-
-
Notifications
You must be signed in to change notification settings - Fork 194
Closed as not planned
Labels
Description
Feature Description
Often, when I use the tool build_sim_id_ws the context is blown away with a single invocation. I'm guessing this is because the build log output is filled with things like compiler commands that don't really matter to my LLM agent. It would be nice if these tools could have a flag that the agent could set that would filter all the cruft out and just present the LLM with warnings, errors, test failures, etc.
My agent often resorts to commands like this instead:
cd /Users/cody/Desktop/MyProject && mkdir -p tmp/build_output && xcodebuild -workspace MyProject.xcworkspace -scheme MyProject -destination "id=[simulator id]" test > tmp/build_output/test_restructured_$(date +"%Y%m%d_%H%M%S").log 2>&1 && echo "Tests completed successfully" || echo "Tests failed - check log file"and if there are issues then it will further grep into the output file to find them. For reference some of my build logs are nearly 20MB of text.
Use Cases
- Building and testing iOS apps in a large codebase
Example Interactions
Prompt: Build and run the tests and fix the errors and warnings that you find.
Reactions are currently unavailable