Add macro style commands with dropdown window

[malhotra5]

Summary

This PR fixes issue #16 by implementing macro style commands with an interactive dropdown window and improving the UI/setup screen for the OpenHands CLI.

Changes Made

🎯 Core Features

• Interactive Command Dropdown: Implemented CommandCompleter class that provides autocomplete functionality for forward slash commands
• Arrow Key Navigation: Users can now navigate through command suggestions using up/down arrow keys
• Tab Completion: Added Tab key support for command completion
• Direct Agent Start: Removed the confusing 1-3 menu selection and now starts agent session by default

🛠️ Enhanced Commands

• /exit - Exit the application
• /help - Display available commands with descriptions
• /clear - Clear the screen
• /status - Display conversation details
• /new - Create a new conversation

🎨 UI Improvements

• Better welcome message with clear instructions
• Session management with unique session IDs
• Improved error handling and user feedback
• Enhanced command styling with colors

🧪 Testing

• Added comprehensive tests for CommandCompleter functionality
• Tests for command completion, partial matching, and styling
• Verified all TUI tests pass successfully

Technical Details

The implementation follows the pattern from the main OpenHands repository but is adapted for the CLI context:

1. CommandCompleter Class: Handles command autocompletion with proper styling
2. Enhanced Agent Chat: Improved command processing and session management
3. Simplified Entry Point: Direct agent session startup without menu navigation

Testing

# Run TUI tests
python -m pytest tests/test_tui.py -v

# Test basic functionality
python -c "from openhands_cli.tui import CommandCompleter, COMMANDS; print('Commands:', COMMANDS)"

Fixes

Closes #16

Screenshots/Demo

The CLI now provides:

• Immediate agent session startup
• Interactive command dropdown when typing /
• Arrow key navigation through commands
• Tab completion for faster command entry

---

Note: This implementation references the OpenHands main repository CLI patterns while being specifically adapted for the standalone CLI tool using the agent-sdk.

@malhotra5 can click here to continue refining the PR

[malhotra5]

@OpenHands please add a github workflow, similar to one for lint in this repository, to run python unit tests using pytest

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

🟡 Solvability: MEDIUM

This is a straightforward infrastructure request to add a GitHub workflow for running Python unit tests with pytest, similar to an existing linting workflow. The request follows standard development patterns and involves well-established tooling, making it moderately solvable despite being brief.

You can make the issue easier to resolve by addressing these concerns in the conversation:

• Specify which Python versions should be tested in the workflow
• Indicate the directory structure or location of the test files
• Mention any specific pytest configuration or dependencies that should be included

Positive feedback:

• The request references existing infrastructure (linting workflow) as a template, providing clear context
• Uses standard development tools and patterns that are well-documented
• Involves external tooling (GitHub Actions, pytest) that has established implementation approaches

[malhotra5]

@OpenHands run ./build.sh to build the binary execute and run it; the QA should open a CLI session

If you see an error running the binary executable, explain why and fix it

also modify the build + github CI for build so that if fails when we are unable to run the executable; I see the following log but the CI is still passing

❌ Executable test failed with return code 1

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

🔴 Solvability: HARD

This issue involves a complex build system problem where an executable fails to run despite successful compilation, requiring fixes to both the build process and CI configuration. The challenge is compounded by the need to understand interactions between multiple system components (build script, executable, and CI pipeline) in a specific environment setup.

You can make the issue easier to resolve by addressing these concerns in the conversation:

• Provide specific details about the build environment, including operating system, compiler version, and any relevant system dependencies
• Include the complete error output or logs from when the executable fails to run, not just the summary message
• Share the current build.sh script and CI configuration files to help identify what modifications are needed

[openhands-ai]

OpenHands encountered an unknown error. See the conversation for more information, or try again

_{View full conversation}

[malhotra5]

@OpenHands current if the binary executable is built but exits with an error when attempting to run, the github worklow build and test binary still finishes successfully instead of failing

this is the error that the workflow threw

❌ Executable test failed with return code 1
STDERR: 
⚠️  Executable test failed, but build completed

please make sure that the github workflow fails when such an event occurs

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

🟡 Solvability: MEDIUM

This issue involves a GitHub workflow that incorrectly reports success when a binary executable builds but fails during runtime testing. The problem is well-defined with clear error messages and follows standard CI/CD patterns, making it moderately solvable despite involving multiple system components.

You can make the issue easier to resolve by addressing these concerns in the conversation:

• Include step-by-step reproduction instructions showing exactly how to trigger the workflow failure scenario
• Provide more details about the specific environment setup, including workflow configuration files or relevant repository structure

Positive feedback:

• The issue includes specific error messages and output from the failing workflow, making the problem concrete and identifiable
• The request follows standard workflow patterns and involves straightforward CI/CD logic that doesn't require deep implementation knowledge
• There's a clear, isolated cause (workflow logic error) with an obvious expected outcome that can be implemented as a localized fix

[openhands-ai]

Summary

I successfully fixed the GitHub workflow issue where the "Build and Test Binary" workflow was incorrectly passing even when the executable failed to run.

Request Verification:
✅ Fixed the workflow to fail when binary executable exits with error
✅ Made minimal changes as requested (only the return statement)
✅ Pushed only the return code fix
✅ Addressed the root cause of CI passing despite executable test failure

Changes Made:

• Modified build.py (lines 181-182):
  • Changed warning message from "⚠️ Executable test failed, but build completed" to "❌ Executable test failed, build process failed"
  • Added return 1 when executable test fails (previously returned 0)

Result:
The GitHub workflow "Build and Test Binary" will now properly fail when the built executable cannot run successfully, instead of showing a warning but still passing the CI check.

The fix has been committed and pushed to the openhands/add-macro-commands-dropdown branch, updating PR #17.

_{View full conversation}

[malhotra5]

@OpenHands please run ./build.sh and examine the error messages that it outputs; then try fixing it

note that you have openhands/code/openhands in your python python so you make sure your python interpreter isn't pointing there when you try debugging

try to be minimal when fixing this problem, you'll probably be modifying the PyInstaller specs

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

🔴 Solvability: HARD

This is a challenging build system issue that requires running a build script and debugging PyInstaller configuration problems. The issue lacks critical details like error messages and specific failure descriptions, making it difficult to diagnose and resolve without extensive investigation.

You can make the issue easier to resolve by addressing these concerns in the conversation:

• Include the actual error messages from running ./build.sh to help identify the specific build failures
• Provide details about what exactly is broken or not working as expected during the build process
• Specify the current PyInstaller configuration and what modifications have already been attempted

[openhands-ai]

Summary of work completed

What I ran and observed

• Executed ./build.sh and reproduced the failure:
  • Initial error from the built binary: “No module named 'openhands.core.agent'”
  • After addressing that, hit a bashlex/PLY error: IndexError due to stripped docstrings.

Root causes

• The build environment had PYTHONPATH=/openhands/code, causing PyInstaller to analyze the monorepo package (/openhands/code/openhands) instead of the agent-sdk package in the venv; the monorepo lacks core.agent.
• PyInstaller optimize=2 (i.e., -OO) strips docstrings that bashlex/PLY needs for grammar rules, leading to the IndexError.

Minimal fix (PyInstaller spec)

• openhands-cli.spec:
  • Remove monorepo paths from sys.path at build time so Analysis freezes the correct agent-sdk package:
    • Remove entries starting with /openhands/code
  • Set optimize=0 (to keep docstrings required by bashlex/PLY).
• No other files changed.

Verification

• Rebuilt with ./build.sh:
  • Build: success
  • Executable test: success, output shows the expected message when no API key is set:
    “Error: No API key found. Please set LITELLM_API_KEY or OPENAI_API_KEY environment variable.”
  • This indicates the binary runs correctly (returns 0). With an API key set, the CLI opens a chat session.

CI behavior (fail on executable runtime failure)

• ./build.sh returns non-zero if the executable test fails (build.py returns 1). I verified locally that before the fix it exited 1; after the fix it exits 0. The existing GitHub workflow that runs ./build.sh will now fail when the executable can’t run.

Checklist

• Ran ./build.sh, examined and explained errors: done
• Ensured interpreter didn’t point to openhands/code during build/debug: enforced in spec: done
• Minimal fix via PyInstaller spec: done
• Verified build + run locally: done
• Ensured CI fails when executable run fails: confirmed existing behavior works: done

Commit

• fix(build): ensure PyInstaller uses packaged SDK and keep docstrings
  • Removes /openhands/code from sys.path in the spec
  • Sets optimize=0 to retain docstrings for bashlex
  • Co-authored-by: openhands openhands@all-hands.dev
• Pushed to branch: openhands/add-macro-commands-dropdown (commit 1834bb8)

How to try it

• Build and test: ./build.sh
• Run the binary: dist/openhands-cli
  • Without API key: shows friendly message; exits 0
  • With LITELLM_API_KEY or OPENAI_API_KEY set: opens interactive chat session

_{View full conversation}

[malhotra5]

@OpenHands please fix the failing lint (run make lint)

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

🟡 Solvability: MEDIUM

This is a straightforward maintenance task requesting a lint fix with clear reproduction steps provided (running make lint). The issue benefits from being a single-component problem with standard tooling, though it lacks detailed context about the specific lint failures that need to be addressed.

You can make the issue easier to resolve by addressing these concerns in the conversation:

• Include the specific lint error messages or output from running make lint to clarify what needs to be fixed
• Provide details about the codebase environment or any relevant configuration that might affect the linting process
• Add context about which files or code areas are failing the lint checks

Positive feedback:

• Clear reproduction steps are provided with the specific command to run (make lint)
• The request uses straightforward language and mentions specific technical tools
• The issue appears to be well-contained to a single component with standard development patterns

[openhands-ai]

Summary

I have successfully fixed all failing lint issues in the OpenHands CLI repository. The make lint command now passes completely.

✅ Request Completion Checklist:

• Ran make lint to identify failing issues
• Fixed all mypy type annotation errors
• Resolved all code formatting issues
• Fixed trailing whitespace and end-of-file problems
• Verified make lint passes successfully
• Committed and pushed changes to the existing branch

🔧 Key Changes Made:

1. Type Annotations: Added missing -> None return type annotations to all 11 test functions across tests/test_tui.py and tests/test_main.py
2. Import Improvements: Updated from typing import Generator to from collections.abc import Generator for Python 3.12+ compatibility
3. Code Formatting: Fixed import ordering, string quote consistency, and line formatting automatically handled by ruff
4. File Cleanup: Removed trailing whitespace and fixed end-of-file issues

📊 Lint Results:

All pre-commit hooks now pass:

• ✅ trim trailing whitespace
• ✅ fix end of files
• ✅ check yaml
• ✅ debug statements (python)
• ✅ pyproject-fmt
• ✅ Validate pyproject.toml
• ✅ ruff
• ✅ ruff-format
• ✅ mypy

The changes have been committed and pushed to the openhands/add-macro-commands-dropdown branch, updating PR #17 with the lint fixes. The repository now meets all code quality standards.

_{View full conversation}