (noticed when fixing #358)

---

Important

Enhance IPython output handling in gptme/tools/python.py by replacing InteractiveShellEmbed with InteractiveShell and adding real-time output streaming.

• IPython Handling:
  • Replace InteractiveShellEmbed with InteractiveShell for IPython instance.
  • Modify _get_ipython() to use InteractiveShell.
• Output Streaming:
  • Add TeeIO class to manage output streaming and filter IPython result prompts.
  • Introduce capture_and_display() context manager to capture and display stdout and stderr in real-time.
• Function Updates:
  • Update execute_python() to use capture_and_display() for real-time output capture and display.

^{This description was created by for 5f90e04. It will automatically update as commits are pushed.}

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to 5dd00ed in 1 minute and 9 seconds

More details

• Looked at 73 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 3 drafted comments based on config settings.

1. gptme/tools/python.py:92

• Draft comment:
  Consider using a thread-safe alternative to io.StringIO for TeeIO to avoid potential issues with multi-threaded code execution.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The TeeIO class is designed to capture and display output in real-time, but it inherits from io.StringIO, which is not thread-safe. This could lead to issues if the code execution involves multi-threading.

2. gptme/tools/python.py:102

• Draft comment:
  Replacing sys.stdout and sys.stderr can interfere with other parts of the program. Consider using a more isolated approach to capture output.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The capture_and_display function replaces sys.stdout and sys.stderr with TeeIO instances. This could interfere with other parts of the program that rely on standard output and error streams.

3. gptme/tools/python.py:135

• Draft comment:
  Ensure result.error_in_exec is not None before accessing its __traceback__ attribute to avoid potential AttributeError.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The execute_python function uses result.error_in_exec to check for execution errors, but it doesn't handle the case where result.error_in_exec might be None. This could lead to an AttributeError if result.error_in_exec is None.

Workflow ID: wflow_E5A9uUkFgeMUftu9

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 97.67442% with 1 line in your changes missing coverage. Please review.

Project coverage is 71.23%. Comparing base (99912f3) to head (5f90e04).
Report is 3 commits behind head on master.

✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
gptme/tools/python.py	97.67%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #357      +/-   ##
==========================================
+ Coverage   71.07%   71.23%   +0.16%     
==========================================
  Files          68       68              
  Lines        5241     5271      +30     
==========================================
+ Hits         3725     3755      +30     
  Misses       1516     1516              

Flag	Coverage Δ
anthropic/claude-3-haiku-20240307	69.47% <97.67%> (+0.15%)	⬆️
openai/gpt-4o-mini	68.05% <88.37%> (+0.35%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ellipsis-dev]

👍 Looks good to me! Incremental review on c6d96a7 in 36 seconds

More details

• Looked at 115 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 3 drafted comments based on config settings.

1. gptme/tools/python.py:105

• Draft comment:
  The import statement for ExecutionResult is unnecessary as it is not used in the function. Consider removing it to clean up the code.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The import statement for ExecutionResult is not used in the function, which is unnecessary and should be removed.

2. gptme/tools/python.py:66

• Draft comment:
  The TeeIO class is defined twice in the code. Please remove the redundant definition to avoid confusion and maintain code clarity.
• Reason this comment was not posted:
  Comment looked like it was already resolved.

3. gptme/tools/python.py:86

• Draft comment:
  The capture_and_display function is defined twice in the code. Please remove the redundant definition to avoid confusion and maintain code clarity.
• Reason this comment was not posted:
  Comment looked like it was already resolved.

Workflow ID: wflow_GylMwy83YbWl0WdA

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.

[ellipsis-dev]

👍 Looks good to me! Incremental review on 5f90e04 in 15 seconds

More details

• Looked at 19 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 1 drafted comments based on config settings.

1. gptme/tools/python.py:153

• Draft comment:
  Good use of checking if tb: before accessing tb.tb_lineno to prevent potential AttributeError. This change makes the code more robust.
• Reason this comment was not posted:
  Confidence changes required: 0%
  The code correctly handles the traceback by checking if tb is not None before accessing tb.tb_lineno. This prevents potential AttributeError if tb is None. The change from the previous version ensures that the code is more robust.

Workflow ID: wflow_29pxl52ZzakhLaEY

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.