Follow-up to the /memory approve fresh-store fix. Both the CLI fallback and
the messaging-gateway handler built a bare MemoryStore() with the hardcoded
default char limits (2200/1375), ignoring the user's configured
memory.memory_char_limit / user_char_limit. A live agent honors those
overrides (agent/agent_init.py), so an approval applied without a live agent
could accept a write the user's lower cap would reject, or vice versa.
Extract a shared tools.memory_tool.load_on_disk_store() factory that reads
the configured limits (falling back to defaults if config can't load) and
wire both the CLI and gateway handlers to it, closing the gap on both
surfaces and de-duplicating the construction block.
The CLI /memory slash handler (cli_commands_mixin._handle_memory_command)
passed self.agent._memory_store straight through, which is None when the
command runs without a live agent — e.g. /memory approve from the Desktop
GUI. The shared write-approval handler then returns "memory store
unavailable" and applies nothing, even with built-in memory enabled and
pending writes present.
Fall back to a freshly loaded on-disk MemoryStore when no live store is
available, mirroring the gateway path (gateway/slash_commands.py). It
persists to the same MEMORY/USER.md and creates MEMORY.md on the first
approved write.
Fixes#46783
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds an optional structured completion contract to the standing-goal loop,
adapted from OpenAI Codex's /goal guidance (a durable objective works best
when it names what done means, how to prove it, what not to break, what's in
scope, and when to stop).
A contract has five optional fields — outcome, verification, constraints,
boundaries, stop_when. When set, the continuation prompt tells the agent to
target the verification surface and respect constraints, and the judge marks
the goal done only when the verification criterion is met with concrete
evidence (command result, file excerpt, test output) instead of a loose
"looks done" claim. This tightens the most common /goal failure mode:
premature completion / endless over-continuation on an underspecified goal.
Two ways to set a contract, both backward compatible (bare /goal <text>
behaves exactly as before):
- /goal draft <objective> — expands plain text into a full contract via the
goal_judge aux model (cache-safe side call), falls back to a free-form goal
if the model is unavailable.
- /goal <text> with inline 'field: value' lines (verify:, constraints:,
boundaries:, stop when:, ...). Plain goals with an incidental colon are not
mangled — only known field prefixes are pulled out.
- /goal show prints the active contract.
Contracts persist in SessionDB.state_meta alongside the goal (survive /resume),
compose with /subgoal criteria, and old goal rows load unchanged. CLI + every
gateway platform via the shared GoalManager engine; zero new model tools.
Tests: +18 in tests/hermes_cli/test_goals.py (parse/serialize/judge-prompt/
draft/fallback), 73/73 green; 42/42 across the broader goal test surface;
live E2E roundtrip (set -> persist -> reload -> contract-aware prompts) green.
* feat(goals): add /goal wait <pid> barrier to park the loop on a background process
The /goal loop re-pokes the agent every turn via the post-turn judge. When a
goal is gated on a long-running background process (CI poller, build, test
matrix, deploy) that produces nothing to judge yet, this spins the agent into
'is it done?' busy-work and burns the turn budget.
/goal wait <pid> [reason] parks the loop: while the PID is alive, the judge is
skipped, no turn is consumed, no continuation fires, and /goal status shows a
parked indicator. The barrier auto-clears the moment the process exits (the
agent's notify_on_complete watcher is the natural wake signal), then the next
turn resumes normal judging. /goal unwait clears it manually; pause/resume/clear
drop it; a dead/stale PID can never wedge the loop.
Wired across CLI, gateway, and the mid-run command guard for parity. Barrier
persists in SessionDB.state_meta (survives /resume); GoalState gains
backward-compatible waiting_on_pid/waiting_reason/waiting_since fields. 12 new
tests; docs updated.
* fix(goals): use gateway.status._pid_exists for liveness, not os.kill(pid,0)
The Windows-footguns CI guard flagged os.kill(pid, 0) in _pid_alive — on
Windows that's not a no-op, it routes to CTRL_C_EVENT and hard-kills the
target's console process group (bpo-14484). Delegate to the canonical
footgun-safe gateway.status._pid_exists (psutil + ctypes/POSIX fallback)
instead, with a direct-psutil last resort.
* feat(goals): judge-driven auto-wait — the loop parks itself, no manual /goal wait
Makes the wait barrier automatic. Every turn the judge is shown the agent's
live background processes (pid, command, uptime, output tail from the
process_registry) alongside the goal + response, and can return a new 'wait'
verdict instead of continue:
{"verdict":"wait","wait_on_pid":N} → park until that process exits
{"verdict":"wait","wait_for_seconds":N} → park until the deadline passes
evaluate_after_turn acts on the directive (sets the barrier, parks the loop)
so the agent isn't re-poked into busy-work while CI/builds/deploys run. Adds a
time-based waiting_until barrier alongside the pid barrier; both auto-clear and
can never wedge the loop. Drivers (CLI, gateway, tui_gateway) feed the live
registry in via gather_background_processes(). Manual /goal wait stays as an
override. Judge verdict contract widened to (verdict, reason, parse_failed,
wait_directive); legacy {"done":bool} shape still accepted.
* test(goals): update kanban _fake_judge to the 4-tuple judge contract
CI test(3) caught it: test_kanban_goal_mode's _fake_judge still returned the
3-tuple (verdict, reason, parse_failed), but the kanban loop now unpacks the
4-tuple (+ wait_directive). Update the fake to return None for the directive
and accept the background_processes kwarg.
* feat(goals): trigger-based wait — park on a process's own signal, not just exit
Addresses two gaps in the judge-driven wait: (1) the judge could only express
'wait until PID exits' or 'wait N seconds', so a long-lived watcher/server that
fires a trigger MID-RUN (and may never exit) couldn't be waited on; (2) the
process's own watch_patterns/notify_on_complete trigger was invisible to the judge.
Adds a session-based barrier (waiting_on_session) that releases on the process's
OWN trigger via process_registry.is_session_waiting(): the session exits, OR (if
started with watch_patterns) its pattern matches — even while the process keeps
running. list_sessions() now surfaces session_id + watch_patterns/watch_hit/
notify_on_complete so the judge sees the trigger and is told to prefer
wait_on_session for trigger processes. Judge verdict gains a {wait_on_session}
directive (preferred over pid). Backward-compatible GoalState field; pid + time
barriers unchanged.
Tests: TestSessionTriggerBarrier (release on mid-run pattern match while alive,
release on exit, unknown-session, full park→trigger→resume, parse, validation,
backcompat load). 105 goal-surface + 85 process_registry tests green.
display.timestamps already drove the [HH:MM] suffix on live submitted and
streamed message labels, but there was no runtime command to toggle it and
/history ignored the setting entirely. Add /timestamps [on|off|status]
(alias /ts) and render [HH:MM] in /history for turns that carry a stored
unix timestamp (resumed sessions). Live unsaved turns without a stored time
are never given a fabricated one. Uses the existing sanctioned non-wire
'timestamp' message key (stripped before the API call in chat_completions),
so message-alternation and prompt-cache invariants are untouched.
* feat(cli): /prompt — compose your next prompt in $EDITOR
Adds /prompt (alias /compose): opens $VISUAL/$EDITOR on a temp markdown
file so you can hand-edit a multi-line prompt, then sends the saved buffer
as the next agent turn. Text after the command pre-seeds the buffer; an
empty save cancels. Reuses the one-shot _pending_agent_seed the interactive
loop already consumes (same mechanism as /blueprint), so no changes to the
input event loop or message pipeline. CLI-only.
* feat(tui): /prompt slash command opens $EDITOR (parity with CLI)
The TUI already opens $EDITOR via Ctrl+G (openEditor), but had no /prompt
slash command like the classic CLI. Wire openEditor into the slash handler
context and register /prompt (alias /compose) to call it; inline text after
the command is dropped into the composer first so it carries into the editor,
matching the CLI's /prompt <text>.
* feat(cli): /reasoning full to show complete thinking, not 10-line clamp
The post-response Reasoning recap box hard-clamped long thinking to the
first 10 lines, so there was no way to see the full reasoning trace after
a turn (live streaming already shows it in full). Add display.reasoning_full
(default off) plus /reasoning full|clamp to toggle it at runtime; the clamp
truncation note now points at the command. Addresses repeated user requests
to show all thinking tokens.
* test(gateway): de-snapshot /reasoning help assertion
The test froze the exact args-hint literal '/reasoning [level|show|hide]',
which the new full/clamp args change to '[level|show|hide|full|clamp]'.
Convert to an invariant: assert /reasoning is in help and carries its core
args, not the exact hint string.
* feat(tui): /reasoning full|clamp parity in tui_gateway
The classic-CLI reasoning_full toggle had no TUI equivalent — typing
/reasoning full in the TUI fell through to parse_reasoning_effort and
errored. The TUI renders thinking as an expand/collapse section (no fixed
10-line recap), so map full -> sections.thinking=expanded (raw, uncapped
via thinkingPreview mode='full') and clamp -> collapsed, persisting
display.reasoning_full for cross-surface config consistency.
* feat(providers): remove google-gemini-cli + google-antigravity OAuth providers
Google now actively bans accounts for third-party tools that piggyback on
Gemini CLI / Antigravity / Code Assist OAuth, and because abuse prevention
sits at a backend layer the ban can extend to the entire Google account
(Gmail/Drive), with a second violation being permanent.
Ref: https://github.com/google-gemini/gemini-cli/discussions/20632
Removes both OAuth inference providers entirely (modules, provider profiles,
auth/runtime/config/models wiring, the /gquota Code Assist quota command,
the antigravity-cli optional skill, desktop + docs surface in en + zh-Hans).
The API-key 'gemini' provider (GOOGLE_API_KEY/GEMINI_API_KEY against
generativelanguage.googleapis.com) is unaffected and stays fully supported.
* fix(skills): keep the antigravity-cli skill — only the OAuth provider is removed
The antigravity-cli optional skill orchestrates the external `agy` binary as
a coding-agent tool via the terminal tool — it does NOT wrap Hermes inference
through the banned google-antigravity OAuth provider, so it carries none of
the account-ban risk that motivated removing that provider. Restore the skill,
its docs page, the sidebar entry, and the optional-skills catalog row. The
google-antigravity / google-gemini-cli inference providers stay fully removed.
* feat(delegation): async background subagents via delegate_task(background=true)
delegate_task(background=true) dispatches a subagent that runs in the
background and returns a handle immediately, so the user and model keep
working while it runs. The full result — plus the original task source —
re-enters the conversation as a new turn when the subagent finishes,
riding the same completion-queue rail as terminal background processes.
- tools/async_delegation.py: daemon-executor registry, capacity cap,
rich self-contained completion event pushed onto the shared
process_registry.completion_queue (type='async_delegation').
- delegate_tool.py: background param + single-task dispatch branch;
batch async rejected (v1).
- process_registry.py: format_process_notification renders the rich
task-source block (goal/context/toolsets/model/status/result).
- gateway/run.py: dedicated _async_delegation_watcher drains + injects
results into the originating session (idle + post-turn), session_key
routing enrichment, shutdown interrupt of dangling delegations.
- config: delegation.max_async_children (default 3).
Reuses the existing idle-drain wiring rather than mutating a running
agent loop, preserving message-role alternation and prompt-cache
invariants. 13 targeted tests; CLI + gateway paths E2E-verified.
* test(delegation): make async non-blocking tests environment-independent
CI 'test (5)' flaked on a cold, 8-worker runner: the first
delegate_task(background=true) call measured 2.27s of one-time setup
(config load + child-agent construction + imports), tripping the
elapsed < 1.0 wall-clock assertion. That assertion was testing setup
overhead, not blocking.
Replace the wall-clock thresholds with the real invariant: dispatch
returns while the child is still gated (active_count == 1, completion
queue empty), which a synchronous impl could not do. Keep only a loose
4s sanity backstop well under the runner's 5s gate.
* fix(delegation): harden async background delegation
Follow-up review fixes:
- Detach background child from parent._active_children at dispatch —
otherwise parent-turn interrupts (Ctrl+C, mid-turn steering), cache
evicts (release_clients), and session close (/new) kill/close the
detached subagent mid-run, defeating the point of background mode.
Lifecycle is owned by the async registry's interrupt_fn.
- Make the capacity check atomic with the record insert (TOCTOU: two
concurrent dispatches could both pass active_count() and exceed the cap).
- TUI dedup: key async_delegation events by delegation_id — the
fallthrough keyed them all as ("", type), suppressing every completion
after the first in the desktop/TUI status feed.
- CLI /stop now interrupts running background delegations and /agents
lists them (they live outside the process registry and were invisible).
- Drop stray unbalanced ']' line from the re-injection block and the
unused _ASYNC_DEFAULT import.
Tests: detach-at-dispatch + concurrent-capacity race added (15 total in
test_async_delegation.py); 137 delegate + 140 process-registry/notify/watch
+ 7 TUI dedup tests pass.
* fix(delegation): harden async background completion drains
Reworks the chat-line UX: pick a recipe by name and the agent asks you for
what it needs, one question at a time, instead of forcing you to hand-type a
slot=val command line.
- /cron-recipe -> lists the catalog
- /cron-recipe <name> -> forgiving name match (exact/prefix/substring/
fuzzy; ambiguous lists candidates), then seeds
the agent with a natural-language fill request
built from the recipe's typed slots + schedule
and prompt templates. The agent asks for each
value one at a time and calls the EXISTING
cronjob tool. No new tool.
- /cron-recipe <name> slot=val -> unchanged deterministic path (fill_recipe ->
create_job) for the dashboard/docs/power user.
Mechanism (no new plumbing, invariant-safe — the seed enters as a normal user
turn, never a synthetic injection):
- shared handler returns RecipeCommandResult{text, agent_seed}; match_recipe()
and build_recipe_seed() are the new shared pieces.
- gateway: dispatch rewrites event.text to the seed and falls through to the
agent (the same pattern /steer uses).
- CLI: handler sets a one-shot self._pending_agent_seed; the interactive loop
consumes it right after process_command() and runs it as the next turn.
The typed-slot schema stays the single source of truth (still validates the
form/inline path via fill_recipe); the agent path just renders those slots into
the questions to ask. Docs updated to lead with the name-then-ask flow.
A 'recipe' is a one-place definition of an automation that every surface
renders natively. The slot schema (cron/recipe_catalog.py) is the single
source of truth; four renderers consume it, and all paths end at the same
cron.jobs.create_job — no second job engine.
Form where there's a screen, conversation where there's a chat line:
- Dashboard / GUI app: a Recipes sub-tab on the Cron page renders each
recipe's typed slots as a form (time-picker, enum dropdown, free-text);
submit POSTs /api/cron/recipes/instantiate which fills + creates the job.
- CLI / TUI / messengers: /cron-recipe lists the catalog, shows a recipe's
fields, or fills + creates from a pasted 'key slot=val' command. The shared
handler (hermes_cli/cron_recipe_cmd.py) names any missing/invalid slot so
the agent can ask a targeted follow-up.
- Docs: a generated Cron Recipes catalog page (website, .mdx + React cards)
shows each recipe with a copy-paste command and a 'Send to App' button.
- Desktop: a hermes:// URL scheme (Electron single-instance lock +
setAsDefaultProtocolClient + open-url/second-instance) routes
hermes://cron-recipe/<key>?slot=val into the chat composer pre-filled.
Typed slots (time/enum/text/weekdays) with defaults: users never type raw
cron — recipes parameterize time-of-day and weekday sets and translate to
cron expressions; a free-text 'schedule' slot is the full-flexibility escape
hatch. Consent-first throughout: nothing schedules without an explicit submit
or send.
Core:
- cron/recipe_catalog.py — CronRecipe + RecipeSlot, 5 curated recipes,
recipe_form_schema / recipe_slash_command / recipe_deeplink /
recipe_catalog_entry renderers, fill_recipe (validate + translate to
create_job kwargs).
- hermes_cli/cron_recipe_cmd.py — shared /cron-recipe handler (CLI + TUI +
gateway never drift). CommandDef + dispatch in commands.py / cli.py /
gateway/run.py.
Dashboard: GET /api/cron/recipes + POST /api/cron/recipes/instantiate
(web_server.py), CronRecipes.tsx gallery+form, Segmented sub-tab on CronPage,
api.ts methods + types.
Desktop: hermes:// scheme end to end (main.cjs deep-link router + ready-queue,
preload onDeepLink/signalDeepLinkReady, global.d.ts types, desktop-controller
composer prefill, electron-builder protocols key).
Docs: extract-cron-recipes.py generator wired into prebuild.mjs,
cron-recipes-catalog.mdx + CronRecipesCatalog React component, sidebar entry.
Generated index json gitignored like skills.json.
Tests: 23 core (catalog/slots/schedule-resolution/validation/renderers/command
handler/generator) + 5 web_server endpoint tests. E2E verified end to end:
slot fill -> create_job -> persisted job with correct schedule/deliver/origin.
The shipped tri-state write_mode (on|off|approve) conflated two concepts —
whether writes are enabled and whether they're gated — so 'on' (writes flow
freely, gate inactive) read like 'gating is on'. Replace it with a single
clear boolean gate that defaults off.
memory.write_approval / skills.write_approval:
false (default) — write freely; the approval gate is off (pre-gate behaviour)
true — require approval: memory foreground prompts inline, memory
background-review + all skill writes stage for review
The old 'off = block all writes' mode is dropped; memory_enabled: false already
disables memory entirely, so a third 'block' state was redundant.
- tools/write_approval.py: get_write_mode/MODE_* → write_approval_enabled() bool;
evaluate_gate() loses the config-driven 'blocked' path (blocked now only comes
from an interactive user denial).
- tools/memory_tool.py, tools/skill_manager_tool.py: comment + behaviour follow.
- hermes_cli/config.py: memory/skills write_mode → write_approval (False);
_config_version 28→29 with a 28→29 migration that renames any persisted
write_mode (approve→true, on/off/unset→false) and drops the old key.
- slash commands: '/memory|/skills mode <on|off|approve>' → 'approval <on|off>'
('mode' kept as a back-compat alias); set_mode_fn callback now takes a bool.
- write_approval_commands.py, cli_commands_mixin.py, gateway/slash_commands.py,
commands.py: handlers + registry args/subcommands updated.
- docs + tests rewritten for the boolean model; added migration tests.
Adds memory.write_mode and skills.write_mode (on|off|approve), applied to
both foreground turns and the background self-improvement review fork — the
source of the unprompted 'wrong assumption' saves users reported.
- on (default): write freely, unchanged behaviour
- off: never write; the tool returns a clean disabled result
- approve: don't commit. Memory foreground writes prompt inline (small,
reviewable in a chat bubble); background memory writes and ALL skill writes
stage to a pending store instead (a SKILL.md is too large to review inline,
and a daemon thread can't block on a prompt)
Review staged writes from CLI or any messaging platform:
/memory pending|approve|reject|mode
/skills pending|approve|reject|diff|mode
Skill review respects the size asymmetry: inline you see a one-line gist;
the full unified diff stays out-of-band (/skills diff, dashboard, or the
staged JSON file).
New: tools/write_approval.py (gate + pending store), hermes_cli/
write_approval_commands.py (shared CLI+gateway handlers). Gates wired at the
single entry points memory_tool() and skill_manage(), using the existing
write-origin ContextVar to distinguish foreground from background_review.
Lift the `_handle_*_command` cluster (2,077 LOC) out of HermesCLI into
hermes_cli/cli_commands_mixin.py; HermesCLI now inherits CLICommandsMixin so
every self.<handler> call resolves unchanged via the MRO. Behavior-neutral.
Import discipline mirrors gateway/slash_commands.py (PR #41886): neutral deps
imported at the mixin module top level; cli.py-internal helpers/constants
(_cprint, _ACCENT, save_config_value, ...) imported lazily inside each handler
via 'from cli import ...' so the mixin never imports cli at module scope.
cli.py 16215 -> 14139 LOC. One test mock repointed (cli.is_browser_debug_ready
-> hermes_cli.cli_commands_mixin.is_browser_debug_ready).
