Follow-up to the salvaged #93517. Two gaps found in review:
- After the last unregister the supervisor stayed resident for the life of
the process (~15 MB + a pipe) in any gateway/cron that ever connected a
stdio server; main's per-server watchdog exited with its server. Close
our write end when the supervised set empties: EOF with nothing
registered makes the supervisor exit without reaping, and the next
register already respawns and replays.
- A child that raced and exited before os.getpgid dropped its group from
coverage entirely. The SDK spawns stdio servers as session leaders
(pgid == pid), so fall back to the pid instead; the prune forgets the
group once nothing in it is alive.
Tests: fake-supervisor release/re-spawn sequence, and a real-process EOF
release (supervisor exits 0, unregistered child untouched). Mutation
checked: disabling the release branch fails both.
Review raised a real gap: we reap by pgid, so a registration is only as
meaningful as the group's identity. A group we deliberately keep registered --
an orphan teardown failed to kill, such as the `node` mcp-remote leaves behind
-- can later exit on its own, after which the kernel may hand that pgid to an
unrelated process owned by the same user. An ungraceful Hermes death while the
registration is stale would then signal a stranger. `_is_safe_target` cannot
catch it, because the value is stale rather than invalid.
Prune registrations whose group has no members left on every registration
change, and tell the supervisor to forget them. Signal 0 is a pure existence
question -- it cannot terminate anything -- so this is cheap and safe to run on
the hot path. An ambiguous answer (EPERM: exists but not ours) keeps the
registration, since dropping real coverage is the more expensive mistake.
This narrows the window rather than closing it: a group can still die and its
pgid be recycled between two probes. Closing it completely means proving group
identity at reap time, e.g. stamping children with a boot-unique env marker and
checking a member still carries it. That was judged not worth putting a `ps`
parse into the one process whose job is to stay simple enough to always work,
so the residual is now documented in the module docstring instead, along with
the note that Hermes's existing killpg-based orphan cleanup already carries the
same exposure (upstream #88350).
Also records why supervisor recovery is deliberately two-step: a failed write
drops the handle, and the next call rebuilds coverage from `_supervised_pgids`,
which -- not the pipe -- is the record of what needs reaping.
The protocol tests register synthetic pgids that were never real process
groups, so they now state that precondition through an `all_groups_alive`
fixture instead of depending on pid-space luck.
29 tests pass. Verified the change adds no failures: same selection, my changes
stashed vs applied, 42 failed either way (pre-existing in a locally rebuilt
venv) with passed going 617 -> 619 for the two new tests. Footgun linter clean.
(cherry picked from commit 3a7d219620a8cb57a7fe2c2346ce3f055320e743)
Every stdio MCP server was wrapped in its own CPython watchdog that polled
getppid() every 2s to notice an ungraceful Hermes exit (kill -9, OOM, crash,
force-quit), since macOS has no PR_SET_PDEATHSIG. That is a whole interpreter
per server for a job that does nothing until the moment Hermes dies: 10.1 MB
physical footprint each, measured on macOS/arm64.
Replace the fleet of pollers with a single supervisor per Hermes process
holding the read end of a pipe only Hermes writes to. Death detection becomes
EOF on that pipe -- exact and instant, rather than up to a poll interval late.
Hermes sends `register <pgid>` / `unregister <pgid>` as servers come and go;
on EOF the supervisor killpg's whatever is still registered, which is exactly
the set whose teardown never ran. A clean shutdown unregisters as it goes, so
EOF then finds nothing to kill.
Servers are now spawned unwrapped. The MCP SDK already starts each stdio child
in its own session, so the pgid recorded for killpg is the server's own group
and the existing cleanup paths reach it unchanged. That also deletes the
signal-forwarding layer the wrapper needed: wrapping had put the real server in
a different session from the pgid being tracked, so a graceful killpg would
have hit only the wrapper.
Measured on a 5-gateway host: 10 watchdogs (~98 MB) -> 5 supervisors (~49 MB).
One supervisor costs about what one watchdog did (9.9 vs 10.1 MB), so the win
is (servers_per_process - 1) x ~10 MB, and a process with no stdio servers now
spawns nothing at all.
The supervisor reads length-capped lines rather than iterating the stream: a
writer that never sends a newline would otherwise grow it without bound, which
it must not be vulnerable to when it is the last defense against leaked
servers. Found by feeding it /dev/zero, where it reached 15 GB.
Verified beyond unit coverage: a real stdio MCP server connects and its tools
are discovered on the unwrapped path; with a live Hermes holding a real
connection, kill -9 reaped the server, its grandchild (in the server's group,
the mcp-remote `node` case), and the supervisor exited on its own. The reap
tests were sabotage-checked in both directions -- a no-op reaper fails all
three, while the test pinning that a cleanly unregistered server survives keeps
passing -- and each wiring half fails independently when removed.
(cherry picked from commit a252d4ce7ff1722f687635fdbf0cff79f538c3f1)
tests/gateway/test_feishu_bot_admission.py builds FeishuAdapter via
object.__new__ and installs dedup state by hand (install_dedup_state), so
the asyncio.Lock added in __init__ was missing -> AttributeError in CI.
Resolve the lock through a getattr-with-create helper and add it to the
shared test helper.
Offloading the fsync to a worker thread (previous commit) lets two inbound
messages have two persists in flight. Without ordering, a slow first flush
lands after the second and the disk file loses the newer token/id/seen-id
(in-memory state stayed correct, so the loss only surfaced after restart).
The loop-side dict could also be mutated while the worker iterated it.
- weixin ContextTokenStore / discord non-conversational tracker: snapshot
the payload on the loop, hand it to the worker, serialize with a per-store
asyncio.Lock.
- feishu dedup: same asyncio.Lock around the offloaded persist (its snapshot
was already taken under _dedup_lock; the write was not).
- Regression tests: two concurrent persists with a slow first write must
leave the union on disk. All three fail on the previous commit.
atomic_json_write() calls os.fsync(), which blocks until the write
reaches stable storage. #83906 offloaded this in channel_directory.py,
gateway/run.py, and gateway/slash_commands.py, but the same blocking
pattern was still present on three platform-adapter message hot paths
that were never touched by that pass:
- plugins/platforms/feishu/adapter.py: _is_duplicate() persists the
dedup cache on every new inbound message (shared by both websocket
and webhook transports). The in-memory bookkeeping stays under
_dedup_lock; the fsync now runs off it via asyncio.to_thread so
concurrent callers on the ws-thread loop aren't blocked on disk I/O.
- gateway/platforms/weixin.py: ContextTokenStore.set() persists the
context-token cache on every inbound message carrying a
context_token, called from the async poll-loop handler
_process_message.
- plugins/platforms/discord/adapter.py: mark_many() persists the
non-conversational message-id tracker, called from both send() and
send_update_prompt() when a message is tagged nonconversational.
Each fix mirrors the existing asyncio.to_thread(atomic_json_write, ...)
pattern from #83906. Regression tests assert the persist call runs on
a different thread than the event loop, mirroring
test_directory_write_runs_off_event_loop_thread in
test_channel_directory.py.
Follow-up to the cherry-picked fix:
- decide force_remove by inspect.signature (the idiom cleanup_vm already
uses) instead of catching TypeError, which would misread a TypeError
raised inside cleanup() as a signature mismatch and call cleanup twice
- skip teardown for ssh: SSHEnvironment has no task-scoped sandbox and its
cleanup() runs sync_back + 'ssh -O exit' on a ControlMaster socket keyed
by user@host:port, shared with the agent's real environment
`_probe_remote_backend` spins up an environment with
`task_id="prompt-backend-probe"` purely to run one `uname` for the system
prompt, then drops the reference without cleaning up. Container backends
default to `container_persistent` / `docker_persist_across_processes`, so
that throwaway sandbox stays up for the entire process lifetime, sitting
idle next to the agent's own `default` sandbox.
Observed on a three-profile Docker-backend host: six running containers
where three were doing any work. Every profile permanently paid for one
extra container whose only job had been a single `uname` at startup.
Fix: wrap the probe in `try/finally` and call
`env.cleanup(force_remove=True)` — the explicit-teardown path that
`DockerEnvironment.cleanup` already documents ("No current caller passes
`force_remove=True`; the parameter is here so the explicit-teardown
semantics can be wired up later"). Backends that inherit the kwarg-less
base `cleanup()` fall back to the bare call.
Verified on a live aarch64 Docker backend: the probe container is now
created, used and reclaimed within the same call. Three regression tests
cover the success path, the exception path, and the kwarg-less fallback;
all three fail without the source change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PLbMzWqHVTPNBFisM8UiNB
Findings from the efficiency review pass on the two salvages, applied as one
small follow-up:
- copilot_auth: check the negative cache BEFORE taking the per-fingerprint
exchange lock. During the 60 s post-failure window, dashboard polls now
raise immediately instead of parking an executor thread behind the
in-flight holder (up to ~50 s) to learn the same answer. Test hangs
without the check (timeout 124), passes with it.
- buzz _localize_inbound_media: download_path.read_bytes() was still
evaluated on the loop as the argument to the offloaded cache call — up to
the 128 MiB inbound cap. Read off the loop too.
- test_list_credential_pool_keeps_loop_responsive: 0.5 s block / 0.25 s
threshold (2x margin) so runner descheduling cannot false-fail it while a
real regression still trips it.
`hermes_home_key` called `Path.resolve()` on every call, which is a
filesystem call. `ToolRegistry.current_scope_key()` calls it, and
`_merged_tools()` calls that, so every single registry lookup went to
the disk: `get_entry`, `get_toolset_for_tool`, `get_emoji`, and the rest.
Anything that looks up many tools in a row paid for it. Building the
deferred-tool catalog for `tool_search` calls `get_entry` once per tool,
so with 400 deferrable tools one search did 800 path resolutions. On
Windows that is `nt._getfinalpathname`, and it took 132ms of the 148ms
the catalog build spent. The tokenizing everyone would suspect first was
under 4ms of it.
The resolved value is now remembered per input path. A process only sees
a handful of home paths, so the dict stays tiny. Measured on the same
400-tool catalog build, the lookups drop from 132ms to 3.1ms with one
entry cached.
A path that does not exist yet is resolved but not remembered, because
the answer can change once it is created, for instance when part of the
path turns out to be a link. A different home path is a different key,
so switching profiles or `HERMES_HOME` gets its own entry rather than
reusing a stale one.
A busy or briefly unreadable cache file must not disable disk loads for
the rest of the process; only missing or permanently malformed files
should mark the cache as loaded.
The OSV preflight is a synchronous network POST to api.osv.dev (up to 10s timeout) on every MCP stdio server start. Tools like `hermes mcp test` and MCP reconnect ladders spawn the same package repeatedly, but the previous in-process cache was empty after every process restart, so each run re-queried OSV and added 5.91x variance to the connection-time span.
Persist the malware-check verdict cache to `<hermes_home>/cache/osv_check.json`. Cache expiry is stored as an absolute wall-clock timestamp, so it survives restarts and monotonic-clock skew. Loading only adds missing keys so an in-memory overwrite (e.g. a test forcing expiry) is not silently reversed by the disk copy. Writes are atomic (temp file + rename) and happen under the existing cache lock.
Fixes the Hermes-owned OSV preflight variance component of #68416. Server-side `initialize` time is outside Hermes' control.
- Adds `hermes_constants.get_hermes_home()` lazy import to keep `tools/osv_check.py` import-safe (stdlib + typing only at module scope).
- Switches cache timestamps from `time.monotonic()` to `time.time()` for persistence compatibility.
- Updates `tests/tools/test_osv_check.py` fixture to isolate disk cache per test via `HERMES_HOME` + `tmp_path`, and adds regression tests for persistence, reload, and disk format.
The background-process poller for non-local backends ran `cat` on the
whole log file every two seconds, then threw away everything except the
part it had not seen yet. The offset it needed was already tracked one
line below, so the full read was pure waste.
Cost of one poll grew with the total output so far, which makes the cost
of a run grow with the square of its length. A job writing 10 MB over an
hour moved about 9 GB across the docker or SSH channel to deliver 10 MB
of output.
The poller now asks the shell for the file size and the bytes after the
offset in one command. Reading the size first and cutting the tail at
that same size keeps the two in step, so a file that grows mid-command
never sends a byte twice. A file that shrank was rotated or truncated,
so the offset drops back to 0 and the buffer is dropped.
The output buffer is now appended to rather than replaced, matching the
local reader loops, and the offset is counted in bytes because the shell
counts bytes.
The previous commit removed cache_media_bytes from teams/adapter.py's import
block but left one bare call at the Bot-Framework image path — a NameError
that the surrounding 'except Exception' swallowed, so every BF image
attachment was silently dropped (ruff F821). Route it through
cache_media_bytes_async like the file's two other sites.
Same bug class at the sites the sweep did not reach:
- buzz _download_attachment / _localize_inbound_media: cache_media_bytes
called from async methods → cache_media_bytes_async.
- photon _dispatch_inbound: _normalize_binary_payload (base64 decode +
cache write of possibly multi-MB payloads) ran on the loop → closure made
async, helper offloaded via asyncio.to_thread.
Tests: cover cache_media_bytes_async (thread-id + kwarg forwarding);
teams/buzz tests re-pointed at the async seam.
Follow-up to the off-loop move: once the credential-pool handlers run on
worker threads, the dashboard's periodic /api/credentials/pool polls can
overlap, and during a DNS outage each poll would have started its own
exchange and abandoned its own hung resolver thread.
- Per-fingerprint threading.Lock around the exchange: concurrent callers
wait on the one in-flight attempt, then hit the positive or negative
cache (bounded worker count, no duplicate network calls).
- _urlopen_bounded: when the hard cap fires and the abandoned worker later
succeeds, close the HTTPResponse instead of leaking the socket.
- Tests (none shipped with the original PR): hard cap + late-close,
single-flight success and failure paths, and the pool endpoint running
off-loop / keeping the loop responsive under a 200 ms blocking read.
The module already binds run_in_threadpool (used by list_profiles_endpoint)
and every sibling router uses the same starlette helper; the nine new
loop.run_in_executor(None, _run) sites now go through that alias so the
file has one offload idiom. Behaviour-identical (both hand the callable to
a worker thread).
Also sweeps the one endpoint the PR left synchronous:
update_profile_model_endpoint's _write_profile_model reads and rewrites
the profile's config.yaml on the event loop.
Two assertions per offloaded site:
- a loop probe, where the stubbed callee records whether an event loop is
running in its own thread — the idiom already used by
tests/hermes_cli/test_cron_dashboard_off_loop.py; and
- a concurrency proof, where the stubbed callee blocks on a threading.Event
while an unrelated request is timed. On the unfixed handlers that request
waits out the whole block; served off the loop it returns in
milliseconds.
The concurrency proof needs a single event loop across requests, so the
client fixture enters the TestClient context manager: that pins one
blocking portal for the whole fixture, where a bare TestClient(app) would
spin up a fresh loop per request and pass even unfixed.
Also covers the status-code mapping through the executor hop (404 on a
missing profile, 400 on a rename collision, 404 from the resolve that stays
on the loop ahead of describe-auto) and the _MISSING sentinel cases: a
desktop.json holding `null` still reports exists=true, an absent one
reports exists=false, and an empty SOUL.md is still distinguishable from a
missing one.
The client fixtures read web_server._SESSION_TOKEN from the module rather
than pinning a literal. web_server resolves that token once at import, so
whichever test file imports it first fixes the value for the session and a
later monkeypatch.setenv is silently ignored — two files hardcoding
different tokens would 401 depending on collection order.
Extends the off-loop suite to the third and last blocking method on the
`UpstreamAdapter` contract, the 401/429 rotation.
As with the two existing pairs, the primary assertion is **thread identity**,
not latency: a latency assertion measured by an HTTP client on the blocked
loop is vacuous, because the client's own timer cannot advance until the
block ends and it therefore reports a fast response on provably frozen code.
* `test_get_retry_credential_runs_off_the_event_loop` records
`threading.get_ident()` inside the fake adapter and compares it to the
loop thread, and checks the rotation still works end to end (rejected
bearer forwarded first, rotated bearer second).
* `test_event_loop_keeps_running_while_the_retry_credential_resolves`
samples a loop-side heartbeat counter from inside the stalled adapter. On
the unfixed handler it records exactly 0 loop iterations across a 0.5s
rotation.
* `test_retry_credential_failure_still_returns_the_upstream_rejection`
guards the error contract the change must leave alone: a raising rotation
is still swallowed and the upstream's own 401 is streamed back, with no
second forward.
A new `_build_rejecting_upstream` harness drives the `status in {401, 429}`
branch by rejecting every bearer except the rotated one.
Extends `test_proxy_off_loop.py` with the `/health` half, using the same
two-assertion shape as the credential tests:
- `test_is_authenticated_runs_off_the_event_loop` compares the thread the
adapter's `is_authenticated` ran on against the loop thread. Before the
fix they are the same ident.
- `test_event_loop_keeps_running_while_health_resolves_auth_state` reads a
loop-side heartbeat counter sampled by the adapter across its own stall.
Before the fix exactly 0 iterations run across 0.5s.
Both also assert the response is unchanged (`200`, `authenticated: true`),
so the offload cannot quietly alter what `/health` reports.
Adds `tests/hermes_cli/test_proxy_off_loop.py`, mirroring the harness in
`test_proxy.py`: the proxy and a fake upstream run as real aiohttp
servers on ephemeral ports under a single `asyncio.run`, guarded by
`pytest.importorskip("aiohttp")` — no pytest-aiohttp dependency.
The primary assertion is thread identity, not latency. A latency
assertion measured with an HTTP client on the blocked loop is vacuous:
the client's own timer cannot advance until the block ends, so it reports
a fast response on code that was provably frozen.
- `test_get_credential_runs_off_the_event_loop` records
`threading.get_ident()` inside the adapter and compares it to the loop
thread. Before the fix both are the same ident.
- `test_event_loop_keeps_running_while_credentials_resolve` runs a
heartbeat task on the loop and has the adapter sample its counter on
entry and exit, so the reading is taken from the loop rather than
through a client that shares it. Before the fix exactly 0 iterations
run across a 0.5s stall; after it, ~50.
- `test_credential_failure_still_maps_to_401` pins the error contract
across the change of call form. It is deliberately not in the
red-before set — it guards behaviour the fix must leave alone.
Submodule and worktree checkouts store .git as a file (gitdir: pointer);
the carry-over only looked at directory names, so that form was still lost
on rollback. Handle files with the same guard. Docstring now states the
deliberate limit: an excluded entry whose skill dir the target snapshot
lacks is dropped with staging (no orphan .git) and is not undoable via the
safety snapshot, which excludes these paths as well.
Excluding nested .git from snapshots has a side effect on rollback: the
staging move takes the whole live skill dir (including its .git) into
.rollback-staging-*, the extract restores the snapshot without it, and the
staging dir is then deleted — so a skill that is itself a git checkout lost
its .git on any rollback. Reproduced: main preserves it, the exclusion-only
branch did not.
After a successful extract, move excluded subtrees from the staged copy back
under their restored skill dir (mirroring how a top-level .git survives by
never being staged). Regression test included.
_safe_skills_path() is the sibling of iter_skills_files(): when a symlink in
skills/ forces a sanitized copy for mount-based backends (Docker/Singularity),
it rglob-copied the whole tree — .hub, .curator_backups, node_modules and all.
Prune EXCLUDED_SKILL_DIRS before descending, same rule as the sync generator,
so the mounted copy never carries (or walks) the bookkeeping trees either.
iter_skills_files() walked the skills tree with a bare rglob("*"), so the
.hub download cache, .archive, curator backups, and any node_modules/.git
under a skill package were uploaded to the sandbox on every sync. The
sandbox never reads them: skill content is resolved host-side.
EXCLUDED_SKILL_DIRS is already the canonical exclusion set, honoured by
discovery and backup. Apply it to the sync path too, across all three
roots iter_skills_files() walks (local, external, project-local), and add
.curator_backups to the set.
Measured on a local install: 900 files / 67.3 MB -> 771 files / 8.4 MB.
This is not just wasted bandwidth on the SSH backend, where the oversized
payload can exceed the 120s _ssh_bulk_upload deadline and surface as the
agent hanging on every tool call.
The filter intentionally does not reuse is_excluded_skill_path(), which
also prunes references/, templates/, assets/ and scripts/ -- those hold
support files and bundled scripts the sandbox does read and execute.
f50b5bb0fa taught the sync retry site to keep the cheap same-provider retry
when a Codex stream dies inside the 60s no-progress window (zero output),
skipping straight to fallback only on a stall or hard-ceiling timeout. The
async site never got that carve-out, so after widening the skip to vision
(#97572) an async vision call on a stillborn stream would have jumped to
fallback where the sync path retries. Both sites now apply the same rule.
Adds the async twin of the vision-skip test and a no-progress-still-retries
guard for the async site.
Issue #54465 established that a same-provider retry after a full-budget
timeout costs a second whole `timeout` window before the fallback chain is
reached, doubling the user-visible stall, and that compression must not pay
it because it sits on a critical path. The guard added for that is spelled
`task == "compression"`, so vision — which sits on the interactive path —
still retries.
The cost is the same and the stall is more visible: the turn holding the
image cannot answer, and because turns are serialised the following user
messages queue behind it. Two sequential full-budget timeouts on an
unhealthy vision provider is a long stall for something the fallback chain
could have served immediately.
Replaces the string comparison at both retry sites (sync `call_llm` and
`async_call_llm`) with `_TIMEOUT_NO_RETRY_TASKS = {"compression", "vision"}`,
so the two paths cannot drift again. Behaviour is unchanged for every other
task: fast blips (a streaming-close or a 5xx) still retry, and only
full-budget timeouts on those two tasks skip straight to fallback.
Tests: vision now falls straight through to fallback with the primary tried
exactly once, and a non-critical task still gets its one same-provider
retry, so the change stays scoped. Reverting the source change fails the
vision test and leaves the scoping test green.
Not the same as #51513, which fixes five separate defects in the vision
fallback chain (capability detection, sync/async client misuse, geo-block
and RemoteProtocolError classification, and chain iteration). This is about
what happens before that chain is reached.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
GitHub answers anonymous fetches with HTTP 401 during outages (and for
renamed/private repos). git then prompts `Username for 'https://github.com':`
on the inherited terminal and `hermes update` sits there — users read it as
Hermes demanding a GitHub login.
Every network git call in the updater (fetch/pull/push, apply + --check +
fork sync) now runs with GIT_TERMINAL_PROMPT=0 / stdin=DEVNULL, so the 401
fails fast into the fetch-failure classifier, which now reports it as a
GitHub-side rejection (likely outage) rather than blaming the user's
credentials. Credential helpers/askpass are left configured so private-fork
origins still authenticate.
Live repro: PTY-attached update --check against a 401 origin hung 15s+ on
the prompt before; exits rc=1 in 0.2s with the diagnosis after.
Same class as #73751 (@Frowtek, pre-main.py decomposition); passive banner
half salvaged from #101421 (@RobbertC5).
A goal quality gate is a shell command persisted by `/goal gate add` and
later executed with `subprocess.run(shell=True)` at every goal turn
boundary (run_gate), with no approval prompt. On the messaging gateway,
slash access is backward-compatible: with no `allow_admin_from` list
configured (the default), every *allowed* chat user is treated as
unrestricted. So an allowed but non-admin remote sender could add an
arbitrary shell command and get authenticated RCE as the Hermes process
account.
Gate ONLY the shell-creating operation (`gate add`) behind the existing
fail-closed explicit-admin check (`_resume_caller_is_admin`, the same one
that guards cross-origin `/resume`). `gate list` / `remove` / `clear`
stay open so a non-admin can still inspect and recover. CLI/TUI/Desktop
`gate add` is local (the user already has host access) and is unchanged.
Salvaged from #91677 by @unsupportedpastels — narrowed to the single
shell-creating op and reusing the existing admin helper rather than
renaming it. Contributor's regression tests preserved.
Co-authored-by: unsupportedpastels <theoldwizard123@pm.me>
On the default Nous config, call_llm acquires the auxiliary client via
_get_cached_client(resolved_model=None), so the cache key's model
element is "". On a 401, _refresh_nous_auxiliary_client rebuilt the
client but keyed the new entry on the resolved wire model (final_model,
e.g. "Hermes-4-405B"). The fresh client therefore landed under a
different key than the lookup, and the stale expired-credential client
under "" was never overwritten: every auxiliary call kept hitting the
dead client, 401ing and forcing a credential portal round-trip on each
request instead of self-healing after the first refresh.
The auto-provider dimensions had the same divergence: call_llm and
async_call_llm dropped task at both acquisition sites, and the async
path additionally dropped main_runtime at acquisition and at both of
its refresh sites, so the refreshed client shadowed the stale one under
a divergent (provider, task, model, runtime) key.
Pass the original lookup model (which may be None) into the refresh as
a separate lookup_model argument used only to build the cache key,
while the resolved model is still stored as the entry's usable model
and returned to the caller. Thread task into both acquisition sites and
main_runtime into the async acquisition and both async refresh sites,
so sync and async compute the same cache key on acquire and on refresh.
The stale client is now overwritten in place instead of lingering under
an orphaned key, preserving the per-model cache keying introduced in
on the default config.
Add end-to-end regression tests that drive the real call_llm and
async_call_llm through the real client cache; the existing 401 tests
patch _get_cached_client wholesale and so cannot observe
acquire/refresh key divergence.
Hermes gathers workspace context by running git against the session
directory automatically — the coding-workspace snapshot, gateway
project-tree build, /diff, @diff|@staged context refs, goal-gate
fingerprint, and -w startup worktree add — before any prompt, tool call,
approval, or trust gate. Those probes ran the system git without
stripping the repository's own config, so a repo delivered as files with
its .git directory intact (a shared zip, sync folder, or USB stick;
git clone never transfers .git/config) could set an execution-sink git
setting and get arbitrary host code execution as the user with nothing
on screen.
- core.fsmonitor / core.hooksPath / pager / editor / credential helper:
neutralized by routing every automatic probe through
noninteractive_git_env(), which pins those keys to inert values via
GIT_CONFIG_* and ignores global/system config. bounded_git_probe (the
reported sink, coding_context._git + tui_gateway.git_probe) now defaults
to that env; worktree-add, working_diff, web_git, context_references,
goals, and subagent_worktree route through it too.
- Attribute-scoped [diff "x"] command=/textconv= drivers: the attacker
names the driver in .gitattributes, so GIT_CONFIG_KEY overrides can't
enumerate them. Added harden_git_argv(), which inserts
--no-ext-diff --no-textconv on diff-rendering subcommands (diff/show/
log/blame) only — status et al reject the flags. Both flags required
(verified empirically; each alone leaves the other live).
Builds on the noninteractive_git_env config-scrubbing from the
gemini-cli #28792 port. Real-git E2E regression suite arms a malicious
repo and asserts every automatic path neutralizes fsmonitor, hooks,
external-diff, and textconv; a baseline test proves the repo is armed.
Addresses review feedback on #84928.
The tick interval was exposed as HERMES_NOUS_KEEPALIVE_INTERVAL_SECONDS.
AGENTS.md reserves .env for credentials and puts behavioural thresholds in
config.yaml, so the knob moves to `nous.keepalive_interval_seconds`,
following the existing `vertex:` section's precedent for non-secret
provider settings. The env var is dropped rather than bridged: it was never
released, so nothing depends on it. Adding a key to a new section is handled
by the deep-merge, so no _config_version bump is required.
test_keepalive_interval_fits_inside_the_token_lifetime asserted
`900 < 899 * 4 - 120`, which is true for any realistic interval and could
never fail. It also tested the wrong value: the configured constant is only
a ceiling, while the schedule that ships is the derived tick. Replaced with
an assertion over the derived tick for each observed lifetime, which does
fail if the derivation constants regress -- verified against both
TICKS_PER_LIFETIME=1 and MIN_INTERVAL_SECONDS=5000.
Also adds coverage for an unreadable config.yaml, which must fall back to
the module default rather than take the keepalive thread down.
pytest tests/hermes_cli/test_nous_auth_keepalive.py -> 9 passed
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Follow-up to the interval fix: ticking faster narrowed the gap but did not
close it, and the hardcoded interval was wrong for short-lifetime accounts.
Two independent problems, both needed:
1. Lifetime is not a constant. Installs have been observed issuing ~3594s
and ~899s (see #35752, which reports expires_in: 899 while this account
reports 3594). Any hardcoded interval is right for one and wrong for the
other. The tick now derives from the lifetime the server actually issued,
capped by the configured interval and floored at 60s. The access token and
the invoke agent key carry separate lifetimes, so the shorter one governs.
2. Ticking faster alone never closes the gap. The refresh only fires once a
credential is within ACCESS_TOKEN_REFRESH_SKEW_SECONDS (120s) of expiry,
so a tick spaced wider than that window steps straight over it. With a 900s
tick against a 3594s lifetime the last tick before expiry still saw 894s
remaining, declined to refresh, and the credential died 6s before the next
one. The keepalive now asks "will this outlive my next tick?" (tick + skew)
rather than reusing the request path's bare skew.
Measured against both observed lifetimes:
lifetime 3594s: was never refreshed proactively; now refreshes 900s early
lifetime 899s: was never refreshed proactively; now refreshes 227s early
The lifetime is re-read every pass rather than cached, since it can change
when the account, plan, or server-side policy does -- exactly the case the
keepalive exists to cover.
min_access_ttl_seconds is threaded through as an optional parameter, so the
request path keeps its existing 120s behaviour and only the keepalive widens
the window.
Nous Portal access tokens carry a one-hour lifetime, and the keepalive only
refreshes once a token is within ACCESS_TOKEN_REFRESH_SKEW_SECONDS (120s) of
expiry. The tick interval was 6 hours, so it could only land inside that
2-minute window by coincidence. In practice every hour rolled over untouched
and the next inference call paid a 401 plus a re-auth round trip.
Observed on a production install: 71 "refreshed Nous runtime credentials
after 401, retrying" events across current logs.
Drop the default interval to 15 minutes, which gives four ticks per token
lifetime and leaves ample margin under the TTL-minus-skew ceiling of 3480s.
Add HERMES_NOUS_KEEPALIVE_INTERVAL_SECONDS so the interval is tunable without
a source edit, matching the existing HERMES_NOUS_TIMEOUT_SECONDS convention.
Zero still disables the thread.
Callers pass no interval, so the default is what actually shipped; the
signature now resolves at call time rather than binding the constant at
import.
Interleaved subagent fan-outs were tagged with the first 4 hex chars of the
delegation id ([b2ac 3/9]), which is attributable but unreadable. Batches are
now numbered in order of appearance per process: [set 1 · 3/9], [set 2 · 1/7].
Desktop /agents already labels groups "Delegation N", so its duplicate hex
badge is dropped.
Assert the invariant — every display read of one session returns the same
transcript — plus the two things that must not grow with it: the model-fed
projection stays compressed, and Undo/Rewind rows stay hidden. Six of these
fail on the unfixed read paths.
Cuts the 41 contributor tests down to 8 pinning the before/after contracts
(out-of-tree provider resolves end to end, copilot-acp unchanged, broken
plugin falls through, flat-install discovery + non-provider kinds untouched).
Adds the create_client hook and process_* fields to the model-provider
plugin developer guide.
`hermes plugins install owner/repo` — and the plugin index behind
`hermes plugins search` — clones into `$HERMES_HOME/plugins/<name>/`, flat, one
directory per plugin. Provider discovery only ever scanned
`$HERMES_HOME/plugins/model-providers/<name>/`.
Nothing joined the two. `PluginManager` does not close the gap either: it
classifies `kind: model-provider` and deliberately skips importing it, because
provider lifecycle is owned by `providers/__init__.py` — which was not looking
in the directory the installer writes to.
So the documented install path half-worked. The CLI reported success, wrote its
install metadata, and the provider silently did not exist: `hermes -m <it>` said
"Unknown provider" and `/model` never listed it. Verified before the fix — a
plugin at `~/.hermes/plugins/<name>/` was NOT FOUND while the identical plugin
at `~/.hermes/plugins/model-providers/<name>/` was discovered.
Discovery now also walks the flat directory, importing only entries whose
manifest declares `kind: model-provider`. Everything else there belongs to
`PluginManager`, which owns its lifecycle and consent flow — importing it here
would run third-party code behind its back, so the tests assert we don't (with
fixtures that register on import, since a fixture that merely raised would be
swallowed by `_import_plugin_dir` and prove nothing).
Manifests are parsed with PyYAML when present and a line scan otherwise, so
provider discovery gains no hard dependency; an unreadable manifest is skipped
rather than allowed to blank the registry.
Co-Authored-By: Junie <junie@jetbrains.com>
An external-process provider is an agent CLI Hermes drives over stdio rather
than an HTTP endpoint. Three things about it were spelled out for one vendor,
and each was a hard stop for any other:
* ``resolve_provider()`` gates on ``PROVIDER_REGISTRY``. Its auto-extend from
``providers/`` covered api-key providers only, so an external-process profile
never entered it and ``hermes -m <that provider>`` died with "Unknown
provider" before a client was ever built.
* ``resolve_runtime_provider()`` keyed the external-process branch on the
literal ``"copilot-acp"``, so anything else silently fell through to the
OpenRouter default instead of its own runtime.
* ``resolve_external_process_provider_credentials()`` hardcoded the binary
(``copilot``), the argv (``--acp --stdio``), the env var names and the
placeholder api_key — so a third-party provider would have been handed
another vendor's CLI.
Now the profile carries what only the provider knows — ``process_command``,
``process_args``, ``process_command_env_vars``, ``process_args_env_var`` — and
the three core paths key on ``auth_type == "external_process"`` instead of a
name. copilot-acp's values move into its profile verbatim, so
``HERMES_COPILOT_ACP_COMMAND`` / ``COPILOT_CLI_PATH`` /
``HERMES_COPILOT_ACP_ARGS`` and its ``copilot-acp`` api_key placeholder behave
exactly as before; the new tests assert that alongside the out-of-tree case at
every step.
The error for a missing binary now names the provider and its own env override
instead of telling every user to install GitHub Copilot CLI.
Co-Authored-By: Junie <junie@jetbrains.com>
``create_openai_client`` was a hardcoded if-ladder: copilot-acp builds an ACP
stdio shim, gemini builds a native client, everything else gets an
``openai.OpenAI``. There was no extension point, so a provider whose wire
protocol is not OpenAI-over-HTTP could only be added by editing this function —
which is exactly why an ACP provider cannot ship outside this tree today, even
though ``providers/__init__.py`` has discovered out-of-tree profiles from
``~/.hermes/plugins/model-providers/`` and pip entry points for a while.
``ProviderProfile.create_client(**client_kwargs)`` closes that gap. It returns
``None`` by default, so every provider that wants the standard client is
unaffected and the existing ladder still runs as the fallback. copilot-acp is
migrated onto it — its hardcoded branch is gone and its profile supplies the
client in three lines, which is the same three lines an external package writes.
Resolution goes by provider name first, then by ``base_url`` prefix, so a
runtime configured only by URL still reaches its profile — matching what the
replaced ``startswith("acp://copilot")`` branch did. A profile that raises is
logged and skipped: a third-party plugin can fail to provide a client, but it
cannot take the turn down.
Also replaces the two ``isinstance`` checks in ``agent/auxiliary_client.py``
that mean "this client is complete, do not wrap it" with capability flags the
client class declares — ``HERMES_SKIP_TRANSPORT_WRAP`` and
``HERMES_SKIP_ASYNC_WRAP``, mirroring ``SUPPORTS_HERMES_TOOL_CALLS`` in
``background_review.py``. Two in-tree consumers (the ACP shim and the Gemini
native client), an out-of-tree client is covered by the same declaration, and
the hot path no longer imports those modules just to type-test.
Co-Authored-By: Junie <junie@jetbrains.com>
N processes sharing one Nous OAuth pool entry hit the hourly expiry
together; each force-refreshed, each rotation invalidated the token a
sibling had just adopted, and processes that lost the auth-store flock
race had their only entry benched ("matched no nous entry ... pool size
0") — ~120 sessions surfaced 401 'out of funds' on Sep 2 2026.
- resolve_nous_runtime_credentials(stale_access_token=): under the store
lock, skip the refresh POST when the on-disk token differs from the one
that failed and is usable (a peer already rotated) — adopt instead.
- credential_pool nous path: adopt a peer-rotated key after the pre-sync,
pass the failed bearer through, and treat a lock TimeoutError as
'retry later', never as an exhausted credential.
- Live 120-process stampede harness: 41 refreshes/9 unrecovered -> 1
refresh/0 unrecovered.
Use the ACP v1 session config contract advertised by session/new: locate the category=model option and apply the selected value through session/set_config_option. Retain session/set_model only as compatibility fallback for pre-configOptions agents. Reject unknown and policy-disabled values before prompting.
Verified against the installed Copilot ACP server: its model config option advertises the account-authorized choices, session/set_config_option returns the updated state, and live prompts route gpt-5.6-terra to Terra and claude-sonnet-5 to Sonnet 5.
Follow-up to the session/set_model wiring, caught in live use: picking an
org-policy-disabled model (claude-fable-5) produced a response claiming to
BE that model while Copilot actually served its default (Claude Sonnet 5).
Two causes:
1. The prompt preamble injected 'Hermes requested model hint: <id>', so
whatever model actually served the session parroted the requested name
back as its identity. Remove the line entirely — the model is applied
for real via session/set_model now, and identity must come from the
backend, not prompt suggestion.
2. session/new advertises policy-disabled ids alongside enabled ones
(_meta.copilotEnablement: 'disabled'); selecting one is accepted but
silently serves the default. Exclude disabled ids from the offered set
so the degrade-with-warning path handles them.
Verified live: requesting claude-fable-5 logs the does-not-offer warning
listing the 23 genuinely enabled models, serves the default, and the
response truthfully self-identifies as Claude Sonnet 5.
kshitijk4poor