Fix remote sync bugs and improve error handling

Fixes two critical issues:
1. Remote clone/sync was never triggered because wmill sync ran before
   checking if workspace was empty. Reordered flow to check/clone/sync
   git repo BEFORE running wmill sync.

2. Push failures were reported as success because GitPython's push()
   doesn't raise exceptions for rejections. Added explicit checking of
   push result flags (ERROR and REJECTED).

Additional improvements:
- When workspace has files but no .git, delete contents and clone from
  remote to ensure proper sync state
- All three cases now properly sync with remote before Windmill overwrites

New flow:
1. Check workspace state and init/clone/sync git repo
2. Run wmill sync (Windmill overwrites files)
3. Commit and push (with proper error detection)

Co-Authored-By: Claude <noreply@anthropic.com>

Dockerfile

@@ -2,15 +2,16 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
-# Install system dependencies
+# Install system dependencies including Node.js
 RUN apt-get update && apt-get install -y \
     git \
     curl \
+    && curl -fsSL https://deb.nodesource.com/setup_20.x | bash - \
+    && apt-get install -y nodejs \
     && rm -rf /var/lib/apt/lists/*
 
-# Install wmill CLI
+# Install wmill CLI via npm
-RUN curl -L https://github.com/windmill-labs/windmill/releases/latest/download/wmill-linux-amd64 -o /usr/local/bin/wmill \
+RUN npm install -g windmill-cli
-    && chmod +x /usr/local/bin/wmill
 
 # Copy requirements and install Python dependencies
 COPY requirements.txt .

app/sync.py

@@ -3,6 +3,7 @@
 Core sync logic for pulling Windmill workspace and pushing to Git.
 """
 import os
+import shutil
 import subprocess
 import logging
 from pathlib import Path
@@ -59,17 +60,18 @@ def run_wmill_sync(config: Dict[str, Any]) -> bool:
 
     logger.info(f"Syncing Windmill workspace '{workspace}' from {WINDMILL_BASE_URL}")
 
-    env = os.environ.copy()
-    env['WM_BASE_URL'] = WINDMILL_BASE_URL
-    env['WM_TOKEN'] = windmill_token
-    env['WM_WORKSPACE'] = workspace
-
     try:
-        # Run wmill sync in the workspace directory
+        # Run wmill sync in the workspace directory with explicit flags
+        # Note: When using --base-url, --token and --workspace are required
         result = subprocess.run(
-            ['wmill', 'sync', 'pull', '--yes'],
+            [
+                'wmill', 'sync', 'pull',
+                '--base-url', WINDMILL_BASE_URL,
+                '--token', windmill_token,
+                '--workspace', workspace,
+                '--yes'
+            ],
             cwd=WORKSPACE_DIR,
-            env=env,
             capture_output=True,
             text=True,
             check=True
@@ -85,9 +87,115 @@ def run_wmill_sync(config: Dict[str, Any]) -> bool:
         raise RuntimeError(f"Failed to sync from Windmill: {e.stderr}")
 
 
+def is_workspace_empty() -> bool:
+    """
+    Check if workspace directory is empty or not initialized as a git repo.
+
+    Returns:
+        bool: True if directory is empty or not a git repository
+    """
+    git_dir = WORKSPACE_DIR / '.git'
+
+    # Check if .git directory exists
+    if not git_dir.exists():
+        # Check if workspace is completely empty or has no meaningful content
+        workspace_contents = list(WORKSPACE_DIR.iterdir())
+        return len(workspace_contents) == 0
+
+    return False
+
+
+def clone_remote_repository(config: Dict[str, Any]) -> Repo:
+    """
+    Clone remote repository to workspace directory.
+
+    Args:
+        config: Configuration dictionary containing git_remote_url, git_token, git_branch
+
+    Returns:
+        Repo: GitPython repository object
+
+    Raises:
+        RuntimeError: If clone fails
+    """
+    git_remote_url = config['git_remote_url']
+    git_token = config['git_token']
+    git_branch = config.get('git_branch', 'main')
+    git_user_name = config.get('git_user_name', 'Windmill Git Sync')
+    git_user_email = config.get('git_user_email', 'windmill@example.com')
+
+    authenticated_url = get_authenticated_url(git_remote_url, git_token)
+
+    logger.info(f"Cloning remote repository from {git_remote_url}")
+
+    try:
+        # Clone the repository
+        repo = Repo.clone_from(authenticated_url, WORKSPACE_DIR)
+
+        # Configure user
+        repo.config_writer().set_value("user", "name", git_user_name).release()
+        repo.config_writer().set_value("user", "email", git_user_email).release()
+
+        # Check if the specified branch exists
+        try:
+            repo.git.checkout(git_branch)
+            logger.info(f"Checked out existing branch '{git_branch}'")
+        except GitCommandError:
+            # Branch doesn't exist, create it as orphan
+            logger.info(f"Branch '{git_branch}' doesn't exist, creating new branch")
+            repo.git.checkout('--orphan', git_branch)
+            # Remove all files from staging (orphan branch starts with staged files)
+            try:
+                repo.git.rm('-rf', '.')
+            except GitCommandError:
+                # If rm fails (no files to remove), that's fine
+                pass
+
+        logger.info("Repository cloned successfully")
+        return repo
+
+    except GitCommandError as e:
+        logger.error(f"Failed to clone repository: {str(e)}")
+        raise RuntimeError(f"Failed to clone remote repository: {str(e)}")
+
+
+def sync_local_with_remote(repo: Repo, config: Dict[str, Any]) -> None:
+    """
+    Sync local repository with remote (fetch and hard reset).
+
+    Args:
+        repo: GitPython Repo object
+        config: Configuration dictionary containing git_branch
+
+    Raises:
+        RuntimeError: If sync fails
+    """
+    git_branch = config.get('git_branch', 'main')
+
+    logger.info(f"Syncing local repository with remote branch '{git_branch}'")
+
+    try:
+        # Fetch from remote
+        origin = repo.remote('origin')
+        origin.fetch()
+        logger.info("Fetched from remote")
+
+        # Reset local branch to match remote
+        try:
+            repo.git.reset('--hard', f'origin/{git_branch}')
+            logger.info(f"Reset local branch to match origin/{git_branch}")
+        except GitCommandError as e:
+            # Branch might not exist on remote yet, which is fine
+            logger.info(f"Branch '{git_branch}' doesn't exist on remote yet, will be created on push")
+
+    except GitCommandError as e:
+        logger.error(f"Failed to sync with remote: {str(e)}")
+        raise RuntimeError(f"Failed to sync local repository with remote: {str(e)}")
+
+
 def init_or_update_git_repo(config: Dict[str, Any]) -> Repo:
     """
-    Initialize Git repository or open existing one.
+    Initialize Git repository, clone from remote if needed, or open existing one.
 
     Args:
         config: Configuration dictionary containing optional git_user_name and git_user_email
@@ -100,18 +208,34 @@ def init_or_update_git_repo(config: Dict[str, Any]) -> Repo:
 
     git_dir = WORKSPACE_DIR / '.git'
 
+    # Check if workspace is empty/uninitialized
+    if is_workspace_empty():
+        # Try to clone from remote
+        logger.info("Workspace is empty, attempting to clone from remote")
+        return clone_remote_repository(config)
+
+    # Repository already exists
     if git_dir.exists():
         logger.info("Opening existing Git repository")
         repo = Repo(WORKSPACE_DIR)
-    else:
-        logger.info("Initializing new Git repository")
-        repo = Repo.init(WORKSPACE_DIR)
 
-        # Configure user
+        # Sync with remote before continuing
-        repo.config_writer().set_value("user", "name", git_user_name).release()
+        sync_local_with_remote(repo, config)
-        repo.config_writer().set_value("user", "email", git_user_email).release()
 
         return repo
+    else:
+        # Workspace has files but no git repo - delete contents and clone
+        logger.info("Workspace has files but no git repository, cleaning and cloning from remote")
+
+        # Delete all contents in workspace
+        for item in WORKSPACE_DIR.iterdir():
+            if item.is_dir():
+                shutil.rmtree(item)
+            else:
+                item.unlink()
+
+        logger.info("Workspace cleaned, attempting to clone from remote")
+        return clone_remote_repository(config)
 
 
 def commit_and_push_changes(repo: Repo, config: Dict[str, Any]) -> bool:
@@ -159,9 +283,21 @@ def commit_and_push_changes(repo: Repo, config: Dict[str, Any]) -> bool:
 
         # Push to remote
         logger.info(f"Pushing to {git_remote_url} (branch: {git_branch})")
-        origin.push(refspec=f'HEAD:{git_branch}', force=False)
+        push_info = origin.push(refspec=f'HEAD:{git_branch}', force=False)
-        logger.info("Push completed successfully")
 
+        # Check push results for errors
+        if push_info:
+            for info in push_info:
+                if info.flags & info.ERROR:
+                    error_msg = f"Push failed: {info.summary}"
+                    logger.error(error_msg)
+                    raise RuntimeError(error_msg)
+                elif info.flags & info.REJECTED:
+                    error_msg = f"Push rejected (non-fast-forward): {info.summary}"
+                    logger.error(error_msg)
+                    raise RuntimeError(error_msg)
+
+        logger.info("Push completed successfully")
         return True
 
     except GitCommandError as e:
@@ -192,12 +328,12 @@ def sync_windmill_to_git(config: Dict[str, Any]) -> Dict[str, Any]:
 
         workspace = config.get('workspace', 'admins')
 
-        # Pull from Windmill
+        # Initialize/update Git repo (must happen BEFORE wmill sync to clone if needed)
-        run_wmill_sync(config)
-
-        # Initialize/update Git repo
         repo = init_or_update_git_repo(config)
 
+        # Pull from Windmill (overwrites files with Windmill workspace content)
+        run_wmill_sync(config)
+
         # Commit and push changes
         has_changes = commit_and_push_changes(repo, config)