btw Addin cannot access R Environment / session objects (unlike `btw_app()`)

Summary

There is a functional discrepancy between launching the btw chat interface via the RStudio Addin vs. calling btw_app() directly: the Addin does not have access to the current R session environment, making env and session tools non-functional when used through the Addin.

Expected behavior

The btw RStudio Addin should behave equivalently to btw_app() with respect to tool availability, including access to the current R environment and session objects (e.g., loaded data frames, model objects, defined variables).

Actual behavior

When launching via the RStudio Addin:

• The tool selection UI correctly reflects the btw.md configuration (e.g., tools: [docs, env, session])
• However, the env and session tools appear to have no access to the live R GlobalEnvironment — objects present in the session are not visible to the LLM

When launching via btw_app() in the console:

• Full access to the current R environment works as expected
• However, btw_app() blocks the R session, making parallel work in RStudio impossible

This creates a forced trade-off with no good solution:

• Addin → non-blocking, but no environment access
• btw_app() → environment access, but session is blocked

Environment

R version: 4.5.2
Platform: macOS (Apple Silicon)
RStudio version: 2026.4.0.347
btw version: 1.1.0
ellmer version: 0.4.0

questions

• Could the Addin be modified to run in the same R process (or share environment state) rather than as a separate process?
• Alternatively, could btw_app() gain a background = TRUE argument that launches non-blocking (e.g., via rstudioapi::jobRunScript()) while retaining environment access?
• Or could environment snapshots be serialized and passed to the Addin process at launch time?

Compare the RStudio Addin launch path with the btw_app() entry point, focusing on how the env and session tools receive R session state. Verify behavior using the issue's example configuration and live GlobalEnvironment objects; done means the Addin retains environment and session access without blocking parallel RStudio work.