Struct CliRunnerBase 

pub struct CliRunnerBase { /* private fields */ }

Shared base struct for all CLI runners.

Holds the common fields (config, model info) that every CLI runner needs. Individual runners wrap this and add only their command-building and response-parsing logic.

A runner holds no conversation state between calls. Every request carries its whole history, and one runner instance serves every caller in the process — so a CLI session id kept here and passed back as --resume would splice one caller’s conversation into the next caller’s turn.

Implementations

impl CliRunnerBase

pub fn new( config: RunnerConfig, default_model: &str, fallback_models: &[&str], ) -> Self

Create a new base with the given config, default model, and fallback model list.

pub fn default_model(&self) -> &str

Get the default model identifier

pub fn available_models(&self) -> &[String]

Get the list of available models

pub async fn health_check(&self, runner_name: &str) -> Result<bool, RunnerError>

Run a --version health check against the runner binary.

Returns true if the binary exits 0, false otherwise.

Errors

Returns RunnerError if the subprocess cannot be spawned.

pub fn check_exit_code( &self, output: &CliOutput, runner_name: &str, ) -> Result<(), RunnerError>

Check CLI exit code and return an error if non-zero.

Logs output byte lengths (not content) and constructs an error message that surfaces the most diagnostic stream available. Stderr is preferred (where well-behaved Unix programs report errors), but some CLIs — notably claude-code exit-1 paths — write the error to stdout instead, so a non-empty stdout is used as a fallback rather than rendering the misleading (no output) the earlier implementation produced. Up to 500 chars are surfaced so multi-line diagnostics aren’t truncated to one line.

Errors

Returns RunnerError::external_service when output.exit_code != 0.