Skip to content

repl.md

Latest commit

 

History

History
83 lines (65 loc) · 3.52 KB

repl.md

File metadata and controls

83 lines (65 loc) · 3.52 KB

The REPL

opp_repl is built around an IPython shell that has all simulation functions pre-loaded. This page describes REPL-specific features that go beyond the standard IPython experience.

Launching

opp_repl --load ~/workspace/omnetpp/omnetpp.opp \
         --load ~/workspace/inet/inet.opp

Key command-line options:

  • --load OPP_FILE — load .opp descriptor file(s); accepts single files, directories (loads all *.opp inside), glob patterns, or the special token @opp for the bundled files. Can be repeated.
  • -p PROJECT — set the default simulation project by name.
  • --mcp-port PORT — start an MCP server on the given TCP port for AI assistant integration. Default is 0, which disables the MCP server entirely. Setting a non-zero port also requires --mcp-token-hash (unless running inside opp_sandbox).
  • --mcp-token-hash HEX — SHA-256 hex hash of the bearer token that MCP clients must present to authenticate. Required when --mcp-port is set, except inside opp_sandbox.
  • --mcp-bypass-token-hash-check — disable bearer-token authentication for the MCP server. Use only in trusted environments; normally --mcp-token-hash is required outside opp_sandbox.

About opp_sandbox: a bubblewrap-based sandbox wrapper shipped with opp_repl (bin/opp_sandbox). It runs opp_repl inside a restricted namespace where the only writable paths are the working directory and explicitly mounted ones. Several features (MCP authentication, overlay-build mounts) treat the sandbox as a trusted environment and relax their requirements when they detect it.

  • -l LEVEL — log level (ERROR, WARN, INFO, DEBUG). Note that opp_repl defaults to INFO, while the opp_* command-line tools default to WARN.
  • --external-command-log-level LEVEL — log level for output from simulations and build tools.
  • --no-handle-exception — show full Python tracebacks instead of short error messages.

Pre-loaded namespace

When the REPL starts, every public function from opp_repl is imported into the top-level namespace — run_simulations, build_project, run_fingerprint_tests, etc. — so you can call them without any prefix.

Convenience variables

Every loaded simulation project is injected as a variable named {name}_project (hyphens and dots become underscores). For example, loading inet.opp creates inet_project, and loading simu5g.opp creates simu5g_project. Use get_simulation_project_variable_names() to list them all.

Autoreload

IPython's %autoreload 2 is enabled at startup. If you edit Python source files (e.g. a user module or opp_repl itself), changes are picked up automatically before each command — no need to restart the REPL.

User module

At startup the REPL tries to import a Python module named after the current OS login user (e.g. import levy). All public names from that module are injected into the namespace. This lets you define personal helper functions, set project defaults, or customize the environment without modifying opp_repl itself. If no such module exists, the import is silently skipped.

Stopping execution

Call stop_execution() (or stop_execution(value)) anywhere in a REPL script to abort the current cell cleanly, without a full traceback. This is useful in multi-line REPL scripts where you want to bail out early.

MCP server

When started with --mcp-port, the REPL exposes its functions to AI assistants via the Model Context Protocol. See MCP server for details.