feat(hermes): auto-heal platforms.api_server.enabled before gateway start (Step 5)

Close the G7 gap from the v3 integration design.

Scenario: A user upgrades Hermes (or manually edits config.yaml and drops
the api_server platform block). The next `gateway start` silently produces
a Gateway that doesn't expose /v1/runs, and ClawPanel's chat/agent flows
fail with confusing errors.

Solution: Run a pre-start guardian that checks
`platforms.api_server.enabled: true` and auto-heals the file when
missing, with a timestamped backup so users can always roll back.

Backend (src-tauri/src/commands/hermes.rs):
- New pure helpers `config_has_api_server_enabled(raw)` and
  `patch_yaml_ensure_api_server(raw)` working directly on YAML strings
  (no serde_yaml dependency needed — we only touch 3 structural keys).
- `ensure_api_server_enabled(app)` wraps the helpers with filesystem
  I/O: reads config.yaml, writes `config.yaml.bak-<epoch>` on mutation,
  writes the patched content back, emits `hermes-config-patched` so
  the frontend can show a transparent toast.
- Called from `hermes_gateway_action` on every `start` action. If
  config.yaml doesn't exist, the guardian is a no-op (configure_hermes
  creates a compliant file on first run).
- 7 new unit tests covering: truthy value variants (true/yes/on/1),
  missing/disabled/commented scenarios, no-op when healthy,
  appending a full platforms block, injecting into existing platforms,
  replacing an explicitly disabled api_server subtree.

Web mode (scripts/dev-api.js):
- Mirrors the three helpers as `_hermesConfigHasApiServerEnabled`,
  `_hermesPatchYamlEnsureApiServer`, `_hermesEnsureApiServerEnabled`.
- Called in `hermes_gateway_action` start path.
- Logs to console.warn instead of emitting a Tauri event.

Frontend (src/engines/hermes/pages/dashboard.js):
- Dashboard subscribes to `hermes-config-patched` and surfaces a toast
  (6s duration) so the user knows the auto-heal happened and where the
  backup lives.

Verified: cargo fmt / cargo clippy -D warnings / cargo test all green
(12 hermes tests pass total: 5 provider registry + 7 guardian).
`node --check scripts/dev-api.js` passes. `npm run build` green.
This commit is contained in:
晴天
2026-04-24 20:58:30 +08:00
parent 31936b4779
commit b25808f7f0
3 changed files with 394 additions and 0 deletions

View File

@@ -6692,6 +6692,11 @@ const handlers = {
const enhanced = hermesEnhancedPath()
const port = hermesGatewayPort()
if (action === 'start') {
// Guardian: ensure platforms.api_server.enabled:true before start.
// Mirrors Rust's ensure_api_server_enabled (see hermes.rs).
try { this._hermesEnsureApiServerEnabled() } catch (e) {
console.warn('[hermes guardian] patch failed:', e.message || e)
}
// 检测是否已运行
const alive = await _tcpProbe('127.0.0.1', port, 300)
if (alive) return 'Gateway 已在运行'
@@ -6845,6 +6850,98 @@ const handlers = {
return []
},
// -----------------------------------------------------------------------
// api_server guardian (Step 5) — mirror of Rust's config_has_api_server_enabled
// + patch_yaml_ensure_api_server + ensure_api_server_enabled. Called before
// every `hermes gateway run` so that an upgrade / manual edit that drops
// `platforms.api_server.enabled: true` is auto-healed.
// -----------------------------------------------------------------------
_hermesConfigHasApiServerEnabled(raw) {
let inPlatforms = false
let inApiServer = false
for (const origLine of raw.split('\n')) {
const hash = origLine.indexOf('#')
const line = hash >= 0 ? origLine.slice(0, hash) : origLine
const trimmed = line.replace(/\s+$/, '')
if (!trimmed) continue
const indent = trimmed.length - trimmed.trimStart().length
if (indent === 0) {
inPlatforms = trimmed.trimStart().startsWith('platforms:')
inApiServer = false
continue
}
if (!inPlatforms) continue
if (indent <= 2) {
inApiServer = trimmed.trimStart().startsWith('api_server:')
continue
}
if (!inApiServer) continue
const t = trimmed.trimStart()
if (t.startsWith('enabled:')) {
const v = t.slice(8).trim().replace(/^['"]|['"]$/g, '').toLowerCase()
return ['true', 'yes', 'on', '1'].includes(v)
}
}
return false
},
_hermesPatchYamlEnsureApiServer(raw) {
if (this._hermesConfigHasApiServerEnabled(raw)) return raw
const lines = raw.split('\n')
const out = []
let platformsFound = false
let i = 0
while (i < lines.length) {
const line = lines[i]
const trimmed = line.replace(/\s+$/, '')
const indent = trimmed.length - trimmed.trimStart().length
if (indent === 0 && trimmed.trimStart().startsWith('platforms:')) {
out.push(line)
platformsFound = true
i++
const accumulated = []
let skipping = false
while (i < lines.length) {
const l = lines[i]
const t = l.replace(/\s+$/, '')
const ind = t.length - t.trimStart().length
if (ind === 0 && t !== '') break
if (ind <= 2) skipping = t.trimStart().startsWith('api_server:')
if (!skipping) accumulated.push(l)
i++
}
out.push(' api_server:')
out.push(' enabled: true')
out.push(...accumulated)
continue
}
out.push(line)
i++
}
if (!platformsFound) {
if (out.length && out[out.length - 1] !== '') out.push('')
out.push('platforms:')
out.push(' api_server:')
out.push(' enabled: true')
}
let content = out.join('\n')
if (!content.endsWith('\n')) content += '\n'
return content
},
_hermesEnsureApiServerEnabled() {
const configPath = path.join(hermesHome(), 'config.yaml')
if (!fs.existsSync(configPath)) return
const raw = fs.readFileSync(configPath, 'utf8')
if (this._hermesConfigHasApiServerEnabled(raw)) return
const ts = Math.floor(Date.now() / 1000)
const backupPath = configPath + `.bak-${ts}`
try { fs.writeFileSync(backupPath, raw) } catch {}
const patched = this._hermesPatchYamlEnsureApiServer(raw)
fs.writeFileSync(configPath, patched)
console.warn(`[hermes guardian] patched config.yaml (api_server.enabled). Backup: ${backupPath}`)
},
// =========================================================================
// .env editor (Step 4) — Web-mode implementations mirroring Rust behavior.
// The managed-key list is duplicated here since Rust's hermes_providers is

View File

@@ -1821,6 +1821,11 @@ pub async fn hermes_gateway_action(
let enhanced = hermes_enhanced_path();
match action.as_str() {
"start" => {
// Guardian: ensure platforms.api_server.enabled:true is present
// before every start. Auto-heal if missing (with a .bak backup).
// See `ensure_api_server_enabled` for rationale.
ensure_api_server_enabled(&app)?;
#[cfg(target_os = "windows")]
{
let home = hermes_home();
@@ -3139,6 +3144,185 @@ pub async fn hermes_memory_write(
Ok("ok".into())
}
// ============================================================================
// api_server guardian (Step 5 / G7)
//
// ClawPanel's Hermes integration requires `platforms.api_server.enabled: true`
// in ~/.hermes/config.yaml so that `hermes gateway run` exposes the
// /v1/runs endpoint we depend on. The setting is written once by
// `configure_hermes`, but several real-world scenarios can remove it:
// * User upgrades Hermes and the new default config.yaml is merged
// without the api_server platform entry.
// * User manually edits config.yaml (via Hermes CLI or text editor).
// * Migration scripts accidentally drop the section.
//
// Rather than silently failing at Gateway start time with an opaque
// "endpoint not found" error, this guardian checks before every start and
// auto-heals the config. A timestamped backup (config.yaml.bak-<epoch>)
// is written before any mutation so users can always roll back.
// ============================================================================
/// Scan a YAML string for `platforms.api_server.enabled: true` and return
/// true only when that exact path exists with a truthy value.
fn config_has_api_server_enabled(raw: &str) -> bool {
let mut in_platforms = false;
let mut in_api_server = false;
for line in raw.lines() {
// Strip comments (crude, but matches the simple YAML we write).
let line = match line.find('#') {
Some(i) => &line[..i],
None => line,
};
let trimmed = line.trim_end();
if trimmed.is_empty() {
continue;
}
let indent = trimmed.len() - trimmed.trim_start().len();
if indent == 0 {
in_platforms = trimmed.trim_start().starts_with("platforms:");
in_api_server = false;
continue;
}
if !in_platforms {
continue;
}
// Inside platforms:
if indent <= 2 {
in_api_server = trimmed.trim_start().starts_with("api_server:");
continue;
}
if !in_api_server {
continue;
}
// Inside platforms.api_server:
let t = trimmed.trim_start();
if let Some(rest) = t.strip_prefix("enabled:") {
let v = rest.trim().trim_matches(|c: char| c == '"' || c == '\'');
return matches!(v.to_ascii_lowercase().as_str(), "true" | "yes" | "on" | "1");
}
}
false
}
/// Produce a patched YAML that guarantees
/// `platforms.api_server.enabled: true` is present, preserving everything
/// else verbatim. If the config already has the setting (as `true`) this
/// returns the original text unchanged.
fn patch_yaml_ensure_api_server(raw: &str) -> String {
if config_has_api_server_enabled(raw) {
return raw.to_string();
}
// Strategy:
// * If `platforms:` exists, inject / replace api_server subtree under it.
// * Otherwise append a new top-level `platforms:` block at EOF.
let lines: Vec<&str> = raw.lines().collect();
let mut out: Vec<String> = Vec::with_capacity(lines.len() + 4);
let mut platforms_found = false;
let mut i = 0;
while i < lines.len() {
let line = lines[i];
let trimmed = line.trim_end();
let indent = trimmed.len() - trimmed.trim_start().len();
if indent == 0 && trimmed.trim_start().starts_with("platforms:") {
// Copy the platforms: header
out.push(line.to_string());
platforms_found = true;
i += 1;
// Accumulate children and drop the existing api_server subtree
// (we'll rewrite it at the top of the block). Keep siblings.
let mut accumulated_children: Vec<String> = Vec::new();
let mut skipping_api_server = false;
while i < lines.len() {
let l = lines[i];
let t = l.trim_end();
let ind = t.len() - t.trim_start().len();
if ind == 0 && !t.is_empty() {
break; // leaving platforms block
}
if ind <= 2 {
skipping_api_server = t.trim_start().starts_with("api_server:");
}
if !skipping_api_server {
accumulated_children.push(l.to_string());
}
i += 1;
}
// Inject a fresh api_server entry at the top of platforms:
out.push(" api_server:".into());
out.push(" enabled: true".into());
out.extend(accumulated_children);
continue;
}
out.push(line.to_string());
i += 1;
}
if !platforms_found {
if let Some(last) = out.last() {
if !last.is_empty() {
out.push(String::new());
}
}
out.push("platforms:".into());
out.push(" api_server:".into());
out.push(" enabled: true".into());
}
let mut content = out.join("\n");
if !content.ends_with('\n') {
content.push('\n');
}
content
}
/// Guardian called from `hermes_gateway_action` on every `start` request.
/// Returns Ok(()) when the config is healthy (either it was already correct
/// or the patch succeeded). Emits `hermes-config-patched` on auto-heal so
/// the frontend can display a transparent toast.
fn ensure_api_server_enabled(app: &tauri::AppHandle) -> Result<(), String> {
let config_path = hermes_home().join("config.yaml");
if !config_path.exists() {
// Nothing to guard — configure_hermes will create a compliant file
// on first run. Don't auto-create here; that's outside the guard's
// responsibility.
return Ok(());
}
let raw = std::fs::read_to_string(&config_path)
.map_err(|e| format!("Failed to read config.yaml: {e}"))?;
if config_has_api_server_enabled(&raw) {
return Ok(());
}
// Back up with a timestamped filename so we never overwrite an earlier
// .bak (rapid re-starts would lose history otherwise).
let ts = std::time::SystemTime::now()
.duration_since(std::time::UNIX_EPOCH)
.map(|d| d.as_secs())
.unwrap_or(0);
let backup_path = config_path.with_extension(format!("yaml.bak-{ts}"));
let _ = std::fs::write(&backup_path, &raw);
let patched = patch_yaml_ensure_api_server(&raw);
std::fs::write(&config_path, &patched)
.map_err(|e| format!("Failed to write config.yaml: {e}"))?;
// Inform the frontend so it can surface a toast. Failure to emit is
// non-fatal — the patch itself already succeeded.
use tauri::Emitter;
let _ = app.emit(
"hermes-config-patched",
serde_json::json!({
"kind": "api_server_enabled",
"backup": backup_path.to_string_lossy(),
"message": "platforms.api_server.enabled 缺失,已自动修复并备份原文件",
}),
);
Ok(())
}
// ============================================================================
// .env editor commands (Step 4 / G6)
//
@@ -3326,3 +3510,108 @@ pub fn hermes_env_delete(key: String) -> Result<(), String> {
std::fs::write(&env_path, content).map_err(|e| format!("Failed to write .env: {e}"))?;
Ok(())
}
// ============================================================================
// Unit tests for the pure YAML helpers (no filesystem I/O).
// ============================================================================
#[cfg(test)]
mod guardian_tests {
use super::{config_has_api_server_enabled, patch_yaml_ensure_api_server};
#[test]
fn detects_enabled_variants() {
let yaml = "\
model:
default: deepseek-chat
platforms:
api_server:
enabled: true
";
assert!(config_has_api_server_enabled(yaml));
for v in ["true", "True", "TRUE", "yes", "on", "1"] {
let y = format!("platforms:\n api_server:\n enabled: {v}\n");
assert!(
config_has_api_server_enabled(&y),
"expected {v} to count as enabled"
);
}
}
#[test]
fn detects_missing_or_disabled() {
assert!(!config_has_api_server_enabled("model:\n default: foo\n"));
assert!(!config_has_api_server_enabled(
"platforms:\n other:\n enabled: true\n"
));
assert!(!config_has_api_server_enabled(
"platforms:\n api_server:\n enabled: false\n"
));
assert!(!config_has_api_server_enabled(
"platforms:\n api_server:\n something: else\n"
));
}
#[test]
fn ignores_commented_enabled() {
let yaml = "platforms:\n api_server:\n # enabled: true\n";
assert!(!config_has_api_server_enabled(yaml));
}
#[test]
fn patch_is_noop_when_already_enabled() {
let yaml = "\
model:
default: x
platforms:
api_server:
enabled: true
";
assert_eq!(patch_yaml_ensure_api_server(yaml), yaml);
}
#[test]
fn patch_appends_when_no_platforms() {
let yaml = "model:\n default: x\n";
let patched = patch_yaml_ensure_api_server(yaml);
assert!(config_has_api_server_enabled(&patched));
assert!(patched.contains("model:"));
assert!(patched.contains("default: x"));
}
#[test]
fn patch_injects_under_existing_platforms() {
let yaml = "\
platforms:
other:
enabled: true
terminal:
backend: local
";
let patched = patch_yaml_ensure_api_server(yaml);
assert!(config_has_api_server_enabled(&patched));
assert!(patched.contains("other:"));
assert!(patched.contains("terminal:"));
assert!(patched.contains("backend: local"));
}
#[test]
fn patch_replaces_disabled_api_server() {
let yaml = "\
platforms:
api_server:
enabled: false
extra: keepme
other:
enabled: true
";
let patched = patch_yaml_ensure_api_server(yaml);
assert!(config_has_api_server_enabled(&patched));
assert!(patched.contains("other:"));
assert!(
!patched.contains("enabled: false"),
"disabled marker should have been removed"
);
}
}

View File

@@ -583,6 +583,14 @@ export function render() {
showGwMsg(evt.payload || '', false)
})
unlisteners.push(unlisten2)
// 监听 config.yaml 自愈事件api_server guardian
const unlisten3 = await tauriListen('hermes-config-patched', async (evt) => {
const { toast } = await import('../../../components/toast.js')
const msg = evt?.payload?.message || 'config.yaml 已自动修复'
toast(msg, 'info', { duration: 6000 })
})
unlisteners.push(unlisten3)
} catch (_) {
// Web 模式下无 Tauri 事件,静默忽略
}