Files
codex/codex-rs/exec/src/lib.rs
Michael Bolin e0c08cea4f feat: add support for --sandbox flag (#1476)
On a high-level, we try to design `config.toml` so that you don't have
to "comment out a lot of stuff" when testing different options.

Previously, defining a sandbox policy was somewhat at odds with this
principle because you would define the policy as attributes of
`[sandbox]` like so:

```toml
[sandbox]
mode = "workspace-write"
writable_roots = [ "/tmp" ]
```

but if you wanted to temporarily change to a read-only sandbox, you
might feel compelled to modify your file to be:

```toml
[sandbox]
mode = "read-only"
# mode = "workspace-write"
# writable_roots = [ "/tmp" ]
```

Technically, commenting out `writable_roots` would not be strictly
necessary, as `mode = "read-only"` would ignore `writable_roots`, but
it's still a reasonable thing to do to keep things tidy.

Currently, the various values for `mode` do not support that many
attributes, so this is not that hard to maintain, but one could imagine
this becoming more complex in the future.

In this PR, we change Codex CLI so that it no longer recognizes
`[sandbox]`. Instead, it introduces a top-level option, `sandbox_mode`,
and `[sandbox_workspace_write]` is used to further configure the sandbox
when when `sandbox_mode = "workspace-write"` is used:

```toml
sandbox_mode = "workspace-write"

[sandbox_workspace_write]
writable_roots = [ "/tmp" ]
```

This feels a bit more future-proof in that it is less tedious to
configure different sandboxes:

```toml
sandbox_mode = "workspace-write"

[sandbox_read_only]
# read-only options here...

[sandbox_workspace_write]
writable_roots = [ "/tmp" ]

[sandbox_danger_full_access]
# danger-full-access options here...
```

In this scheme, you never need to comment out the configuration for an
individual sandbox type: you only need to redefine `sandbox_mode`.

Relatedly, previous to this change, a user had to do `-c
sandbox.mode=read-only` to change the mode on the command line. With
this change, things are arguably a bit cleaner because the equivalent
option is `-c sandbox_mode=read-only` (and now `-c
sandbox_workspace_write=...` can be set separately).

Though more importantly, we introduce the `-s/--sandbox` option to the
CLI, which maps directly to `sandbox_mode` in `config.toml`, making
config override behavior easier to reason about. Moreover, as you can
see in the updates to the various Markdown files, it is much easier to
explain how to configure sandboxing when things like `--sandbox
read-only` can be used as an example.

Relatedly, this cleanup also made it straightforward to add support for
a `sandbox` option for Codex when used as an MCP server (see the changes
to `mcp-server/src/codex_tool_config.rs`).

Fixes https://github.com/openai/codex/issues/1248.
2025-07-07 22:31:30 -07:00

250 lines
8.7 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
mod cli;
mod event_processor;
use std::io::IsTerminal;
use std::io::Read;
use std::path::Path;
use std::path::PathBuf;
use std::sync::Arc;
pub use cli::Cli;
use codex_core::codex_wrapper;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use codex_core::config_types::SandboxMode;
use codex_core::protocol::AskForApproval;
use codex_core::protocol::Event;
use codex_core::protocol::EventMsg;
use codex_core::protocol::InputItem;
use codex_core::protocol::Op;
use codex_core::protocol::TaskCompleteEvent;
use codex_core::util::is_inside_git_repo;
use event_processor::EventProcessor;
use tracing::debug;
use tracing::error;
use tracing::info;
use tracing_subscriber::EnvFilter;
pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
let Cli {
images,
model,
config_profile,
full_auto,
dangerously_bypass_approvals_and_sandbox,
cwd,
skip_git_repo_check,
color,
last_message_file,
sandbox_mode: sandbox_mode_cli_arg,
prompt,
config_overrides,
} = cli;
// Determine the prompt based on CLI arg and/or stdin.
let prompt = match prompt {
Some(p) if p != "-" => p,
// Either `-` was passed or no positional arg.
maybe_dash => {
// When no arg (None) **and** stdin is a TTY, bail out early unless the
// user explicitly forced reading via `-`.
let force_stdin = matches!(maybe_dash.as_deref(), Some("-"));
if std::io::stdin().is_terminal() && !force_stdin {
eprintln!(
"No prompt provided. Either specify one as an argument or pipe the prompt into stdin."
);
std::process::exit(1);
}
// Ensure the user knows we are waiting on stdin, as they may
// have gotten into this state by mistake. If so, and they are not
// writing to stdin, Codex will hang indefinitely, so this should
// help them debug in that case.
if !force_stdin {
eprintln!("Reading prompt from stdin...");
}
let mut buffer = String::new();
if let Err(e) = std::io::stdin().read_to_string(&mut buffer) {
eprintln!("Failed to read prompt from stdin: {e}");
std::process::exit(1);
} else if buffer.trim().is_empty() {
eprintln!("No prompt provided via stdin.");
std::process::exit(1);
}
buffer
}
};
let (stdout_with_ansi, stderr_with_ansi) = match color {
cli::Color::Always => (true, true),
cli::Color::Never => (false, false),
cli::Color::Auto => (
std::io::stdout().is_terminal(),
std::io::stderr().is_terminal(),
),
};
let sandbox_mode = if full_auto {
Some(SandboxMode::WorkspaceWrite)
} else if dangerously_bypass_approvals_and_sandbox {
Some(SandboxMode::DangerFullAccess)
} else {
sandbox_mode_cli_arg.map(Into::<SandboxMode>::into)
};
// Load configuration and determine approval policy
let overrides = ConfigOverrides {
model,
config_profile,
// This CLI is intended to be headless and has no affordances for asking
// the user for approval.
approval_policy: Some(AskForApproval::Never),
sandbox_mode,
cwd: cwd.map(|p| p.canonicalize().unwrap_or(p)),
model_provider: None,
codex_linux_sandbox_exe,
};
// Parse `-c` overrides.
let cli_kv_overrides = match config_overrides.parse_overrides() {
Ok(v) => v,
Err(e) => {
eprintln!("Error parsing -c overrides: {e}");
std::process::exit(1);
}
};
let config = Config::load_with_cli_overrides(cli_kv_overrides, overrides)?;
let mut event_processor =
EventProcessor::create_with_ansi(stdout_with_ansi, !config.hide_agent_reasoning);
// Print the effective configuration and prompt so users can see what Codex
// is using.
event_processor.print_config_summary(&config, &prompt);
if !skip_git_repo_check && !is_inside_git_repo(&config) {
eprintln!("Not inside a Git repo and --skip-git-repo-check was not specified.");
std::process::exit(1);
}
// TODO(mbolin): Take a more thoughtful approach to logging.
let default_level = "error";
let _ = tracing_subscriber::fmt()
// Fallback to the `default_level` log filter if the environment
// variable is not set _or_ contains an invalid value
.with_env_filter(
EnvFilter::try_from_default_env()
.or_else(|_| EnvFilter::try_new(default_level))
.unwrap_or_else(|_| EnvFilter::new(default_level)),
)
.with_ansi(stderr_with_ansi)
.with_writer(std::io::stderr)
.try_init();
let (codex_wrapper, event, ctrl_c) = codex_wrapper::init_codex(config).await?;
let codex = Arc::new(codex_wrapper);
info!("Codex initialized with event: {event:?}");
let (tx, mut rx) = tokio::sync::mpsc::unbounded_channel::<Event>();
{
let codex = codex.clone();
tokio::spawn(async move {
loop {
let interrupted = ctrl_c.notified();
tokio::select! {
_ = interrupted => {
// Forward an interrupt to the codex so it can abort any inflight task.
let _ = codex
.submit(
Op::Interrupt,
)
.await;
// Exit the inner loop and return to the main input prompt. The codex
// will emit a `TurnInterrupted` (Error) event which is drained later.
break;
}
res = codex.next_event() => match res {
Ok(event) => {
debug!("Received event: {event:?}");
if let Err(e) = tx.send(event) {
error!("Error sending event: {e:?}");
break;
}
},
Err(e) => {
error!("Error receiving event: {e:?}");
break;
}
}
}
}
});
}
// Send images first, if any.
if !images.is_empty() {
let items: Vec<InputItem> = images
.into_iter()
.map(|path| InputItem::LocalImage { path })
.collect();
let initial_images_event_id = codex.submit(Op::UserInput { items }).await?;
info!("Sent images with event ID: {initial_images_event_id}");
while let Ok(event) = codex.next_event().await {
if event.id == initial_images_event_id
&& matches!(
event.msg,
EventMsg::TaskComplete(TaskCompleteEvent {
last_agent_message: _,
})
)
{
break;
}
}
}
// Send the prompt.
let items: Vec<InputItem> = vec![InputItem::Text { text: prompt }];
let initial_prompt_task_id = codex.submit(Op::UserInput { items }).await?;
info!("Sent prompt with event ID: {initial_prompt_task_id}");
// Run the loop until the task is complete.
while let Some(event) = rx.recv().await {
let (is_last_event, last_assistant_message) = match &event.msg {
EventMsg::TaskComplete(TaskCompleteEvent { last_agent_message }) => {
(true, last_agent_message.clone())
}
_ => (false, None),
};
event_processor.process_event(event);
if is_last_event {
handle_last_message(last_assistant_message, last_message_file.as_deref())?;
break;
}
}
Ok(())
}
fn handle_last_message(
last_agent_message: Option<String>,
last_message_file: Option<&Path>,
) -> std::io::Result<()> {
match (last_agent_message, last_message_file) {
(Some(last_agent_message), Some(last_message_file)) => {
// Last message and a file to write to.
std::fs::write(last_message_file, last_agent_message)?;
}
(None, Some(last_message_file)) => {
eprintln!(
"Warning: No last message to write to file: {}",
last_message_file.to_string_lossy()
);
}
(_, None) => {
// No last message and no file to write to.
}
}
Ok(())
}