Hamilton VSCode extension 0.1.2 fails to detect installed LSP in Positron; dataflow graph never appears

## Expected behavior
The Hamilton extension should detect the installed language server and render the Hamilton dataflow graph.

## Actual
A notification popup repeatedly reports that the language server needs to be installed.

The Hamilton dataflow graph never appears.

The extension activates, but the graph view remains unavailable.

## Library & System Information
OS: Windows 11
IDE: Positron
Extension: DAGWorks.hamilton-vsc 0.1.2
Python: CPython 3.14.3
Environment manager: uv
*Project path contains spaces:*
C:\Users\\OneDrive - Camden\workspace\PyDev v01\...

LSP:
sf-hamilton-lsp 0.2.0
apache-hamilton-lsp (installed as dependency)

Installed the extension via VSIX/Marketplace.

## Additional context
Confirmed Positron is using the intended .venv interpreter.

uv run python -c "import hamilton_lsp; print(hamilton_lsp.__file__)"

returns:

C:\...\site-packages\hamilton_lsp\__init__.py

also

uv pip show sf-hamilton-lsp

returns:

Name: sf-hamilton-lsp
Version: 0.2.0
Location: \.venv\Lib\site-packages

and

Get-ChildItem .\.venv\Scripts\*hamilton*

shows:

hamilton-lsp.exe

Found in the log:

TypeError: Cannot read properties of undefined (reading 'forEach')

at Object.PR (...\dagworks.hamilton-vsc-0.1.2\out\extension.js)

The extension appears to activate, but the graph never renders.

ExtensionService#_doActivateExtension DAGWorks.hamilton-vsc,
activationEvent: 'onView:hamilton.dataflowWebview'

The issue occurs despite:

- hamilton_lsp being importable
- hamilton-lsp.exe existing in .venv\Scripts
- Positron using the same interpreter where the LSP is installed

Reproduce the failure in Positron on Windows with the project's .venv and sf-hamilton-lsp 0.2.0 installed, then inspect the Hamilton VSCode extension activation logs around the reported TypeError in out/extension.js. Confirm the extension detects hamilton-lsp.exe and renders the dataflow graph without repeated installation notifications.

Written by the indexing model from the issue text.