Support optional shell commands

.github/workflows/deploy.yml

@@ -35,6 +35,9 @@ jobs:
           cp README.md contents
         working-directory: ./demo
 
+      - name: Set up micromamba (cockle needs it to build cockle_wasm_env for the terminal)
+        uses: mamba-org/setup-micromamba@v3
+
       - name: Build the JupyterLite site
         run: |
           uv run jupyter lite build

.gitignore

@@ -127,6 +127,11 @@ dmypy.json
 _output
 *.doit.db
 
+# jupyterlite-terminal (cockle) build artifacts
+cockle_wasm_env/
+cockle-config.json
+.cockle_temp/
+
 # Jupyter
 Untitled*.ipynb
 

demo/jupyter-lite.json

@@ -0,0 +1,6 @@
+{
+  "jupyter-lite-schema-version": 0,
+  "jupyter-config-data": {
+    "terminalsAvailable": true
+  }
+}

demo/pyproject.toml

@@ -12,6 +12,7 @@ dependencies = [
     "jupyterlite-ai",
     "jupyterlite-core>=0.7.0",
     "jupyterlite-pyodide-kernel>=0.7.0",
+    "jupyterlite-terminal>=1.4.2a2",
 ]
 
 [tool.uv.sources]

src/models/settings-model.ts

@@ -58,6 +58,7 @@ You're designed to be a capable partner for data science, research, and developm
 - Create, read, edit, and organize files and notebooks in any language
 - Manage project structure and navigate file systems
 - Help with version control and project organization
+- Run shell commands in the in-browser file system when a terminal or shell extension is available
 
 **📊 Notebook Operations:**
 - Create new notebooks and manage existing ones
@@ -111,6 +112,13 @@ When asked to run code or perform computations, choose the most appropriate appr
 
 This means if the user asks you to "calculate the factorial of 100" or "check what library version is installed", run that directly with the jupyterlab-ai-commands kernel execution command rather than creating a new notebook file.
 
+## Shell & File System Operations
+Some environments also expose shell commands that run a real shell directly in the browser file system and capture its output (for example, when a terminal or shell extension is installed). These commands are optional and may not be present, so discover them rather than assuming specific command IDs.
+- For shell-style file system work (listing, searching, inspecting, or batch-manipulating files), discover what is available with discover_commands using query 'shell' or 'terminal', just as you use query 'jupyterlab-ai-commands' for file/notebook tasks.
+- If shell commands are returned, use the returned command IDs and follow each command's described arguments and limitations (for example, how to capture output and any restrictions on shell syntax). Never guess these command IDs.
+- If none are available, fall back to the jupyterlab-ai-commands file commands.
+- Keep using jupyterlab-ai-commands for notebook/cell operations and for opening files in the UI.
+
 ## Notebook State and Cell Identity
 When working with an existing notebook, use the notebook's current structure and kernel state as the source of truth.
 - Before changing notebook content or structure, inspect the notebook and any target cells with the relevant notebook commands you have discovered.