HexLab98 6bbe55dd09 fix(gateway): bound the agent cache by memory, not just count and age ... The per-session agent cache is capped at 128 entries with a 1h idle TTL, and neither bound knows how many bytes it holds. Each cached agent pins _session_messages -- the full transcript including tool output, tens of MB on a session with 100+ tool calls -- so a gateway serving many chats keeps every warm transcript resident: agents that took a turn inside the TTL are never idle-swept, and the idle sweep additionally defers finalizable sessions until they expire. RSS climbs until the cgroup throttles and SIGTERM can no longer flush inside systemd's stop timeout. Add the missing bound. Each session-expiry watcher tick compares the process's anonymous RSS against a budget and, when over, sheds LRU agents through the same soft-eviction path the cap enforcer uses, then runs malloc_trim so the freed arenas actually return to the OS. Evicted sessions rebuild their transcript from the persisted session on the next turn. Three classes of session are never shed: agents mid-turn, the most recently used ones, and any session whose transcript has not finished reaching disk (_last_flushed_db_idx vs len(_session_messages) -- the same divergence the FTS write-corruption guard reacts to when it preserves live history). memory_high_mb defaults to "auto", deriving the budget from the cgroup limit the gateway runs under, so a MemoryHigh/MemoryMax on the unit is respected without a second number to keep in sync. The two existing bounds become configurable alongside it under agent.agent_cache. protect_recent is clamped to half the cache: a couple of sessions can exhaust the budget on their own, and a fixed MRU guard would then protect everything and leave the gateway climbing with nothing it would shed. Fixes #80764		2026-08-07 21:02:33 +05:30
..
dashboard_auth	…
observability	Merge updated client resource metrics into active-install metrics	2026-08-04 15:10:41 -07:00
proxy	…
subcommands	…
web_routers	refactor(cron): polish registration partial-failure surfaces	2026-08-07 17:45:06 +05:30
__init__.py	chore: release v0.20.0 (2026.8.3)	2026-08-03 09:57:23 -07:00
_early_recovery.py	…
_parser.py	feat(kanban): let a task pin its own thinking depth	2026-08-01 16:09:56 -05:00
_scan_venv_blockers.py	…
_startup_fast.py	refactor: single chokepoint for the pre-import version fast path	2026-08-02 21:13:12 +05:30
_subprocess_compat.py	chore: suppress windows-footgun false positive on gated killpg	2026-08-04 17:33:36 -07:00
active_sessions.py	…
agent_import.py	refactor(cli): review follow-ups for the config.yaml import guard	2026-08-02 14:55:32 +05:30
approval_mode.py	…
approvals_suggest.py	…
auth.py	fix: show explicit member spend cap message instead of 'no credits'	2026-08-07 19:47:40 +05:30
auth_commands.py	…
azure_detect.py	…
backup.py	fix(backup): serialize and atomically publish snapshots	2026-08-03 23:48:55 +05:30
bang_shell.py	…
banner.py	fix(cli): route keyless first run into provider onboarding instead of a broken chat	2026-08-01 15:34:16 -07:00
blueprint_cmd.py	refactor(cron): polish registration partial-failure surfaces	2026-08-07 17:45:06 +05:30
browser_connect.py	…
build_info.py	…
bundles.py	…
callbacks.py	fix(approval): classify CLI/TUI approval timeouts separately from explicit denials	2026-08-02 20:21:59 -07:00
checkpoints.py	…
claw.py	…
cli_agent_setup_mixin.py	fix(agent): finish the #80622 bug class — sibling predicates, refund ordering, prompt carve-out, honest skip response	2026-08-07 19:44:35 +05:30
cli_billing_mixin.py	…
cli_commands_mixin.py	fix(agent): finish the #80622 bug class — sibling predicates, refund ordering, prompt carve-out, honest skip response	2026-08-07 19:44:35 +05:30
cli_output.py	…
clipboard.py	…
codex_models.py	…
codex_runtime_plugin_migration.py	…
codex_runtime_switch.py	…
colors.py	…
commands.py	feat: /refine — run the memory/skill self-improvement review on demand	2026-08-05 22:40:51 -07:00
completion.py	…
config.py	fix(config): make get_env_value scope-aware — the last scope-blind credential reader	2026-08-01 16:42:51 -07:00
config_defaults.py	fix(gateway): bound the agent cache by memory, not just count and age	2026-08-07 21:02:33 +05:30
config_migrations.py	…
console_engine.py	fix(console): handle string SystemExit code in _capture_output	2026-08-05 11:48:03 +05:30
container_boot.py	fix(install): support non-pid-1 container entrypoints	2026-08-01 10:52:34 -07:00
context_switch_guard.py	…
copilot_auth.py	test: harden cold-start regression tests + debug-log the env-var skip	2026-08-03 21:16:09 +05:30
credential_lifecycle.py	…
cron.py	…
curator.py	…
curses_ui.py	…
dashboard_procs.py	…
dashboard_register.py	…
debug.py	fix(debug): say where a client-side log lives instead of "(file not found)" (#78687)	2026-08-04 11:53:32 -06:00
default_soul.py	…
dep_ensure.py	fix(runtime): resolve Hermes-managed Node and uv before bare PATH	2026-08-01 21:17:51 -04:00
diagnostics_upload.py	…
dingtalk_auth.py	…
doctor.py	feat(providers): add Actual Computer inference provider	2026-08-05 14:08:32 -07:00
dump.py	feat(gateway): session activity watchdog, stall notify, compress timeout (#72424)	2026-08-02 16:16:36 -07:00
env_loader.py	fix(env): make config.yaml authoritative for terminal.backend (#29186)	2026-08-02 18:21:58 -07:00
fallback_cmd.py	…
fallback_config.py	…
focus_view.py	…
gateway.py	fix: align helper PID check with Python parser + dedupe drain-wait	2026-08-07 06:55:27 +05:30
gateway_enroll.py	…
gateway_windows.py	…
goals.py	feat(goals): quality gates — deterministic commands that must pass before /goal completes	2026-08-05 22:32:39 -07:00
gui_uninstall.py	feat(desktop): register a Linux launcher entry for `hermes desktop`	2026-08-05 11:47:05 -04:00
heartbeat.py	feat: /heartbeat — recurring session re-entry prompt fired when idle	2026-08-05 22:32:55 -07:00
hooks.py	feat(hooks): outbound webhooks — push signed lifecycle events to external HTTP endpoints	2026-08-02 15:01:11 -07:00
init_command.py	…
input_sanitize.py	…
inventory.py	…
journey.py	…
kanban.py	…
kanban_db.py	fix(tui): avoid writable Kanban opens on empty polls	2026-08-03 20:40:28 +05:30
kanban_decompose.py	…
kanban_diagnostics.py	…
kanban_specify.py	…
kanban_swarm.py	…
lifecycle.py	…
linux_desktop_entry.py	feat(desktop): register a Linux launcher entry for `hermes desktop`	2026-08-05 11:47:05 -04:00
logs.py	…
main.py	feat(update): emit an action-scoped terminal receipt from hermes update	2026-08-05 10:34:18 -06:00
managed_scope.py	…
managed_uv.py	fix(update): survive a hermes_constants cached from before the update	2026-08-01 23:06:47 -05:00
mcp_catalog.py	fix(catalog): validate http+api_key manifests declare the header's env key	2026-08-03 22:55:36 +05:30
mcp_config.py	…
mcp_picker.py	…
mcp_security.py	…
mcp_startup.py	fix(mcp): ensure MCP discovery completes before agent build in non-interactive sessions	2026-08-01 12:27:31 +05:30
mem_trim.py	fix(mem): 5s force floor so delegate-batch closes don't stack trims	2026-08-02 22:44:38 +05:30
memory_oauth.py	…
memory_setup.py	…
middleware.py	…
migrate.py	…
moa_cmd.py	…
moa_config.py	…
model_catalog.py	perf(model-picker): serve stale model caches instantly, refresh in background	2026-08-01 14:52:18 -07:00
model_cost_guard.py	…
model_normalize.py	fix(models): a model id missing its vendor prefix says so instead of 404ing (#78856)	2026-08-04 19:35:57 +00:00
model_search.py	…
model_setup_flows.py	…
model_switch.py	fix(model-switch): treat models dict as metadata, not allowlist	2026-08-04 08:52:31 -07:00
models.py	feat(providers): add Actual Computer inference provider	2026-08-05 14:08:32 -07:00
nous_account.py	fix: show explicit member spend cap message instead of 'no credits'	2026-08-07 19:47:40 +05:30
nous_auth_keepalive.py	…
nous_billing.py	…
nous_subscription.py	…
npm_engine.py	Merge branch 'main' into ethie/bundled-node-path-windows-layout	2026-08-01 21:27:26 -05:00
onepassword_secrets_cli.py	…
oneshot.py	fix(mcp): ensure MCP discovery completes before agent build in non-interactive sessions	2026-08-01 12:27:31 +05:30
pairing.py	…
partial_compress.py	…
pets.py	…
platforms.py	…
plugins.py	fix(gateway): split check_fn (passive probe) from ensure_deps_fn (active installer)	2026-08-07 13:28:43 +05:30
plugins_cmd.py	…
portal_cli.py	…
profile_describer.py	…
profile_distribution.py	refactor(utils): move mode+owner preservation into atomic_write_text	2026-08-06 05:00:17 +05:30
profiles.py	refactor: trim verbose comments + drop redundant default_flow_style kwarg	2026-08-05 11:59:33 +05:30
projects_cmd.py	…
projects_db.py	…
prompt_size.py	fix(prompt_size): search volatile tier for skills block after the stable->volatile move	2026-08-03 23:23:39 +05:30
prompt_stash.py	…
provider_catalog.py	…
providers.py	feat(providers): add Actual Computer inference provider	2026-08-05 14:08:32 -07:00
proxy_cli.py	…
psutil_android.py	…
pt_input_extras.py	…
pty_bridge.py	…
pty_session.py	…
relaunch.py	…
route_identity.py	…
runtime_provider.py	fix: adapt Actual provider salvage to current main	2026-08-05 14:08:32 -07:00
secret_prompt.py	…
secrets_cli.py	…
security_advisories.py	…
security_audit.py	fix(security): cache OSV malware preflight verdicts and stop double component discovery (#75485)	2026-08-01 10:47:20 -07:00
security_audit_startup.py	…
send_cmd.py	feat(gateway): simplex channel enumeration + show configured platforms in hermes send --list	2026-08-02 15:08:45 -07:00
service_manager.py	…
session_export.py	…
session_export_html.py	…
session_export_md.py	…
session_filters.py	…
session_listing.py	…
session_recap.py	…
session_recovery.py	fix(state): deduplicate session system prompts	2026-08-03 20:37:17 +05:30
sessions_cmd.py	fix(cli): back up state.db before clean-markers writes by default	2026-08-04 11:26:15 +05:30
setup.py	fix(runtime): resolve Hermes-managed Node and uv before bare PATH	2026-08-01 21:17:51 -04:00
setup_hidden_env.py	…
setup_whatsapp_cloud.py	…
skills_config.py	…
skills_hub.py	refactor: share category detection and prune vendored SKILL.md hits	2026-08-02 13:31:16 +05:30
skin_cmd.py	refactor: trim verbose comments + drop redundant default_flow_style kwarg	2026-08-05 11:59:33 +05:30
skin_engine.py	…
slack_cli.py	…
slash_exec.py	…
sqlite_runtime.py	…
sqlite_safe_read.py	fix(cli): untrack sqlite connections only after close succeeds	2026-07-31 23:18:12 -07:00
sqlite_util.py	…
status.py	fix: address self-review findings on the check_fn/ensure_deps_fn split	2026-08-07 13:28:43 +05:30
stdio.py	…
suggestions_cmd.py	refactor(cron): polish registration partial-failure surfaces	2026-08-07 17:45:06 +05:30
telegram_managed_bot.py	…
timefmt.py	chore: round-2 review nits (re-review #9)	2026-08-02 16:16:36 -07:00
timeouts.py	…
tips.py	fix(gateway): defer in-band restart until active turns finish (#77184)	2026-08-03 09:57:58 +05:30
tools_config.py	fix(tools): reuse subscription features for toolset listing	2026-08-03 16:07:16 +05:30
toolset_validation.py	…
uninstall.py	refactor(utils): move mode+owner preservation into atomic_write_text	2026-08-06 05:00:17 +05:30
update_cmd.py	feat(update): emit an action-scoped terminal receipt from hermes update	2026-08-05 10:34:18 -06:00
update_lock.py	…
urllib_security.py	…
vercel_auth.py	…
voice.py	…
web_deps.py	…
web_git.py	…
web_models.py	feat(profiles): REST export/import + extra_files overlay hook	2026-08-04 12:07:29 -06:00
web_server.py	refactor(cron): polish registration partial-failure surfaces	2026-08-07 17:45:06 +05:30
webhook.py	…
win_pty_bridge.py	…
windows_ssh_runtime.py	fix(desktop): keep Windows SSH runtime at machine root	2026-08-01 14:30:16 -07:00
write_approval_commands.py	…
xai_retirement.py	refactor(utils): move mode+owner preservation into atomic_write_text	2026-08-06 05:00:17 +05:30