Remove Streamlit dependency and migrate to Flask dashboard

- Remove streamlit from pyproject.toml dependencies
- Update CLI dash command to launch Flask instead of Streamlit
- Add environment variable support for Flask host/port configuration
- Remove mock_streamlit fixture from tests
- Update README to reflect Flask-based dashboard
- Fix mypy error: use os.environ instead of subprocess.os.environ

Author: bangyen

README.md

@@ -49,7 +49,7 @@ emergent/
 │   ├── training/        # Training loops and optimization
 │   ├── experiments/     # Population dynamics and ablation studies
 │   ├── analysis/        # Language analysis and evaluation
-│   ├── apps/            # CLI and Streamlit dashboard
+│   ├── apps/            # CLI interface
 │   └── tracking/        # MLflow experiment tracking
 ├── tests/               # Unit and integration tests
 ├── docs/                # Documentation and figures

dashboard/main.py

@@ -151,4 +151,9 @@ def get_stats() -> Response:
 
 
 if __name__ == "__main__":
-    app.run(debug=True, host="0.0.0.0", port=5000)
+    import os
+
+    host = os.environ.get("FLASK_RUN_HOST", "0.0.0.0")
+    port = int(os.environ.get("FLASK_RUN_PORT", "5000"))
+
+    app.run(debug=True, host=host, port=port)

pyproject.toml

@@ -36,7 +36,6 @@ dependencies = [
     "numpy>=1.21.0",
     "matplotlib>=3.5.0",
     "seaborn>=0.11.0",
-    "streamlit>=1.28.0",
     "pandas>=1.5.0",
     "scipy>=1.7.0",
     "mlflow>=2.0.0",

src/langlab/apps/cli.py

@@ -570,47 +570,35 @@ def parse_heldout(heldout_str: str) -> list:
 
 
 @main.command()
-@click.option("--port", default=8888, help="Port to run the dashboard on")
-@click.option("--host", default="localhost", help="Host to run the dashboard on")
+@click.option("--port", default=5000, help="Port to run the dashboard on")
+@click.option("--host", default="0.0.0.0", help="Host to run the dashboard on")
 def dash(port: int, host: str) -> None:
-    """Launch the interactive Streamlit dashboard for visualizing language emergence."""
+    """Launch the Flask dashboard for visualizing language emergence."""
+    import os
     import subprocess
     import sys
-    import os
+    from pathlib import Path
 
-    # Get the path to the app.py file
-    app_path = os.path.join(os.path.dirname(__file__), "app.py")
+    dashboard_dir = Path(__file__).parent.parent.parent.parent / "dashboard"
+    app_path = dashboard_dir / "main.py"
 
-    if not os.path.exists(app_path):
-        click.echo("Error: Dashboard app not found", err=True)
+    if not app_path.exists():
+        click.echo("Error: Dashboard not found at dashboard/main.py", err=True)
         return
 
     logger.info(f"Launching dashboard on {host}:{port}")
     click.echo("Launching Language Emergence Dashboard...")
-    click.echo(f"Dashboard will be available at: http://{host}:{port}")
+    click.echo(f"Dashboard will be available at: http://localhost:{port}")
     click.echo("Press Ctrl+C to stop the dashboard")
 
     try:
-        # Launch Streamlit
-        cmd = [
-            sys.executable,
-            "-m",
-            "streamlit",
-            "run",
-            app_path,
-            "--server.headless",
-            "true",
-            "--server.port",
-            str(port),
-            "--server.address",
-            host,
-        ]
-        subprocess.run(cmd, check=True)
+        cmd = [sys.executable, str(app_path)]
+        env = {**os.environ, "FLASK_RUN_PORT": str(port), "FLASK_RUN_HOST": host}
+        subprocess.run(cmd, check=True, env=env)
     except subprocess.CalledProcessError as e:
         logger.error(f"Failed to launch dashboard: {e}")
         click.echo(
-            "Error: Failed to launch dashboard. Make sure Streamlit is installed.",
-            err=True,
+            "Error: Failed to launch dashboard. Make sure Flask is installed.", err=True
         )
     except KeyboardInterrupt:
         click.echo("\nDashboard stopped.")

tests/conftest.py

@@ -226,27 +226,6 @@ def sample_ablation_params() -> Dict[str, Any]:
     }
 
 
-@pytest.fixture
-def mock_streamlit() -> Generator[Any, None, None]:
-    """Mock Streamlit for testing dashboard components."""
-    with patch("streamlit.set_page_config"), patch("streamlit.title"), patch(
-        "streamlit.markdown"
-    ), patch("streamlit.sidebar"), patch("streamlit.selectbox"), patch(
-        "streamlit.warning"
-    ), patch(
-        "streamlit.success"
-    ), patch(
-        "streamlit.error"
-    ), patch(
-        "streamlit.metric"
-    ), patch(
-        "streamlit.plotly_chart"
-    ), patch(
-        "streamlit.pyplot"
-    ) as mock_plt:
-        yield mock_plt
-
-
 @pytest.fixture
 def mock_torch_load() -> Generator[Any, None, None]:
     """Mock torch.load for testing checkpoint loading."""