[codex] simplify memory read metrics (#28164)

## Why

Memory read telemetry currently reconstructs the executable shell
command after a tool call finishes. That duplicates shell, login-policy,
and cwd resolution owned by the tool handlers, and can diverge from the
environment-specific command that unified exec actually ran.

## What changed

- Expose the existing restricted shell-script parser directly for raw
script text.
- Parse `shell_command` and `exec_command` input into plain command argv
before classifying memory reads.
- Preserve all-or-nothing safe-command validation for multi-command
scripts.
- Remove cwd resolution, shell selection, and the unnecessary async
boundary from memory read metric emission.

## Testing

- `just test -p codex-shell-command`
- `cargo check -p codex-core`
This commit is contained in:
pakrym-oai
2026-06-15 08:28:02 -07:00
committed by GitHub
Unverified
parent bbcfed8ac2
commit 42ad752f36
6 changed files with 36 additions and 58 deletions
+6 -40
View File
@@ -5,20 +5,15 @@ use crate::tools::handlers::unified_exec::ExecCommandArgs;
use codex_memories_read::usage::MEMORIES_USAGE_METRIC;
use codex_memories_read::usage::memories_usage_kinds_from_command;
use codex_protocol::models::ShellCommandToolCallParams;
use std::path::PathBuf;
pub(crate) async fn emit_metric_for_tool_read(invocation: &ToolInvocation, success: bool) {
let Some((command, _)) = shell_command_for_invocation(invocation) else {
pub(crate) fn emit_metric_for_tool_read(invocation: &ToolInvocation, success: bool) {
let Some(command) = shell_script_for_invocation(invocation) else {
return;
};
let kinds = memories_usage_kinds_from_command(&command);
if kinds.is_empty() {
return;
}
let success = if success { "true" } else { "false" };
let tool_name = flat_tool_name(&invocation.tool_name);
for kind in kinds {
for kind in memories_usage_kinds_from_command(&command) {
invocation.turn.session_telemetry.counter(
MEMORIES_USAGE_METRIC,
/*inc*/ 1,
@@ -31,7 +26,7 @@ pub(crate) async fn emit_metric_for_tool_read(invocation: &ToolInvocation, succe
}
}
fn shell_command_for_invocation(invocation: &ToolInvocation) -> Option<(Vec<String>, PathBuf)> {
fn shell_script_for_invocation(invocation: &ToolInvocation) -> Option<String> {
let ToolPayload::Function { arguments } = &invocation.payload else {
return None;
};
@@ -42,39 +37,10 @@ fn shell_command_for_invocation(invocation: &ToolInvocation) -> Option<(Vec<Stri
) {
(None, "shell_command") => serde_json::from_str::<ShellCommandToolCallParams>(arguments)
.ok()
.map(|params| {
if !invocation.turn.config.permissions.allow_login_shell
&& params.login == Some(true)
{
#[allow(deprecated)]
let cwd = invocation.turn.resolve_path(params.workdir).to_path_buf();
return (Vec::new(), cwd);
}
let use_login_shell = params
.login
.unwrap_or(invocation.turn.config.permissions.allow_login_shell);
let command = invocation
.session
.user_shell()
.derive_exec_args(&params.command, use_login_shell);
#[allow(deprecated)]
let cwd = invocation.turn.resolve_path(params.workdir).to_path_buf();
(command, cwd)
}),
.map(|params| params.command),
(None, "exec_command") => serde_json::from_str::<ExecCommandArgs>(arguments)
.ok()
.and_then(|params| {
let command = crate::tools::handlers::unified_exec::get_command(
&params,
invocation.session.user_shell(),
&invocation.turn.unified_exec_shell_mode,
invocation.turn.config.permissions.allow_login_shell,
)
.ok()?;
#[allow(deprecated)]
let cwd = invocation.turn.resolve_path(params.workdir).to_path_buf();
Some((command.command, cwd))
}),
.map(|params| params.cmd),
(Some(_), _) | (None, _) => None,
}
}
@@ -26,9 +26,7 @@ pub use write_stdin::WriteStdinHandler;
#[derive(Debug, Deserialize)]
pub(crate) struct ExecCommandArgs {
cmd: String,
#[serde(default)]
pub(crate) workdir: Option<String>,
pub(crate) cmd: String,
#[serde(default)]
shell: Option<String>,
#[serde(default)]
+1 -1
View File
@@ -569,7 +569,7 @@ impl ToolRegistry {
Ok((_, success)) => *success,
Err(_) => false,
};
emit_metric_for_tool_read(&invocation, success).await;
emit_metric_for_tool_read(&invocation, success);
let post_tool_use_payload = if success {
let guard = response_cell.lock().await;
guard
+11 -4
View File
@@ -1,6 +1,7 @@
use codex_protocol::parse_command::ParsedCommand;
use codex_shell_command::bash::parse_shell_script_into_commands;
use codex_shell_command::is_safe_command::is_known_safe_command;
use codex_shell_command::parse_command::parse_command;
use codex_shell_command::parse_command::parse_shell_script;
pub use crate::metrics::MEMORIES_USAGE_METRIC;
@@ -25,12 +26,18 @@ impl MemoriesUsageKind {
}
}
pub fn memories_usage_kinds_from_command(command: &[String]) -> Vec<MemoriesUsageKind> {
if !is_known_safe_command(command) {
pub fn memories_usage_kinds_from_command(command: &str) -> Vec<MemoriesUsageKind> {
let Some(commands) = parse_shell_script_into_commands(command) else {
return Vec::new();
};
if !commands
.iter()
.all(|command| is_known_safe_command(command))
{
return Vec::new();
}
parse_command(command)
parse_shell_script(command)
.into_iter()
.filter_map(|command| match command {
ParsedCommand::Read { path, .. } => get_memory_kind(path.display().to_string()),
+8 -5
View File
@@ -94,6 +94,12 @@ pub fn try_parse_word_only_commands_sequence(tree: &Tree, src: &str) -> Option<V
Some(commands)
}
/// Parses a shell script consisting only of plain commands joined by safe operators.
pub fn parse_shell_script_into_commands(script: &str) -> Option<Vec<Vec<String>>> {
let tree = try_parse_shell(script)?;
try_parse_word_only_commands_sequence(&tree, script)
}
pub fn extract_bash_command(command: &[String]) -> Option<(&str, &str)> {
let [shell, flag, script] = command else {
return None;
@@ -114,9 +120,7 @@ pub fn extract_bash_command(command: &[String]) -> Option<(&str, &str)> {
/// joined by safe operators.
pub fn parse_shell_lc_plain_commands(command: &[String]) -> Option<Vec<Vec<String>>> {
let (_, script) = extract_bash_command(command)?;
let tree = try_parse_shell(script)?;
try_parse_word_only_commands_sequence(&tree, script)
parse_shell_script_into_commands(script)
}
/// Returns the parsed argv for a single shell command in a here-doc style
@@ -322,8 +326,7 @@ mod tests {
use pretty_assertions::assert_eq;
fn parse_seq(src: &str) -> Option<Vec<Vec<String>>> {
let tree = try_parse_shell(src)?;
try_parse_word_only_commands_sequence(&tree, src)
parse_shell_script_into_commands(src)
}
#[test]
+9 -5
View File
@@ -1818,7 +1818,11 @@ fn parse_find_query_and_path(tail: &[String]) -> (Option<String>, Option<String>
fn parse_shell_lc_commands(original: &[String]) -> Option<Vec<ParsedCommand>> {
// Only handle bash/zsh here; PowerShell is stripped separately without bash parsing.
let (_, script) = extract_bash_command(original)?;
Some(parse_shell_script(script))
}
/// Parses command metadata from a Bash-compatible shell script.
pub fn parse_shell_script(script: &str) -> Vec<ParsedCommand> {
if let Some(tree) = try_parse_shell(script)
&& let Some(all_commands) = try_parse_word_only_commands_sequence(&tree, script)
&& !all_commands.is_empty()
@@ -1831,9 +1835,9 @@ fn parse_shell_lc_commands(original: &[String]) -> Option<Vec<ParsedCommand>> {
// Commands arrive in source order; drop formatting helpers while preserving it.
let filtered_commands = drop_small_formatting_commands(all_commands);
if filtered_commands.is_empty() {
return Some(vec![ParsedCommand::Unknown {
return vec![ParsedCommand::Unknown {
cmd: script.to_string(),
}]);
}];
}
// Build parsed commands, tracking `cd` segments to compute effective file paths.
let mut commands: Vec<ParsedCommand> = Vec::new();
@@ -1939,11 +1943,11 @@ fn parse_shell_lc_commands(original: &[String]) -> Option<Vec<ParsedCommand>> {
})
.collect();
}
return Some(commands);
return commands;
}
Some(vec![ParsedCommand::Unknown {
vec![ParsedCommand::Unknown {
cmd: script.to_string(),
}])
}]
}
/// Return true if this looks like a small formatting helper in a pipeline.