autoforge-ai 0.1.2 → 0.1.4

package/server/services/assistant_chat_session.py

@@ -25,7 +25,7 @@ from .assistant_database import (
     create_conversation,
     get_messages,
 )
-from .chat_constants import API_ENV_VARS, ROOT_DIR
+from .chat_constants import ROOT_DIR
 
 # Load environment variables from .env file if present
 load_dotenv()
@@ -258,15 +258,11 @@ class AssistantChatSession:
         system_cli = shutil.which("claude")
 
         # Build environment overrides for API configuration
-        sdk_env: dict[str, str] = {}
-        for var in API_ENV_VARS:
-            value = os.getenv(var)
-            if value:
-                sdk_env[var] = value
-
-        # Determine model from environment or use default
-        # This allows using alternative APIs (e.g., GLM via z.ai) that may not support Claude model names
-        model = os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
+        from registry import get_effective_sdk_env
+        sdk_env = get_effective_sdk_env()
+
+        # Determine model from SDK env (provider-aware) or fallback to env/default
+        model = sdk_env.get("ANTHROPIC_DEFAULT_OPUS_MODEL") or os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
 
         try:
             logger.info("Creating ClaudeSDKClient...")

package/server/services/expand_chat_session.py

@@ -22,7 +22,7 @@ from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
 from dotenv import load_dotenv
 
 from ..schemas import ImageAttachment
-from .chat_constants import API_ENV_VARS, ROOT_DIR, make_multimodal_message
+from .chat_constants import ROOT_DIR, make_multimodal_message
 
 # Load environment variables from .env file if present
 load_dotenv()
@@ -154,16 +154,11 @@ class ExpandChatSession:
         system_prompt = skill_content.replace("$ARGUMENTS", project_path)
 
         # Build environment overrides for API configuration
-        # Filter to only include vars that are actually set (non-None)
-        sdk_env: dict[str, str] = {}
-        for var in API_ENV_VARS:
-            value = os.getenv(var)
-            if value:
-                sdk_env[var] = value
-
-        # Determine model from environment or use default
-        # This allows using alternative APIs (e.g., GLM via z.ai) that may not support Claude model names
-        model = os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
+        from registry import get_effective_sdk_env
+        sdk_env = get_effective_sdk_env()
+
+        # Determine model from SDK env (provider-aware) or fallback to env/default
+        model = sdk_env.get("ANTHROPIC_DEFAULT_OPUS_MODEL") or os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
 
         # Build MCP servers config for feature creation
         mcp_servers = {

package/server/services/process_manager.py

@@ -227,6 +227,46 @@ class AgentProcessManager:
         """Remove lock file."""
         self.lock_file.unlink(missing_ok=True)
 
+    def _cleanup_stale_features(self) -> None:
+        """Clear in_progress flag for all features when agent stops/crashes.
+
+        When the agent process exits (normally or crash), any features left
+        with in_progress=True were being worked on and didn't complete.
+        Reset them so they can be picked up on next agent start.
+        """
+        try:
+            from autoforge_paths import get_features_db_path
+            features_db = get_features_db_path(self.project_dir)
+            if not features_db.exists():
+                return
+
+            from sqlalchemy import create_engine
+            from sqlalchemy.orm import sessionmaker
+
+            from api.database import Feature
+
+            engine = create_engine(f"sqlite:///{features_db}")
+            Session = sessionmaker(bind=engine)
+            session = Session()
+            try:
+                stuck = session.query(Feature).filter(
+                    Feature.in_progress == True,  # noqa: E712
+                    Feature.passes == False,  # noqa: E712
+                ).all()
+                if stuck:
+                    for f in stuck:
+                        f.in_progress = False
+                    session.commit()
+                    logger.info(
+                        "Cleaned up %d stuck feature(s) for %s",
+                        len(stuck), self.project_name,
+                    )
+            finally:
+                session.close()
+                engine.dispose()
+        except Exception as e:
+            logger.warning("Failed to cleanup features for %s: %s", self.project_name, e)
+
     async def _broadcast_output(self, line: str) -> None:
         """Broadcast output line to all registered callbacks."""
         with self._callbacks_lock:
@@ -288,6 +328,7 @@ class AgentProcessManager:
                     self.status = "crashed"
                 elif self.status == "running":
                     self.status = "stopped"
+                self._cleanup_stale_features()
                 self._remove_lock()
 
     async def start(
@@ -320,6 +361,9 @@ class AgentProcessManager:
         if not self._check_lock():
             return False, "Another agent instance is already running for this project"
 
+        # Clean up features stuck from a previous crash/stop
+        self._cleanup_stale_features()
+
         # Store for status queries
         self.yolo_mode = yolo_mode
         self.model = model
@@ -359,12 +403,22 @@ class AgentProcessManager:
             # stdin=DEVNULL prevents blocking if Claude CLI or child process tries to read stdin
             # CREATE_NO_WINDOW on Windows prevents console window pop-ups
             # PYTHONUNBUFFERED ensures output isn't delayed
+            # Build subprocess environment with API provider settings
+            from registry import get_effective_sdk_env
+            api_env = get_effective_sdk_env()
+            subprocess_env = {
+                **os.environ,
+                "PYTHONUNBUFFERED": "1",
+                "PLAYWRIGHT_HEADLESS": "true" if playwright_headless else "false",
+                **api_env,
+            }
+
             popen_kwargs: dict[str, Any] = {
                 "stdin": subprocess.DEVNULL,
                 "stdout": subprocess.PIPE,
                 "stderr": subprocess.STDOUT,
                 "cwd": str(self.project_dir),
-                "env": {**os.environ, "PYTHONUNBUFFERED": "1", "PLAYWRIGHT_HEADLESS": "true" if playwright_headless else "false"},
+                "env": subprocess_env,
             }
             if sys.platform == "win32":
                 popen_kwargs["creationflags"] = subprocess.CREATE_NO_WINDOW
@@ -425,6 +479,7 @@ class AgentProcessManager:
                 result.children_terminated, result.children_killed
             )
 
+            self._cleanup_stale_features()
             self._remove_lock()
             self.status = "stopped"
             self.process = None
@@ -502,6 +557,7 @@ class AgentProcessManager:
         if poll is not None:
             # Process has terminated
             if self.status in ("running", "paused"):
+                self._cleanup_stale_features()
                 self.status = "crashed"
                 self._remove_lock()
             return False

package/server/services/spec_chat_session.py

@@ -19,7 +19,7 @@ from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
 from dotenv import load_dotenv
 
 from ..schemas import ImageAttachment
-from .chat_constants import API_ENV_VARS, ROOT_DIR, make_multimodal_message
+from .chat_constants import ROOT_DIR, make_multimodal_message
 
 # Load environment variables from .env file if present
 load_dotenv()
@@ -140,16 +140,11 @@ class SpecChatSession:
         system_cli = shutil.which("claude")
 
         # Build environment overrides for API configuration
-        # Filter to only include vars that are actually set (non-None)
-        sdk_env: dict[str, str] = {}
-        for var in API_ENV_VARS:
-            value = os.getenv(var)
-            if value:
-                sdk_env[var] = value
-
-        # Determine model from environment or use default
-        # This allows using alternative APIs (e.g., GLM via z.ai) that may not support Claude model names
-        model = os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
+        from registry import get_effective_sdk_env
+        sdk_env = get_effective_sdk_env()
+
+        # Determine model from SDK env (provider-aware) or fallback to env/default
+        model = sdk_env.get("ANTHROPIC_DEFAULT_OPUS_MODEL") or os.getenv("ANTHROPIC_DEFAULT_OPUS_MODEL", "claude-opus-4-5-20251101")
 
         try:
             self.client = ClaudeSDKClient(

package/server/websocket.py

@@ -640,9 +640,7 @@ class ConnectionManager:
         self._lock = asyncio.Lock()
 
     async def connect(self, websocket: WebSocket, project_name: str):
-        """Accept a WebSocket connection for a project."""
-        await websocket.accept()
-
+        """Register a WebSocket connection for a project (must already be accepted)."""
         async with self._lock:
             if project_name not in self.active_connections:
                 self.active_connections[project_name] = set()
@@ -727,16 +725,22 @@ async def project_websocket(websocket: WebSocket, project_name: str):
     - Agent status changes
     - Agent stdout/stderr lines
     """
+    # Always accept WebSocket first to avoid opaque 403 errors
+    await websocket.accept()
+
     if not validate_project_name(project_name):
+        await websocket.send_json({"type": "error", "content": "Invalid project name"})
         await websocket.close(code=4000, reason="Invalid project name")
         return
 
     project_dir = _get_project_path(project_name)
     if not project_dir:
+        await websocket.send_json({"type": "error", "content": "Project not found in registry"})
         await websocket.close(code=4004, reason="Project not found in registry")
         return
 
     if not project_dir.exists():
+        await websocket.send_json({"type": "error", "content": "Project directory not found"})
         await websocket.close(code=4004, reason="Project directory not found")
         return
 
@@ -879,8 +883,7 @@ async def project_websocket(websocket: WebSocket, project_name: str):
                 break
             except json.JSONDecodeError:
                 logger.warning(f"Invalid JSON from WebSocket: {data[:100] if data else 'empty'}")
-            except Exception as e:
-                logger.warning(f"WebSocket error: {e}")
+            except Exception:
                 break
 
     finally:

package/temp_cleanup.py

@@ -0,0 +1,148 @@
+"""
+Temp Cleanup Module
+===================
+
+Cleans up stale temporary files and directories created by AutoForge agents,
+Playwright, Node.js, and other development tools.
+
+Called at Maestro (orchestrator) startup to prevent temp folder bloat.
+
+Why this exists:
+- Playwright creates browser profiles and artifacts in %TEMP%
+- Node.js creates .node cache files (~7MB each, can accumulate to GBs)
+- MongoDB Memory Server downloads binaries to temp
+- These are never cleaned up automatically
+
+When cleanup runs:
+- At Maestro startup (when you click Play or auto-restart after rate limits)
+- Only files/folders older than 1 hour are deleted (safe for running processes)
+"""
+
+import logging
+import shutil
+import tempfile
+import time
+from pathlib import Path
+
+logger = logging.getLogger(__name__)
+
+# Max age in seconds before a temp item is considered stale (1 hour)
+MAX_AGE_SECONDS = 3600
+
+# Directory patterns to clean up (glob patterns)
+DIR_PATTERNS = [
+    "playwright_firefoxdev_profile-*",  # Playwright Firefox profiles
+    "playwright-artifacts-*",           # Playwright test artifacts
+    "playwright-transform-cache",       # Playwright transform cache
+    "mongodb-memory-server*",           # MongoDB Memory Server binaries
+    "ng-*",                             # Angular CLI temp directories
+    "scoped_dir*",                      # Chrome/Chromium temp directories
+]
+
+# File patterns to clean up (glob patterns)
+FILE_PATTERNS = [
+    ".78912*.node",   # Node.js native module cache (major space consumer, ~7MB each)
+    "claude-*-cwd",   # Claude CLI working directory temp files
+    "mat-debug-*.log",  # Material/Angular debug logs
+]
+
+
+def cleanup_stale_temp(max_age_seconds: int = MAX_AGE_SECONDS) -> dict:
+    """
+    Clean up stale temporary files and directories.
+
+    Only deletes items older than max_age_seconds to avoid
+    interfering with currently running processes.
+
+    Args:
+        max_age_seconds: Maximum age in seconds before an item is deleted.
+                        Defaults to 1 hour (3600 seconds).
+
+    Returns:
+        Dictionary with cleanup statistics:
+        - dirs_deleted: Number of directories deleted
+        - files_deleted: Number of files deleted
+        - bytes_freed: Approximate bytes freed
+        - errors: List of error messages (for debugging, not fatal)
+    """
+    temp_dir = Path(tempfile.gettempdir())
+    cutoff_time = time.time() - max_age_seconds
+
+    stats = {
+        "dirs_deleted": 0,
+        "files_deleted": 0,
+        "bytes_freed": 0,
+        "errors": [],
+    }
+
+    # Clean up directories
+    for pattern in DIR_PATTERNS:
+        for item in temp_dir.glob(pattern):
+            if not item.is_dir():
+                continue
+            try:
+                mtime = item.stat().st_mtime
+                if mtime < cutoff_time:
+                    size = _get_dir_size(item)
+                    shutil.rmtree(item, ignore_errors=True)
+                    if not item.exists():
+                        stats["dirs_deleted"] += 1
+                        stats["bytes_freed"] += size
+                        logger.debug(f"Deleted temp directory: {item}")
+            except Exception as e:
+                stats["errors"].append(f"Failed to delete {item}: {e}")
+                logger.debug(f"Failed to delete {item}: {e}")
+
+    # Clean up files
+    for pattern in FILE_PATTERNS:
+        for item in temp_dir.glob(pattern):
+            if not item.is_file():
+                continue
+            try:
+                mtime = item.stat().st_mtime
+                if mtime < cutoff_time:
+                    size = item.stat().st_size
+                    item.unlink(missing_ok=True)
+                    if not item.exists():
+                        stats["files_deleted"] += 1
+                        stats["bytes_freed"] += size
+                        logger.debug(f"Deleted temp file: {item}")
+            except Exception as e:
+                stats["errors"].append(f"Failed to delete {item}: {e}")
+                logger.debug(f"Failed to delete {item}: {e}")
+
+    # Log summary if anything was cleaned
+    if stats["dirs_deleted"] > 0 or stats["files_deleted"] > 0:
+        mb_freed = stats["bytes_freed"] / (1024 * 1024)
+        logger.info(
+            f"Temp cleanup: {stats['dirs_deleted']} dirs, "
+            f"{stats['files_deleted']} files, {mb_freed:.1f} MB freed"
+        )
+
+    return stats
+
+
+def _get_dir_size(path: Path) -> int:
+    """Get total size of a directory in bytes."""
+    total = 0
+    try:
+        for item in path.rglob("*"):
+            if item.is_file():
+                try:
+                    total += item.stat().st_size
+                except (OSError, PermissionError):
+                    pass
+    except (OSError, PermissionError):
+        pass
+    return total
+
+
+if __name__ == "__main__":
+    # Allow running directly for testing/manual cleanup
+    logging.basicConfig(level=logging.DEBUG)
+    print("Running temp cleanup...")
+    stats = cleanup_stale_temp()
+    mb_freed = stats["bytes_freed"] / (1024 * 1024)
+    print(f"Cleanup complete: {stats['dirs_deleted']} dirs, {stats['files_deleted']} files, {mb_freed:.1f} MB freed")
+    if stats["errors"]:
+        print(f"Errors (non-fatal): {len(stats['errors'])}")