llm-rovodev 0.1.1

LLM plugin exposing the RovoDev model for the llm CLI

llm-rovodev

Plugin for LLM adding a model called rovodev that shells out to acli rovodev run and returns the parsed response.

Installation

Install this plugin in the same environment as LLM:

llm install llm-rovodev

To learn more about installing and using the Rovo Dev CLI itself, see: Introducing Rovo Dev CLI – AI-Powered Development in your terminal.

Usage

This plugin adds a model called rovodev. You can execute it like this:

llm -m rovodev "What is 2+2?"

By default it parses the box-drawn "Response" section from the external acli rovodev run output and prints just that content.

Return raw stdout (including banners and box drawing) using -o raw:

llm -m rovodev -o raw true "What is 2+2?"

Notes:

• The plugin executes an external CLI: acli rovodev run. Override the binary path via ACLI_BIN if needed (e.g. ACLI_BIN=/path/to/acli).
• Advanced features are intentionally not supported: no API keys, no schemas, no tools, no attachments, no token usage tracking, etc.

Handling blank/no response from Rovo Dev CLI

This plugin relies on the external acli rovodev run to emit a box-drawn "Response" section on stdout. If no such Response block is found, the plugin will exit with an error:

Error: No Response block found in rovodev CLI output. Enable LLM_ROVODEV_DEBUG=1 to capture full logs or run the rovodev CLI directly to debug.

Notes and tips:

• This can occur if the external CLI decides not to produce a response for certain inputs (for example multi-line or file-derived prompts).
• Enable debug logging to capture full stdout/stderr from the subprocess and the exact command invoked: LLM_ROVODEV_DEBUG=1.
• You can also run the underlying CLI directly to compare behavior: acli rovodev run "your prompt".
• The -o raw option only changes how output is displayed when a Response block exists; it does not bypass the error on missing Response blocks. With LLM, booleans for -o are passed using two arguments, for example: -o raw true.

Debug logging

Enable verbose debug logs by setting LLM_ROVODEV_DEBUG=1 for a run. These logs include the effective command, redacted environment, verbatim prompt, timing info, and full stdout/stderr from the external CLI. By default logs go to stderr if your terminal is interactive.

To capture both logs and the normal model output into a file using shell redirection, enable debug and redirect stdout:

LLM_ROVODEV_DEBUG=1 llm -m rovodev "your prompt" > out.txt

• When stdout is redirected (non-tty), the plugin automatically emits debug logs to stdout so > out.txt contains both logs and the model's output.
• If you prefer logs to remain separate, keep using a TTY and redirect stderr only:

LLM_ROVODEV_DEBUG=1 llm -m rovodev "your prompt" 2> debug.log > out.txt

Development

To set up this plugin locally:

python3 -m venv venv
source venv/bin/activate
pip install -e .
pip install pytest
pytest