The 30-slot default could not fit Hermes's ~50 built-in commands, so
every skill command (and 20 built-ins) were silently dropped from the
Telegram \`/\` menu by default — they only worked when typed manually.
Raising the default to 60 keeps all built-ins plus common skill commands
visible out of the box while staying under Telegram's ~4KB payload limit.
Users can still tune it via platforms.telegram.extra.command_menu.
Adds a configurable Telegram BotCommand menu cap and priority list via
platforms.telegram.extra.command_menu (max_commands clamped 1..100;
priority_mode prepend|append|replace). Default cap stays 30; hidden
commands remain invokable when typed and /commands lists the full set.
Salvaged from PR #42021. Cherry-picked onto current main; the original
edited gateway/platforms/telegram.py, now relocated to
plugins/platforms/telegram/adapter.py.
Selective --clone / --clone-from / --clone-config copied .env but not
auth.json, silently dropping the credential pool — including OAuth tokens
(Anthropic `claude /login`, Codex, xAI) that never land in .env. A profile
cloned from an OAuth-authenticated default therefore resolved a different
provider (or none) than the source under provider: auto. --clone-all already
carried auth.json via the full copytree; only the selective path missed it.
Add auth.json to _CLONE_CONFIG_FILES and tighten it to 0o600 after copy,
matching .env semantics.
On Windows, start_server() served uvicorn via a bare asyncio.run(_serve()),
which uses the default ProactorEventLoop. uvicorn's socket-serving stack
assumes a SelectorEventLoop on win32 (uvicorn/loops/asyncio.py forces it, and
uvicorn.Server.run threads config.get_loop_factory() into its runner for
exactly this reason). Driving uvicorn on the proactor loop makes
server.startup() bind a socket that never accepts: the dashboard and desktop
backend print "Skipping web UI build" then hang forever with the port
LISTENING but no TCP handshake completing.
Fix is win32-scoped to keep the blast radius minimal: POSIX keeps the exact
asyncio.run(_serve()) it had (its default loop is already a SelectorEventLoop /
uvloop, which is what uvicorn serves on). Only on Windows do we mirror
uvicorn.Server.run and run on the loop factory uvicorn picks, with a fallback
to WindowsSelectorEventLoopPolicy for uvicorn < 0.36.
Fixes hermes dashboard and hermes desktop (the Electron app spawns a
hermes dashboard backend). The gateway symptom in the report has a separate
root cause (no uvicorn) and is not addressed here.
The dashboard Profiles view showed "Gateway stopped" for a gateway that
is in fact running — while the sidebar status strip and `hermes gateway
status` (CLI) both correctly showed it running. Reported on v0.17.0
running the gateway + dashboard in one Docker container.
Root cause: three liveness surfaces with three detection strengths, all
reading the same `gateway.pid`:
- `hermes gateway status` -> find_gateway_pids() (process-table scan)
- sidebar /api/status -> get_running_pid() + gateway_state.json PID
fallback + health-URL probe
- Profiles view -> _check_gateway_running() = get_running_pid()
ONLY, no fallback
`get_running_pid()` short-circuits to None the moment the runtime lock
(`gateway.lock`) doesn't register as held by the *calling* process —
which is always true when the reader is a separate process from the
gateway (the dashboard is its own s6 service in the container), and also
for any launch-service-managed gateway that left a fresh
`gateway_state.json` but no live PID file. So the Profiles view alone
reported the live gateway as stopped.
Fix: give _check_gateway_running the same fallback the sidebar already
has — after the pid-file/lock check misses, validate the PID recorded in
that profile's gateway_state.json against the live process table via the
existing get_runtime_status_running_pid(). read_runtime_status() gains an
optional path arg so a profile's state file can be read without mutating
the process-global HERMES_HOME (preserving the contextvar-based profile
isolation the dashboard relies on). Backward compatible: every existing
caller passes no argument.
Tests: a regression test that fails pre-fix (live gateway, lock check
returns None -> must still report running) and a guard test that a
'stopped' state file is never reported running even with a live PID.
Profiles without their own messaging token inherit the default
profile's token via os.getenv, hit a token collision, and exit with
startup_failed. s6 restarts them immediately, creating ~30MB tirith
sandbox dirs in /tmp each cycle — filling the disk in hours (#51228).
Changes:
- gateway/restart.py: add GATEWAY_FATAL_CONFIG_EXIT_CODE = 78
- gateway/run.py: set exit_code=78 on non-retryable startup errors
(token collision, no platforms)
- hermes_cli/service_manager.py: add _render_finish_script() that
translates exit 78 → exit 125 (s6 permanent failure)
- hermes_cli/container_boot.py: write finish script alongside run
script during profile registration
The s6 finish script pattern follows docker/s6-rc.d/dashboard/finish.
Closes#51228
The cron ticker only runs inside the gateway (_start_cron_ticker); there
is no standalone cron daemon. When the gateway isn't running, next_run_at
passes but jobs never fire and last_run_at stays null — and manual
'hermes cron run' (which bypasses the ticker) appears to work, masking
the real cause. This is the most common cron support report (#51038).
cron list already warned; extend the same warning to cron create (the
moment the user is most likely to hit this) via a shared helper, and add
a pointer to 'hermes cron status'. Silent when a gateway is running, so
the gateway /cron path is unaffected.
Builds on @wgu9's runtime-tracking fix: now that find_gateway_pids() can
see a no-supervisor `gateway restart` runtime, have stop_profile_gateway()
fall back to an orphan-aware, profile-scoped reap (SIGTERM then SIGKILL)
when the pidfile/runtime record is missing or stale. Closes the duplicate-
accumulation path in #51325 — a follow-up restart now kills the prior
orphan instead of stacking another listener on :8644. Gated on
not supports_systemd_services() so a transient `gateway restart` argv on
supervised hosts is never killed.
Also adds the AUTHOR_MAP entry for the salvaged contributor.
_ensure_uv_for_termux only checked resolve_uv() (the managed
$HERMES_HOME/bin/uv) before falling back to pip, so a uv installed via
`pkg install uv` lives on PATH but is invisible to the helper. Combined
with the cherry-picked wheel-only fallback, a Termux user with no managed
uv still hit `pip install uv`, which has no Android wheel and tried to
source-build the Rust crate, OOM-killing low-memory devices.
Probe shutil.which("uv") right after the Termux guard and reuse it before
pip. Add a regression test that keeps resolve_uv() returning None while a
uv exists on PATH and asserts pip is never invoked.
Register a per-instance wakeUrl and forward it to the connector at
self-provision so a suspended gateway can be poked awake when buffered
work arrives (pairs with the connector-side WakePoker).
- relay_wake_url() resolver (env GATEWAY_RELAY_WAKE_URL, then
gateway.relay_wake_url in config.yaml), mirroring relay_instance_id()
- thread wake_url through _post_provision (adds wakeUrl to the body only
when set) + self_provision_relay (resolve, forward, log)
- hermes gateway enroll --wake-url <url> persists GATEWAY_RELAY_WAKE_URL
- document the §5.2 wake poke in relay-connector-contract.md §3.3
- tests: relay_wake_url resolution (env/config/absent), provision
forwarding, body-only-when-set (6 new; 130 relay tests pass)
The actual reconnect+drain on wake is Unit B's loop; this unit only
wires the wake SIGNAL. Opt-in: absent wakeUrl => connector never pokes.
* fix(windows): harden gateway scheduled task
* fix(windows): launch gateway scheduled task via console-less wscript
The Scheduled Task ran the gateway through cmd.exe, which allocates a
console. During logon Windows broadcasts CTRL_CLOSE_EVENT to console
process groups, reaping cmd.exe and the half-initialized gateway with
STATUS_CONTROL_C_EXIT (0xC000013A) - which Task Scheduler treats as a
user cancel, so RestartOnFailure never fires and the gateway vanishes on
every reboot (issue #45599 root cause #1).
Add a console-less .vbs launcher (wscript.exe -> pythonw.exe, both
GUI-subsystem) mirroring the gateway.cmd env + argv, and point the task
action at it. The .cmd stays for the Startup-folder fallback and /Run.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Jeff <jeffrobodie@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Open-ended skill learning across every surface. /learn <free text> takes a
description of any source — a directory, a URL, the workflow you just walked
the agent through, or pasted notes — and the live agent gathers it with the
tools it already has (read_file/search_files, web_extract, the conversation,
the pasted text), then authors a SKILL.md via skill_manage following the
house authoring standards (<=60-char description, the standard section order,
Hermes-tool framing, no invented commands).
No engine, no model-tool footprint, works on any terminal backend (local,
Docker, remote): /learn builds a standards-guided prompt and hands it to the
agent as a normal turn.
- agent/learn_prompt.py: shared standards-guided prompt builder
- /learn registry entry (both surfaces) + CLI handler (inject onto input
queue) + gateway handler (rewrite turn, fall through, /blueprint pattern)
- tui_gateway command.dispatch returns a send directive -> TUI + dashboard chat
- dashboard Skills page 'Learn a skill' panel (dir + URL + open-ended text)
composes a /learn request and runs it in chat
- docs (slash-commands ref + skills feature page), 11 targeted tests
Inspired by OpenAI Codex's Record & Replay and the /learn concept from #47234
(dir-distillation engine); reworked to be open-ended and engine-free per
review.
When a session rotates id on compression, _sync_session_key_after_compress()
re-anchored the session_key, approval-notify routing, yolo state, and slash
worker — but never moved the active-session lease, which stayed keyed to the
pre-compression id. And _find_live_session_by_key() matched live sessions on
the stale session_key, not the live agent's current agent.session_id. After
compression a resume/create path failed to recognize the existing live agent
and could build a SECOND live agent against the same DB continuation -> forked
lineage / cross-session message mixing.
- active_sessions.transfer_active_session(): move a lease in place to the new
id under the exclusive file lock (no slot drop).
- gateway _transfer_active_session_slot(): call it inside
_sync_session_key_after_compress(); on the rare fallback (entry pruned)
RESERVE the new slot before releasing the old lease (reserve-before-release),
so a concurrent gateway at the session cap cannot grab the freed slot in a
release-then-reacquire window and leave this session with no lease; if the
reserve fails, keep the existing lease (review fix).
- _session_lookup_key(): make live-session lookup authoritative on
agent.session_id, wired into all stale-session_key consumers
(_find_live_session_by_key, _session_live_item, _live_session_payload) —
fixes the whole lookup class.
Co-authored-by: kshitijk4poor <82637225+kshitijk4poor@users.noreply.github.com>
By default `hermes slack manifest` opts the app into Slack's AI Assistant
container (assistant_view feature + assistant:write scope +
assistant_thread_* events). Slack then renders DMs as the right-hand
Assistant split-pane, where every exchange is a thread and bare slash
commands (/help, /new, ...) are not delivered as normal command events —
they only work when the bot is @mentioned. There was no way to opt out
short of hand-editing the generated JSON.
Add --no-assistant to emit a flat-DM manifest that omits those three
pieces, so DMs render as a normal chat and slash commands dispatch
inline. The regular messaging surface (Messages tab, slash commands,
Socket Mode, channel + DM scopes/events) is preserved in both modes.
Default behaviour is unchanged (assistant mode still on).
Tests: cover both manifest modes and the argparse wiring.
Adds a per-platform display.reasoning_style setting (code | blockquote |
subtext) controlling how the show_reasoning summary renders on the gateway.
Discord defaults to "subtext" (-# small grey metadata text); every other
platform keeps the fenced code block. Resolves through the existing
display.platforms.<platform>.reasoning_style override chain.
* Revert "fix(cron): scope job execution to its owning profile (#32091 follow-up) (#50993)"
This reverts commit 660e36f097.
* Revert "fix(cron): anchor cron storage at the default root home (not the active profile)"
This reverts commit a5c09fd176.
* feat(memory): OAuth token storage and refresh for the Honcho provider
* feat(memory): refresh the Honcho OAuth token in the client and session
* feat(memory): zero-CLI loopback OAuth authorization flow
* feat(memory): generic memory-provider OAuth connect endpoints
* feat(desktop): memory-provider OAuth connect link
* feat(memory): CLI OAuth sign-in with source-tagged authorize links
* fix(memory): IP-literal loopback redirect and consent config_path on the authorize link
* fix(memory): profile-scope the memory-provider OAuth endpoints
* refactor(desktop): generic memory-provider OAuth client functions
* docs(memory): trim OAuth module docstrings to the invariants
* docs(memory): document OAuth connect as an optional auth method
* fix(memory): send home-relative display path to consent, not the absolute path
* perf(memory): cache OAuth token expiry in memory to skip the hot-path disk read
* fix(memory): log OAuth refresh failures at warning, not debug
* feat(memory): fall back to an OS-assigned loopback port when 8765 is taken
* test(memory): cover the desktop Connect launcher, status, and provider dispatch
* fix(desktop): keep the memory-provider dropdown one size regardless of connect state
* fix(desktop): move the memory connect link to the description line, leaving the dropdown untouched
* refactor(memory): move OAuth connect routes out of web_server into a memory-layer router
* refactor(desktop): import MemoryConnect directly, drop the single-export barrel
* fix(memory): launch CLI OAuth sign-in right after the auth choice, not after the wizard
* fix(desktop): auto-clear the OAuth error state instead of leaving it sticky
* test(honcho): isolate auth-method prompt from deployment-shape wizard tests
main's wizard suite scripts the cloud prompts without the OAuth auth-method step; auto-answer it in the shared helper so the answer lists stay shape-only.
* docs(honcho): document query-adaptive reasoning level (reasoningHeuristic)
README never mentioned reasoningHeuristic and listed reasoningLevelCap as an orphaned cap with the wrong default (— vs "high"). Add the query-adaptive scaling note + the reasoningHeuristic/reasoningLevelCap rows (grouped under Dialectic & Reasoning), matching the wording already on the hosted honcho.md page, and add a pointer from the memory-providers overview.
* fix(honcho): default the CLI peer prompt to the OAuth consent name
The CLI runs the grant with apply_config=False, so the peerName the user just entered at consent was dropped and the wizard's 'Your name' prompt fell back to $USER. Surface it as a transient OAuthCredential.consent_peer_name (set even when config isn't merged) and seed the prompt default from it.
* feat(honcho): split OAuth client_id by surface (cli=hermes-agent, desktop=hermes-desktop)
resolve_endpoints now picks the client_id from the initiating surface and
threads it through authorize -> token exchange -> persisted grant -> refresh,
so the CLI and desktop register as distinct OAuth clients. Surface-specific
env overrides (HONCHO_OAUTH_CLIENT_ID_CLI/_DESKTOP) win over the generic
HONCHO_OAUTH_CLIENT_ID, which still overrides every surface.
* feat(honcho): show OAuth vs API key in status; detect existing OAuth in setup
status now prints 'Auth: OAuth (clientId, token valid Xm/expired)' instead of
masking the OAuth access token as a generic API key; setup notes an existing
OAuth grant when re-run.
* docs(honcho): drop 'shared pool' wording from unified observation mode help
* fix(honcho): cross-process lock around OAuth refresh to prevent grant revocation
The in-process threading lock can't stop a sibling process (another profile or
the desktop app sharing honcho.json) from replaying the single-use refresh
token and tripping reuse-detection, which revokes the whole grant. Guard the
read-refresh-persist section with an OS file lock on <config>.lock so only one
process rotates at a time; the others re-read the freshly-persisted token.
Best-effort: platforms without flock degrade to in-process serialization.
* refactor(honcho): one OAuth client (hermes-agent) for all surfaces
Collapse the per-surface client_id split. CLI and desktop now use a single
client_id (hermes-agent); consent branding/UI still adapt via the source query
param. One grant identity means no clientId-vs-refresh-token desync that could
get the grant revoked. HONCHO_OAUTH_CLIENT_ID still overrides for self-hosting.
* fix(honcho): per-session resolves to session_id, never remapped by title
Reorder resolve_session_name so stable identifiers win over labels: gateway
per-chat key first, then the per-session session_id, then the cwd map / title.
A (possibly auto-generated) title can no longer remap a live per-session
conversation onto a second Honcho session mid-stream — fixes the desktop, which
is per-conversation via session_id. Consequence: a gateway's per-chat key now
also wins over a title (titles never remap a stable id).
The card was macOS-only. cua-driver also runs on Windows and Linux, so
fold `cua-driver doctor` (cross-platform binary/health probes) into a
single OS-aware `ready` signal:
- macOS: ready == both TCC grants; keeps the permission rows + grant flow.
- Windows/Linux: no TCC toggles, so ready == driver health, with a
per-OS note (SmartScreen/UIAccess on Windows; X11/XWayland on Linux).
`computer_use_status()` replaces the macOS-only `permissions_status()` and
surfaces `platform`, `ready`, `can_grant`, and the doctor `checks` (non-ok
ones render as warnings). CLI `permissions status`, the REST endpoint, and
the desktop card all key off the one payload. Grant stays macOS-only (400
elsewhere — nothing to grant).
Computer Use already worked through the desktop backend (the cua-driver
toolset enables + installs via Settings -> Skills & Tools), but there was
no in-app way to see or grant the two macOS permissions it needs, so "give
a model my Mac" was tribal knowledge.
The grants attach to cua-driver's OWN TCC identity (com.trycua.driver /
the installed CuaDriver.app), not Hermes -- so no app entitlement is
involved. cua-driver 0.5+ exposes `permissions status/grant`, which we wrap:
- tools/computer_use/permissions.py: thin client over the two subcommands
- hermes computer-use permissions {status,grant}: CLI parity
- GET /api/tools/computer-use/status, POST .../permissions/grant: desktop REST
- ComputerUsePanel: live Accessibility + Screen Recording state with a
Grant button (dialog attributed to CuaDriver), shown in the expanded
Computer Use toolset row. Binary install stays in the existing provider
post-setup runner.
Follow-ups: i18n the card copy; a "Stop driver" control (cua-driver stop)
for the runaway-`serve` case.
Adds auxiliary.background_review.{provider,model} (default auto = main chat
model — unchanged). Set it to a different, cheaper model and the post-turn
self-improvement review runs there for ~3-5x lower cost.
Cache-aware by design: the main chat is warm in the prompt cache, so the
default full-history replay on the main model is cheap cache reads — left
exactly as-is. A different model can't reuse that cache (different key), so
when (and only when) routed to a different model the fork replays a compact
digest instead of the full transcript, minimising what it cold-writes on the
aux model. Same model -> full replay; different model -> digest.
Quality holds in benchmarks: memory capture identical, skill near-identical.
Nothing changes unless you opt in by naming a different model.
Co-authored-by: Hermes Agent <noreply@nousresearch.com>
The #32091 fix moved every profile's cron jobs into one shared root store,
but never wired the execution-scoping half it recommended: a job still ran
under whichever profile's ticker picked it up, not its owning profile. So a
job created under `hermes -p donna` could execute with the root profile's
.env / config.yaml / credentials.
- jobs.py: create_job auto-captures the active profile (explicit profile=
override available) and stores it on the job; resolve_profile_home() maps a
profile name to its HERMES_HOME; legacy jobs backfill to 'default'.
- scheduler.py: run_job applies the job's profile via a scoped HERMES_HOME
override (env var + in-process ContextVar) before any .env/config/script
load, restored in finally. tick() routes profile-mismatched jobs to the
single-worker sequential pool so the env mutation can't race.
- cronjob tool threads profile through (NOT exposed in the model schema, to
avoid cross-profile privilege escalation); hermes cron add gains --profile.
E2E verified against a temp HERMES_HOME with a real profile dir: a root-profile
ticker runs a profile='donna' job with HERMES_HOME=donna during execution and
restores the ticker env afterward.
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>
`hermes computer-use install` refused to install on Linux, Windows, and
macOS x86_64 because the pre-install asset probe was hitting the wrong
GitHub endpoint AND duplicating tag-resolution logic the upstream
installer already does correctly.
`_check_cua_driver_asset_for_arch()` queried
`https://api.github.com/repos/trycua/cua/releases/latest`. On trycua/cua:
- cua-driver-rs releases (the binary the installer fetches) are marked
**prerelease** on every cut. GitHub's `/releases/latest` explicitly
skips prereleases.
- The Python package releases (`cua-agent`, `cua-computer`, `cua-train`)
are non-prerelease and end up as the "latest" instead.
Live API check today:
$ curl -sf https://api.github.com/repos/trycua/cua/releases/latest \
| jq '{tag:.tag_name, asset_count: (.assets|length)}'
{ "tag": "agent-v0.8.3", "asset_count": 0 }
The probe sees zero assets, prints "Latest CUA release has no Linux
x86_64 asset", and skips install on every Linux / Windows / macOS-x86_64
host — even though the cua-driver-rs-v0.6.0 release ships 19 binary
assets covering all those platforms.
Filtering `/releases?per_page=N` for the `cua-driver-rs-v*` prefix
fixes the bug, but it duplicates tag-resolution logic the upstream
`_install-rust.sh` already does correctly via `CUA_DRIVER_RS_BAKED_VERSION`
(auto-baked by CD on every release, with a `/releases?per_page=N` API
fallback for dev checkouts). The right answer is to trust that
contract instead of mirroring it in Python where it can drift.
Two paths get the same outcome without the probe:
1. **Fresh install**: run `install.sh` directly. It has the baked
release tag, fetches the right asset, and errors with a clear
message on missing-arch downloads. No preflight needed.
2. **Upgrade path**: `cua_driver_update_check()` (separately added)
shells `cua-driver check-update --json` against the installed
binary, which returns the canonical update answer from the same
source the installer uses.
- `hermes_cli/tools_config.py`: delete `_check_cua_driver_asset_for_arch`
and its two call sites in `install_cua_driver`. Replace with an
inline comment near the top of the module explaining the rationale.
- `tests/hermes_cli/test_install_cua_driver.py`: drop the
`TestCheckCuaDriverAssetForArch` block. Add `TestArchProbeRemoval`
with three regressions:
- `test_probe_function_is_gone` — asserts the deleted helpers stay
deleted.
- `test_fresh_install_does_not_call_github_api` — asserts the
install path doesn't hit GitHub directly from Python anymore.
- `test_upgrade_with_binary_does_not_call_github_api_directly` —
same for the upgrade path.
All 9 `test_install_cua_driver` tests pass.
Reported by @teknium1 while testing on a headed Ubuntu host.
Salvages #50469 by @libre-7.
_dashboard_local_update_managed_externally() previously blocked every containerized dashboard from the local update API, even when the running install was a bind-mounted git checkout that can be updated with hermes update.
Allow the dashboard updater only for git installs inside containers, while keeping hosted /opt/data, docker, and pip installs managed externally. Pip remains blocked because its apply path mutates the running container filesystem and is not the self-managed checkout case.
Adds regression coverage for docker, git, and pip install-method handling inside containers, and maps the contributor email for release attribution.
Co-authored-by: libre-7 <libre-7@users.noreply.github.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.
/simplify-code (LOW, flagged by two reviewers): the source tags 'user' /
'project' / 'bundled' were bare string literals scattered across the discovery
scrub and the two mount-time refuse guards. A typo in any one site (e.g.
'users') would SILENTLY disable a security gate with no error — the exact
failure mode this RCE boundary must not have.
Introduce a shared module-level _NON_BUNDLED_PLUGIN_SOURCES frozenset referenced
by both the discovery scrub and the (now single) mount guard, so the
auto-import policy lives in one place. The two mount guards collapse into one
gate that still emits the distinct per-source operator message via a map (no
loss of guidance). Behavior unchanged: 39 RCE-bypass tests pass, and the
constant is mutation-checked (typo'ing it fails the bypass tests).
Defence-in-depth (discovery scrub + mount refuse) is retained intentionally.
* feat(computer_use): disable cua-driver telemetry by default, add opt-in
cua-driver ships anonymous PostHog usage telemetry ENABLED by default
upstream (fires cua_driver_install / cua_driver_doctor events to
eu.i.posthog.com). Hermes now disables it for our users unless they
explicitly opt in.
- New config key `computer_use.cua_telemetry` (default false) in
DEFAULT_CONFIG.
- `cua_backend.cua_driver_child_env()` injects
`CUA_DRIVER_RS_TELEMETRY_ENABLED=0` into the child env when telemetry is
disabled (the default); leaves the var untouched on opt-in so the driver
uses its own default. Reads config fail-safe — any error defaults to
telemetry off.
- Routed every cua-driver spawn site through the policy: MCP backend
(StdioServerParameters env), `cua_driver_update_check`, doctor's
health_report Popen, the install.sh/install.ps1 runner, and the
`--version` / status probes.
- Docs: new Telemetry subsection in computer-use.md (EN).
- Tests: tests/computer_use/test_cua_telemetry.py — default disables,
explicit-false disables, opt-in leaves var untouched, config-failure
fails safe, inherited-enabled is overridden off.
Verified live on Linux against the real cua-driver-rs 0.6.0 binary: with
the var=0 the driver reports "telemetry: disabled via
CUA_DRIVER_RS_TELEMETRY_ENABLED" and sends no event; with it unset it logs
"sending event: cua_driver_doctor". 213 computer_use + install tests green.
* fix(dashboard): fold computer_use config category into agent tab
The new computer_use.cua_telemetry key created a single-field dashboard
config category, tripping test_no_single_field_categories (web_server's
invariant that categories with <2 fields must be merged to avoid tab
sprawl). Add computer_use -> agent to _CATEGORY_MERGE, matching the
existing onboarding/telegram single-field folds.
The install pre-flight asset probe queried trycua/cua's `releases/latest`,
which floats across the monorepo's components (agent-*, computer-*, lume-*,
train-*) — most ship zero binary assets. So the probe false-negatived and
hard-blocked `install_cua_driver` (line 770: `if not probe: return False`)
BEFORE the upstream installer ran, on Linux, Windows, and Intel macOS — even
though the installer it gates resolves the right tag and would have succeeded.
Net effect: the normal enable path (`hermes tools` → Computer Use post-setup,
and `hermes computer-use install`) refused to install on every platform this
PR claims to support.
Fix: list `/releases?per_page=100`, pick the newest `cua-driver-rs-v*` tag,
and match its assets on OS-token + arch — mirroring what the upstream
`install.sh` already does. Fail open if no driver release surfaces (installer
remains the source of truth). Adds an OS-token gate so a darwin asset can't
satisfy a Linux probe.
Tests: updated the install-probe fixtures to the list-of-releases shape with
`cua-driver-rs-v*` tags + OS-token asset names; added a regression guard
(`test_releases_latest_tag_ignored_picks_driver_rs_tag`) for the monorepo
floating-latest case. 25/25 install + 192 computer_use tests green.
Verified live: probe returns True for all six platform/arch combos against
the real GitHub releases API.
The runtime gate (check_computer_use_requirements) and the hermes tools
platform_gate both enable linux alongside darwin/win32, but several
docstrings/comments still described Linux as "alpha, gated off until it
flips upstream" — contradicting the code that ships it. Bring the prose in
line with the gate that's actually live:
- tool.py / cua_backend.py module docstrings: Linux is enabled (X11 today,
Wayland via XWayland), not gated off.
- toolsets.py description and hermes tools display name: (macOS/Windows) ->
(macOS/Windows/Linux).
No behavior change — the gate already allowed all three platforms.
Make the computer_use toolset platform-agnostic by driving cua-driver on
macOS, Windows, and Linux. Consumes the 8 cua-driver decoupling surfaces
(capability discovery, structuredContent AX tree, opaque element_token,
click button enum, explicit mimeType, machine-readable manifest,
structured list_windows, structured health_report), each degrading
gracefully on older drivers.
Adds `hermes computer-use doctor` (drives cua-driver health_report with a
per-OS check matrix and an exit 0/1/2 ok/degraded/blocked contract), full
typed wrappers for the previously-uncovered cua-driver tools plus a generic
call_tool escape hatch, per-session agent-cursor lifecycle, platform-aware
system-prompt guidance (host-deterministic, cache-safe), and honors
HERMES_CUA_DRIVER_CMD end-to-end.
Replaces the macOS-only skills/apple/macos-computer-use skill with a
cross-platform skills/computer-use skill, and refreshes the EN + zh-Hans
docs.
Supersedes #44221 (Windows-enablement salvage of #30660).
Co-authored-by: Teknium <127238744+teknium1@users.noreply.github.com>
* 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.
OpenCode Go (and OpenCode Zen) showed only a subset of the models they
serve in the desktop/CLI model picker — e.g. opencode-go rendered 13 of
19, silently dropping minimax-m3/m2.7/m2.5, glm-5/5.1, deepseek-v4-flash.
Root cause: the picker dedup in build_models_payload strips any model
from an aggregator row that overlaps a user-defined provider's catalog
(so a local proxy isn't shadowed by OpenRouter). It gated on
is_aggregator(), which is True for opencode-go/zen because their flat
/v1/models returns bare IDs the model-switch resolver searches. But
those are flat-namespace RESELLERS, not routing aggregators — every
model they list is first-party, so deduping them against a user proxy
that happens to serve a same-named model guts their own catalog.
Fix: add is_routing_aggregator() (True only for true routers like
OpenRouter and custom:* proxies; False for opencode-go/zen) and gate the
picker dedup on it. is_aggregator() is unchanged so model-switch flat
catalog resolution keeps working. Both desktop entry points
(model.options JSON-RPC and /api/model/options REST) and hermes model
share build_models_payload, so all surfaces get the full list.
Fixes#47077
The post-update gateway resume path (`_resume_windows_gateways_after_update`)
only relaunched gateways that were *running* when the update began — it
enumerates live PIDs in `_pause_windows_gateways_for_update` and respawns
exactly those. A gateway that had already died between updates (e.g. it was
launched attached to a terminal/TUI that later closed, taking the child with
it) was never brought back: the Startup-folder / Scheduled-Task autostart
entry only fires on the next login, not after an in-place update.
So a Desktop-GUI update (which runs `hermes update --yes --gateway`) on a box
whose gateway had quietly died would complete with no gateway running, and the
user had no indication anything should have come up.
Fix: when no gateway is running at pause time but an autostart entry is
installed (`gateway_windows.is_installed()` — an explicit "I want a gateway"
signal), return a `cold_start_if_installed` token. The resume step then does a
fresh detached spawn via `gateway_windows._spawn_detached()` — the same
windowless `pythonw` + `CREATE_BREAKAWAY_FROM_JOB` path `hermes gateway start`
uses. It re-checks liveness immediately before spawning so a concurrent start
(autostart entry firing) can't produce a duplicate.
Gateway-less users (no autostart entry) get nothing forced on them — the
pause step still returns None for them. POSIX is unaffected: enabled systemd
units already restart via `Restart=always`.
Windows-only; best-effort throughout (logs at debug and no-ops on any error).
Tests: pause returns the cold-start token only when installed, returns None
when not installed, resume cold-starts on the token, and resume skips the
cold-start when a gateway is already running.
* chore: re-trigger CI (workflows did not dispatch on prior head)
* fix(update): don't count across shallow-clone boundary (bogus '12492 commits behind')
Installer checkouts are shallow (git clone --depth 1). The CLI banner and
hermes update --check both did a plain git fetch (silently unshallowing the
repo) then git rev-list --count HEAD..origin/main, which counts across the
shallow boundary and prints a huge nonsense number like '12492 commits behind'.
Detect shallow up front, fetch with --depth 1 to preserve the boundary, and
compare tip SHAs instead of counting:
- banner _check_via_local_git: returns UPDATE_AVAILABLE_NO_COUNT when behind
(renders as 'update available') instead of the bogus count.
- _cmd_update_check: reports presence-only on shallow clones.
Full clones keep the exact count path unchanged. Mirrors the desktop fix in
apps/desktop/electron/main.cjs (commit 2950c6fa2).
Defense-in-depth for the dashboard plugin auto-import path. The web server
auto-imports and mounts the Python backend (dashboard/manifest.json -> api file)
of plugins found in ~/.hermes/plugins/ (user) and ./.hermes/plugins/ (project),
not just bundled plugins. So any plugin that reaches one of those dirs gets
arbitrary Python executed on the next dashboard start.
NOTE ON THREAT MODEL: #43719's originally-documented delivery chain (a public
--insecure dashboard + open API used to git clone a malicious repo into
~/.hermes/plugins/) is ALREADY mitigated on main — since the June 2026
hermes-0day hardening, a non-loopback bind ALWAYS requires an auth provider and
--insecure no longer bypasses the auth gate. This change is therefore NOT
closing that (now-authenticated) network path; it removes the residual
'arbitrary code executes merely because a plugin is on disk' hazard, which still
applies when a plugin arrives by other means: a socially-engineered git clone,
a supply-chain drop, an authenticated-but-malicious actor, or a future
regression in the auth gate. Untrusted on-disk code should not auto-execute.
Restrict dashboard backend Python auto-import to BUNDLED plugins only. User and
project plugins may still extend the dashboard UI via static JS/CSS, but their
api Python file is never auto-imported. Two layers: _discover_dashboard_plugins
scrubs api/_api_file for user/project sources (and bundled wins name conflicts
so a non-bundled plugin cannot shadow a trusted backend route);
_mount_plugin_api_routes re-refuses user/project at mount time. Tightens the
prior GHSA-5qr3-c538-wm9j / #29156 hardening (bundled+user) to bundled-only.
Salvaged from #44472 (@egilewski) onto current main.
A cron job that sets `enabled_toolsets` to a list of *native* toolsets (e.g.
`["web", "terminal"]`) silently got ZERO MCP tools, while a job with no
per-job list got every globally-enabled MCP server. `_resolve_cron_enabled_
toolsets` returned the per-job list verbatim, bypassing the MCP-merge that the
platform-fallback branch performs via `_get_platform_tools`. So
`discover_mcp_tools()` registered the MCP tools into the registry, but
`get_tool_definitions(enabled_toolsets=...)` kept only the named native
toolsets — the agent then rejected every `mcp_*` call as "Unknown tool". (R2
of #23997.)
Fix: `_merge_mcp_into_per_job_toolsets` layers MCP membership onto a per-job
allowlist with the SAME semantics as `_get_platform_tools`:
* `no_mcp` sentinel present -> no MCP servers (sentinel stripped)
* one or more MCP server names already listed -> treat as an allowlist
* otherwise -> union in every globally-enabled MCP server
To avoid duplicating the "which MCP servers are enabled" computation (it
already existed inline in `_get_platform_tools`), this extracts a shared
`enabled_mcp_server_names(config)` helper in `hermes_cli.tools_config` and has
BOTH the gateway/CLI platform resolver and the cron per-job resolver call it —
so every path agrees on MCP membership (extend, don't duplicate).
Note: the issue's *headline* — bare MCP server names rejected, registry never
includes them — was already fixed on main (commits c10fea8d2 + 04918345e,
both before the issue was filed). This PR closes the remaining cron-specific
gap (R2). The `server:*` / `mcp:server` alias-notation rejection (R1) and the
quiet-mode silent-drop (R3) are tracked separately.
Salvaged from #32788 by sherman-yang (credited below). Reworked to reuse the
shared `enabled_mcp_server_names` helper instead of re-implementing the MCP
membership set in cron/scheduler.py.
Fixes#23997
Co-authored-by: sherman-yang <58446328+sherman-yang@users.noreply.github.com>
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.
Authorization to message the agent is the gate, not the file extension.
Previously the inbound-attachment allowlist (SUPPORTED_DOCUMENT_TYPES) was
opt-OUT on Discord (allow_any_attachment defaulted false) and had no bypass
at all on Telegram/Slack — so an .html (or any non-allowlisted type) was
dropped or hard-rejected before the agent saw it.
Now every authorized upload is cached and surfaced to the agent regardless
of type:
- base.cache_media_bytes(): unknown types cache as octet-stream (or the
caller-supplied MIME) instead of returning None — fixes the chokepoint
that Teams/Telegram-media route through.
- discord/telegram/slack adapters: removed the allowlist reject/skip; any
non-media attachment is typed DOCUMENT and cached. Known types keep their
precise MIME.
- Text inlining now gates on a shared _TEXT_INJECT_EXTENSIONS set (text +
code + config + markup) instead of a blind UTF-8 decode, so binary formats
(PDF/zip/docx) with ASCII headers are never inlined.
- gateway/run.py emits the path-pointing context note for every DOCUMENT,
including non text/application MIME types.
- discord.allow_any_attachment is now a documented no-op kept for config
back-compat.
Validation: 357 gateway tests pass; E2E confirms .html/.bin/custom types
cache, known types stay precise, PDFs are not inlined.
