Fall back to configured instruction files if AGENTS.md isn't available (#4544)

Allow users to configure an agents.md alternative to consume, but warn
the user it may degrade model performance.

Fixes #4376
This commit is contained in:
easong-openai
2025-10-01 11:19:59 -07:00
committed by GitHub
parent 2f370e946d
commit 400a5a90bf
5 changed files with 117 additions and 6 deletions

View File

@@ -141,6 +141,9 @@ pub struct Config {
/// Maximum number of bytes to include from an AGENTS.md project doc file.
pub project_doc_max_bytes: usize,
/// Additional filenames to try when looking for project-level docs.
pub project_doc_fallback_filenames: Vec<String>,
/// Directory containing all Codex state (defaults to `~/.codex` but can be
/// overridden by the `CODEX_HOME` environment variable).
pub codex_home: PathBuf,
@@ -670,6 +673,9 @@ pub struct ConfigToml {
/// Maximum number of bytes to include from an AGENTS.md project doc file.
pub project_doc_max_bytes: Option<usize>,
/// Ordered list of fallback filenames to look for when AGENTS.md is missing.
pub project_doc_fallback_filenames: Option<Vec<String>>,
/// Profile to use from the `profiles` map.
pub profile: Option<String>,
@@ -1038,6 +1044,19 @@ impl Config {
mcp_servers: cfg.mcp_servers,
model_providers,
project_doc_max_bytes: cfg.project_doc_max_bytes.unwrap_or(PROJECT_DOC_MAX_BYTES),
project_doc_fallback_filenames: cfg
.project_doc_fallback_filenames
.unwrap_or_default()
.into_iter()
.filter_map(|name| {
let trimmed = name.trim();
if trimmed.is_empty() {
None
} else {
Some(trimmed.to_string())
}
})
.collect(),
codex_home,
history,
file_opener: cfg.file_opener.unwrap_or(UriBasedFileOpener::VsCode),
@@ -1811,6 +1830,7 @@ model_verbosity = "high"
mcp_servers: HashMap::new(),
model_providers: fixture.model_provider_map.clone(),
project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
project_doc_fallback_filenames: Vec::new(),
codex_home: fixture.codex_home(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
@@ -1871,6 +1891,7 @@ model_verbosity = "high"
mcp_servers: HashMap::new(),
model_providers: fixture.model_provider_map.clone(),
project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
project_doc_fallback_filenames: Vec::new(),
codex_home: fixture.codex_home(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
@@ -1946,6 +1967,7 @@ model_verbosity = "high"
mcp_servers: HashMap::new(),
model_providers: fixture.model_provider_map.clone(),
project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
project_doc_fallback_filenames: Vec::new(),
codex_home: fixture.codex_home(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
@@ -2007,6 +2029,7 @@ model_verbosity = "high"
mcp_servers: HashMap::new(),
model_providers: fixture.model_provider_map.clone(),
project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
project_doc_fallback_filenames: Vec::new(),
codex_home: fixture.codex_home(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,

View File

@@ -1,6 +1,7 @@
//! Project-level documentation discovery.
//!
//! Project-level documentation can be stored in files named `AGENTS.md`.
//! Project-level documentation is primarily stored in files named `AGENTS.md`.
//! Additional fallback filenames can be configured via `project_doc_fallback_filenames`.
//! We include the concatenation of all files found along the path from the
//! repository root to the current working directory as follows:
//!
@@ -17,8 +18,8 @@ use std::path::PathBuf;
use tokio::io::AsyncReadExt;
use tracing::error;
/// Currently, we only match the filename `AGENTS.md` exactly.
const CANDIDATE_FILENAMES: &[&str] = &["AGENTS.md"];
/// Default filename scanned for project-level docs.
pub const DEFAULT_PROJECT_DOC_FILENAME: &str = "AGENTS.md";
/// When both `Config::instructions` and the project doc are present, they will
/// be concatenated with the following separator.
@@ -152,8 +153,9 @@ pub fn discover_project_doc_paths(config: &Config) -> std::io::Result<Vec<PathBu
};
let mut found: Vec<PathBuf> = Vec::new();
let candidate_filenames = candidate_filenames(config);
for d in search_dirs {
for name in CANDIDATE_FILENAMES {
for name in &candidate_filenames {
let candidate = d.join(name);
match std::fs::symlink_metadata(&candidate) {
Ok(md) => {
@@ -173,6 +175,22 @@ pub fn discover_project_doc_paths(config: &Config) -> std::io::Result<Vec<PathBu
Ok(found)
}
fn candidate_filenames<'a>(config: &'a Config) -> Vec<&'a str> {
let mut names: Vec<&'a str> =
Vec::with_capacity(1 + config.project_doc_fallback_filenames.len());
names.push(DEFAULT_PROJECT_DOC_FILENAME);
for candidate in &config.project_doc_fallback_filenames {
let candidate = candidate.as_str();
if candidate.is_empty() {
continue;
}
if !names.contains(&candidate) {
names.push(candidate);
}
}
names
}
#[cfg(test)]
mod tests {
use super::*;
@@ -202,6 +220,20 @@ mod tests {
config
}
fn make_config_with_fallback(
root: &TempDir,
limit: usize,
instructions: Option<&str>,
fallbacks: &[&str],
) -> Config {
let mut config = make_config(root, limit, instructions);
config.project_doc_fallback_filenames = fallbacks
.iter()
.map(std::string::ToString::to_string)
.collect();
config
}
/// AGENTS.md missing should yield `None`.
#[tokio::test]
async fn no_doc_file_returns_none() {
@@ -347,4 +379,45 @@ mod tests {
let res = get_user_instructions(&cfg).await.expect("doc expected");
assert_eq!(res, "root doc\n\ncrate doc");
}
/// When AGENTS.md is absent but a configured fallback exists, the fallback is used.
#[tokio::test]
async fn uses_configured_fallback_when_agents_missing() {
let tmp = tempfile::tempdir().expect("tempdir");
fs::write(tmp.path().join("EXAMPLE.md"), "example instructions").unwrap();
let cfg = make_config_with_fallback(&tmp, 4096, None, &["EXAMPLE.md"]);
let res = get_user_instructions(&cfg)
.await
.expect("fallback doc expected");
assert_eq!(res, "example instructions");
}
/// AGENTS.md remains preferred when both AGENTS.md and fallbacks are present.
#[tokio::test]
async fn agents_md_preferred_over_fallbacks() {
let tmp = tempfile::tempdir().expect("tempdir");
fs::write(tmp.path().join("AGENTS.md"), "primary").unwrap();
fs::write(tmp.path().join("EXAMPLE.md"), "secondary").unwrap();
let cfg = make_config_with_fallback(&tmp, 4096, None, &["EXAMPLE.md", ".example.md"]);
let res = get_user_instructions(&cfg)
.await
.expect("AGENTS.md should win");
assert_eq!(res, "primary");
let discovery = discover_project_doc_paths(&cfg).expect("discover paths");
assert_eq!(discovery.len(), 1);
assert!(
discovery[0]
.file_name()
.unwrap()
.to_string_lossy()
.eq(DEFAULT_PROJECT_DOC_FILENAME)
);
}
}

View File

@@ -2,6 +2,7 @@ use super::*;
use crate::app_event::AppEvent;
use crate::app_event_sender::AppEventSender;
use crate::test_backend::VT100Backend;
use crate::tui::FrameRequester;
use codex_core::AuthManager;
use codex_core::CodexAuth;
use codex_core::config::Config;

View File

@@ -36,9 +36,13 @@ pub(crate) fn compose_agents_summary(config: &Config) -> String {
Ok(paths) => {
let mut rels: Vec<String> = Vec::new();
for p in paths {
let file_name = p
.file_name()
.map(|name| name.to_string_lossy().to_string())
.unwrap_or_else(|| "<unknown>".to_string());
let display = if let Some(parent) = p.parent() {
if parent == config.cwd {
"AGENTS.md".to_string()
file_name.clone()
} else {
let mut cur = config.cwd.as_path();
let mut ups = 0usize;
@@ -53,7 +57,7 @@ pub(crate) fn compose_agents_summary(config: &Config) -> String {
}
if reached {
let up = format!("..{}", std::path::MAIN_SEPARATOR);
format!("{}AGENTS.md", up.repeat(ups))
format!("{}{}", up.repeat(ups), file_name)
} else if let Ok(stripped) = p.strip_prefix(&config.cwd) {
stripped.display().to_string()
} else {