-
Notifications
You must be signed in to change notification settings - Fork 127
Fix agent confusion about app commands (#30) #4045
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Establishes CLAUDE.md as single source of truth for operational commands (npm run dev, validate, deploy). Updates validation to dynamically discover package.json scripts instead of hardcoding. Apps.tmpl now references CLAUDE.md for complete guidance. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <[email protected]>
|
|
||
| **⚠️ CRITICAL**: Use ONLY these commands. Do not improvise alternatives. | ||
|
|
||
| ### Local Development |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
are you sure it is agent's concern? imo it should not know anything but validate
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
template must set the way it should be executed, dont you think?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is certainly the agent's concern. I regularly tell the agent to run the app locally and open it in the browser so I can inspect the progress of the app.
The agent needs to know how to build and run the app locally.
| What do you want to do? | ||
| │ | ||
| ├─ "Test/run locally" → npm run dev | ||
| ├─ "Check ready to deploy" → databricks experimental apps-mcp tools validate . |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
i'd hint invoke_databricks_cli reference here
|
Commit: e5ed0b2
17 failing tests:
Top 31 slowest tests (at least 2 minutes):
|
| errorPrefix: "Failed to run tests", | ||
| displayName: "Tests", | ||
| }, | ||
| {name: "install", command: "npm install", errorPrefix: "Failed to install dependencies", displayName: "Install"}, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why do we need to parse package.json?
AFAIK, this is functionally equivalent to the old implementation. The old implementation will always work even if thescrips are missing since we are running with --if-present
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
wanted to have a single source of truth, but I agree the code looks less clean , so will roll this back until I have cleaner idea of implementing this
experimental/apps-mcp/templates/appkit/template/{{.project_name}}/CLAUDE.md
Outdated
Show resolved
Hide resolved
Resolved conflicts in:
- experimental/apps-mcp/lib/prompts/apps.tmpl
- experimental/apps-mcp/templates/appkit/template/{{.project_name}}/CLAUDE.md
Kept main branch versions for consistency with updated documentation style.
|
|
||
| >>> [CLI] bundle deploy | ||
| Building python_artifact... | ||
| Error: build failed python_artifact, error: exit status 127, output: bash: uv: command not found |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This seems to be specific output to your local run and will cause tests to fail on CI, could you revert?
|
|
||
| >>> find my_pydabs -mindepth 1 ! -name pyproject.toml ! -regex .*/resources.* -delete | ||
|
|
||
| >>> ruff format --quiet --diff --check my_pydabs |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
same as above, but you can just install ruff and it should succeed
|
closing due to feedback |
…irmation Two key issues addressed: 1. Agent was opening deployed URLs instead of localhost when user asked to "open a dev copy" or see the app during development. 2. Agent was deploying too early without user confirmation. Changes: - apps.tmpl: Add localhost guidance and deployment confirmation requirement - CLAUDE.md: Add "Deployment Procedures" section with pre-deployment checklist - CLAUDE.md: Add "Local Development vs Deployed Apps" section with decision tree The agent will now: - Prefer localhost:8000 during active development - Ask "Ready to deploy to [environment]?" before deploying - Only use deployed URLs after deployment with user approval Addresses feedback from PR #4045 discussion.
…rmation Three key issues addressed: 1. Agent was opening deployed URLs instead of localhost when user asked to "open a dev copy" or see the app during development. 2. Agent was deploying too early without user confirmation. 3. Need to clarify npm run dev (development) vs npm start (production). Changes: - apps.tmpl: Add localhost guidance and deployment confirmation requirement - apps.tmpl: Clarify "npm run dev" ALWAYS for development, NEVER "npm start" - CLAUDE.md: Add "Deployment Procedures" section with pre-deployment checklist - CLAUDE.md: Add "Local Development vs Deployed Apps" section with decision tree - CLAUDE.md: Add visual box explaining "npm run dev vs npm start" The agent will now: - ALWAYS use "npm run dev" during development (NEVER npm start) - Prefer localhost:8000 during active development - Ask "Ready to deploy to [environment]?" before deploying - Only use deployed URLs after deployment with user approval Addresses feedback from PR #4045 discussion.
#4056) Fix apps-mcp agent behavior: localhost preference and deployment confirmation Two key issues addressed: 1. Agent was opening deployed URLs instead of localhost when user asked to "open a dev copy" or see the app during development. 2. Agent was deploying too early without user confirmation. Changes: - apps.tmpl: Add localhost guidance and deployment confirmation requirement - CLAUDE.md: Add "Deployment Procedures" section with pre-deployment checklist - CLAUDE.md: Add "Local Development vs Deployed Apps" section with decision tree The agent will now: - Prefer localhost:8000 during active development - Ask "Ready to deploy to [environment]?" before deploying - Only use deployed URLs after deployment with user approval Addresses feedback from PR #4045 discussion. ## Changes Prompt updates with checklist how to deploy when and how to run app. ## Why issue 21 | Victor | The deployment was really early, maybe a confirmation step first. issue 30 | Lennart | Agent is opening deployed apps rather than “dev copy” (unless I ask about localhost very explicitly) ## Tests make test
Establishes CLAUDE.md as single source of truth for operational commands (npm run dev, validate, deploy). Updates validation to dynamically discover package.json scripts instead of hardcoding. Apps.tmpl now references CLAUDE.md for complete guidance.