_attach_child now mirrors a pending parent stop with the same split
interrupt() uses for its own fan-out (hard -> hard_interrupt, soft ->
interrupt), so a redirect is not turned into a cancel on a child that was
attached late. _restore_parent_cancellation collapses to re-attaching the
rejected unit's children: the replay is the attach step's job now.
Test fixture: _Batch gained origin_session_history_delivery on main after
the salvaged PR was written.
Co-authored-by: illidan <noequal666@gmail.com>
Fold the whitespace-only case into the empty-file miss test and drop
test_overwrite_is_all_or_nothing: it wrote sequentially, so it exercised
nothing about os.replace atomicity that the publish test does not.
CI runs the suite as ~96 concurrent per-file pytest subprocesses and does
not install filelock, so the adapter-guard cache in
tests/gateway/conftest.py runs lock-free there. The verdict was published
with Path.write_text (open-truncate-write): a sibling reading between the
truncate and the write saw an empty file and raised pytest.UsageError(""),
failing a healthy tests/gateway/relay/* file with a bare 'ERROR:' line
(flaky-retry frames in CI runs 32930243538, 32932008269, 32933117564).
Fix the class, not the window:
- _write_guard_cache_atomic: publish via staging file + os.replace so
readers see the old state or the complete verdict, never a torn write.
Staging name (.tmp-*) is invisible to the stale-fingerprint eviction
globs so a sibling's eviction can't unlink it mid-publish.
- _read_guard_cache: empty/vanished cache is a miss (rescan), never a
violation verdict.
Repro: 24 concurrent lock-free subprocesses over 3 relay files fired the
empty-ERROR failure 8x in 8 rounds on origin/main; 0x in 10 rounds after
the fix. Regression tests sabotage-verified.
The Python-repr pass ran after the MCP probe header scrub and collapsed
'Authorization': 'Digest ***' to '***', erasing the scheme word that scrub
deliberately keeps (tests/hermes_cli/test_mcp_probe_redaction.py went red
on rebase). Skip values that already carry a mask marker.
Port from OpenHands/software-agent-sdk#4508: their dict-entry secret
redaction was uppercase-only and leaked mixed-case keys (UserPassword,
sessionToken). Apply the same case-insensitive treatment to the Python
mapping-repr pass: a casefolded credential suffix (apikey/token/secret/
password/passwd/credential) now qualifies a key, while metadata names
(TOKEN_COUNT, password_policy, tokenizer) stay untouched.
A /p/<profile>/ cron_job route resolved and fired the job from the gateway's
default home: `_fire_cron_job` ran `execute_job_for_event` outside
`_profile_scope`, so `cron/jobs.json` lookups (`resolve_job_ref`,
`claim_job_for_fire`) and the run's config/secrets came from the wrong
profile — the agent-mode path already scopes its run, this path did not.
`asyncio.to_thread` copies contextvars, so wrapping the call is enough.
Route-level `skills` were also being injected into the rendered prompt on
cron_job routes even though the docs say they are ignored (the job's own
skills apply); skip `_apply_skills` for cron_job routes so the per-run
context stays plain event text.
Test proven red on the pre-fix tree (home resolved to the default profile),
green after.
ChatGPT Work's Aug 25 2026 release lets scheduled tasks fire from app
events (new Gmail message, Slack activity, GitHub PR feedback) instead
of polling on a cadence. This ports the pattern by composing two
existing Hermes subsystems: a webhook route can now set cron_job to
fire an existing cron job on each inbound event.
- gateway/platforms/webhook.py: cron_job route mode — after the same
HMAC auth / rate limit / filters / script / idempotency as agent
routes, the rendered prompt becomes transient per-run context and the
job fires through execute_job_for_event on a worker thread (202
Accepted immediately). Startup validation rejects cron_job +
deliver_only.
- tools/cronjob_tools.py: execute_job_for_event() — public wrapper over
the shared claimed-run body (_execute_job_now), so event fires share
at-most-once claiming, in-flight dedupe, delivery, and [SILENT]
handling with scheduler and manual runs.
- hermes webhook subscribe --cron-job: creates event-trigger
subscriptions; job ref validated (and canonicalized to the job ID) at
create time.
- Docs: webhooks.md route table + Event-Triggered Cron Jobs section,
cron.md capability list, zh-Hans mirrors.
- Tests: tests/gateway/test_webhook_cron_trigger.py (adapter + unit),
CLI tests in test_webhook_cli.py.
Port from openclaw/openclaw#120597: Docker Desktop, OrbStack, and Podman
expose host bind mounts as fuse.virtiofs or 9p. SQLite WAL shared-memory
is not coherent across the VM boundary and fails SILENTLY under write
pressure (zero-filled pages), so the existing reactive marker fallback
(_WAL_INCOMPAT_MARKERS) never fires — the mode must be refused before
the pragma.
apply_wal_with_fallback now checks the DB path against
/proc/self/mountinfo (longest mount-point prefix wins; per-directory
cached) and applies journal_mode=DELETE with a one-shot WARNING naming
the fix (native volume instead of a bind mount). On-disk WAL databases
are still never live-downgraded; require_wal callers get
WalUnsupportedError. Non-Linux hosts and unreadable mount tables
conservatively skip detection (behavior unchanged).
Sabotage-verified: disabling only the wiring makes
test_fresh_db_on_cross_vm_fs_gets_delete fail.
audio_param_key is a new _build_payload seam; one invariant covers it (Wan
emits `audio`, veo3.1 still emits `generate_audio`) plus start_image_url
and the integer duration.
Alibaba's Wan 3.0 generation went live on FAL (was "Coming Soon" since
the Wan 2.7 rollout). Adds two families to FAL_FAMILIES:
- wan-3.0: alibaba/wan-3.0/{text,image}-to-video — 2-30s integer
durations, 480p/720p/1080p, native audio, adaptive aspect handling
($0.05-$0.20/s by resolution)
- wan-3.0-prime: alibaba/wan-3.0-prime/{text,image}-to-video — premium
tier, same surface ($0.068-$0.28/s)
Schema quirks mapped from the FAL OpenAPI/llms.txt:
- i2v takes `start_image_url` (image_param_key, Kling-4K pattern)
- duration is a JSON integer (duration_int)
- audio toggle key is `audio`, not `generate_audio` — new generic
`audio_param_key` family flag in _build_payload (veo3.1 regression
asserted unchanged)
The streaming accumulator dropped delta.reasoning_details entirely — only
non-streaming responses preserved the OpenRouter unified reasoning replay
data (signatures, encrypted reasoning blocks). Providers that require the
reasoning_details sequence passed back on the next turn (OpenRouter
reasoning models, Anthropic signed thinking via shims) lost continuity on
every streamed turn, which is nearly all turns.
- Accumulate delta.reasoning_details (attr or model_extra) during
streaming and attach the merged list to the final mock message, where
_build_assistant_message's existing passthrough persists it.
- _append_streamed_reasoning_detail merges consecutive reasoning.text /
reasoning.summary fragments into one logical entry (OpenRouter streams
them as word-level deltas; unmerged they bloat the replayed signature
payload) while encrypted/opaque entries stay discrete. Later fragments
backfill signature/id/format/index fields the first fragment omitted.
Ported from earendil-works/pi#8605 (commit c5ad7c1b0), credit
@cristinaponcela for the merge-fragments pattern.
Tests: tests/run_agent/test_streamed_reasoning_details.py (8 tests;
sabotage-verified: disabling accumulation fails the E2E test, disabling
merging fails 4 unit tests). Neighboring suites green (test_streaming.py
40 passed).
Rebased onto main (~9700 commits): the branch predates the god-file
decomposition and the tests/state -> tests/hermes_state move.
Fixes for seams that moved:
- cron memory contract patches cron.scheduler_delivery._resolve_origin and
hermes_state_registry.acquire (where run_job now reads them).
- state.db conformance imports _live_writer_holds_db from
hermes_state_repair and drives the public create_quick_snapshot.
- update receipt: serve runtimes are reconciled in their own unit
vocabulary since #100479, so the "full accounting" row names the serve
unit instead of borrowing a gateway relaunch.
Deleted (change-detectors / source-greps / duplicates / dead code):
- source-text scan of hermes_state*.py for "maintenance-shaped" defs and
the symbol REGISTRY it fed (renames are not regressions).
- per-op exact-outcome table under a live writer -> one invariant: refuse
with a lock error or report zero work, DB stays intact.
- copy_db_and_verify pins: the symbol is a revert-scheduled plugin-compat
pointer, not production code (check_compat_pointers.py).
- cron ON-direction tests already pinned by tests/cron/test_scheduler.py,
plus a tautology that never called production code.
- exact warning-wording asserts in the env deprecation truth table.
- same-file duplicates: fixed-point (implied by idempotence), tripwire
round-trips, absent-registry, sequential "race" re-enactment.
Two weeks of closed issues/merged PRs show the same areas regenerating:
each salvage pinned its instance while the class invariant had no test.
These suites pin the invariants themselves:
- tests/conformance/test_profile_write_tripwire.py — no writes to the
default profile tree while a profile is active (#88532#92662#89190#89625#92156); reusable tripwire fixture, 4 surfaces
- tests/hermes_cli/test_env_deprecation_truthtable.py — 18-row truth
table for the Deprecated-.env warning (#88829#89016#89389#90299)
- tests/cron/test_cron_memory_contract.py — cron<->memory contract that
flipped twice in Aug (#91269 -> #91384 -> #91447)
- tests/agent/test_injected_param_strip_retry_registry.py — every
strippable injected param x real 400 shapes must strip-and-retry;
unknown params must still fail (#90257#89897#91164#89503)
- tests/agent/test_transcript_decoration_idempotence.py — f(f(x))==f(x)
law + 4-breakpoint budget for apply_anthropic_cache_control (#90971)
- tests/state/test_state_db_maintenance_conformance.py — registry-
enumerated maintenance ops refuse/degrade under a live writer; copies
of corrupt DBs are refused or flagged (#91839#90806#90613#88235)
- tests/tools/test_bot_mode_canonical_chat_resolution.py — canonical
Bot Chat resolution is idempotent, never mints, unique per profile,
race-safe (#92040#90705#92692#90005#90732, PR #92129)
- tests/hermes_cli/test_update_receipt_truthfulness.py — receipts:
crash never claims success; success requires full fleet accounting;
refusal != failure (#91283#91439#92902#92780)
117 tests, all sabotage-verified (each suite proven to FAIL when its
bug class is reintroduced).
The shell `find ~/.hermes/skills ~/.hermes/hermes-agent/optional-skills ...`
snippet hardcoded a dev-clone path that does not exist on user installs; the
loader already substitutes ${HERMES_SKILL_DIR} (skills.template_vars), which
is what every other bundled skill uses. Drop the upstream-agent path mention.
Regenerated the docs page so it matches SKILL.md (also fixes the comfyui
related-skill link, which now lives under optional).
Optional skill: scroll-as-timeline landing pages on a deterministic
CSS/JS engine, with interview → page grammar → signature move workflow
and screenshot-based scroll verification. Engine and scripts vendored
verbatim; asset generation re-anchored on image_generate with the
upstream kie.ai flow kept as an optional path.
- The Kanban plugin cannot import `@/lib/ime` (plugin fence), so its
`./ime-enter` twin duplicated the predicate. Export `isSubmitEnter` from
`@hermes/plugin-sdk` and drop the twin so one helper owns the policy.
- `BoardNameField` in kanban/board-switcher.tsx (main's refactor of the
create/rename board dialogs) submitted on composition Enter; guarded.
- Telegram allowed-ID input and the clarify-card textarea submitted on the
post-compositionend keyCode-229 Enter; both now use `isSubmitEnter`.
Widens the salvaged Kanban fix (PR #94611 by @huklaa) to the whole bug
class, following cloudflare-os PR #291 which centralized one
isImeComposing predicate and swept every Enter-submit site.
- New shared helper src/lib/ime.ts (isImeComposing / isSubmitEnter):
handles nativeEvent.isComposing (React), event.isComposing (DOM), and
the legacy Chromium/Safari keyCode 229 commit-Enter that arrives after
compositionend.
- Sweeps 21 previously unguarded Enter-submit sites: dialogs (project,
worktree, profile-remote-override, pet rename), settings fields
(credential keys, model API key, quick-entry shortcut, attachment
size, combobox), quick entry, pet overlay composer, pet generate +
hatch, review ship bar, file rename, preview browser address bar,
model catalog menu, MCP setup + approval strips, Kanban drawer.
- Upgrades two partial guards (onboarding, session-actions-menu) that
checked isComposing but missed keyCode 229.
- find-in-page: Enter step no longer fires mid-composition (typed CJK
search queries jumped the viewport on every conversion commit).
Validation: tsc clean, eslint clean, 135 tests green across ime/find-bar/
composer/kanban suites; sabotage run (guard removed) fails the new test.
Pasting exactly one http(s) link while composer text is selected now
turns the selection into a markdown link ([selected text](url)) instead
of replacing it — the behavior every rich text editor ships.
- resolveExactLinkPaste(): recognizes a clipboard payload that is exactly
one supported link (bare or <...>-wrapped, host required, no prose or
trailing punctuation) and returns the href.
- selectionLinkLabel(): the selected composer text eligible for linking;
rejects collapsed selections, selections spanning ref chips or line
breaks, and whitespace-only selections.
- markdownLinkFor(): builds the markdown link, escaping square brackets.
- handlePaste wires the three together ahead of the @url: chip path;
any non-qualifying paste falls through to existing behavior.
Adapted from Buzz's TipTap link-mark approach to Hermes' contenteditable
composer: we emit a markdown link (the composer's native rich construct)
rather than a ProseMirror mark.
The prompt template is built from _PROMPT_GOOD_EXAMPLES, so asserting the
template contains them can never fail independently of the code it checks.
The behavioural invariants (echo rejected, greeting allowed, near-miss
passes) stay.
Small title models parroting a prompt example back verbatim produced
sessions named "Fix login button on mobile" with no relation to the
conversation. The example lines in _TITLE_PROMPT_TEMPLATE now render
from _PROMPT_GOOD_EXAMPLES so the guard set and prompt cannot drift,
and generate_title rejects exact (case-insensitive, wrapper-stripped)
echoes so the instant derived title survives instead. 'Friendly
greeting' stays allowed — it is prescribed output for bare greetings.
Anthropic's /v1/models is cursor-paginated with a default page size of 20.
Both hermes fetchers read a single unpaginated page, so any model past the
first page silently vanished from the /model picker and provider catalogs.
- hermes_cli/models.py _fetch_anthropic_models(): request limit=1000 and
follow has_more/last_id (bounded, repeated-cursor guarded, de-duped)
- plugins/model-providers/anthropic fetch_models(): same pagination walk,
and it now honors the base_url argument instead of hardcoding
api.anthropic.com
- tests: live-HTTP paginated-server regression tests for both fetchers,
incl. single-page and stuck-cursor termination; updated the two URL-pinning
pool-discovery tests for the ?limit=1000 contract
Live anonymous probes (2026-09-13, HermesAgent UA, 4 rounds): nemotron-3.5-lightning-free
returned zero bytes for >90s on every attempt and nemotron-3-ultra-free took ~40s, while
mimo-v2.5-free answered 200 in 2-4s each time. An aux model (titles, compression, memory)
that hangs is worse than a delisted one, so the default moves to the model that works.
The hy3-free fixture swap left two identical nemotron-3.5-lightning-free
assertions; use mimo-v2.5-free so the line still covers a second live
free-tier slug.
The OpenCode Zen relay no longer serves hy3-free (since ~2026-08-31) or
laguna-s-2.1-free (new, verified 2026-09-09): both are gone from the live
GET /zen/v1/models catalog and anonymous chat completions return
401 {"type":"ModelError","message":"Model <id> is not supported"}
(2 probes >=60s apart, x-opencode-session header present).
- hermes_cli/models_catalog_static.py: remove both slugs from the
opencode-free offline floor and the opencode-zen discovery floor;
document the delist dates in the catalog comment.
- plugins/model-providers/opencode-free: default_aux_model moves from the
dead laguna-s-2.1-free to nemotron-3.5-lightning-free (fastest surviving
anonymous model).
- tests: swap fixtures off the dead slugs; extend the floor-exclusion
invariant to cover both.
The live revalidation path already hides them when the relay is reachable;
this fixes the OFFLINE floor and the aux default, which would otherwise
offer/route to models that 401.
data:image/... entries were treated as local paths and silently skipped as
"unreadable". They now ride as image_url parts only (never pasted into the
text hint, never appended to a text-mode goal). Skips and forwarding
failures log at warning since the caller explicitly asked for the images;
decide_image_input_mode gets the child's requested_provider like the CLI
and gateway callers.
Tests collapse to five invariants, including one that drives _ChildRun
and asserts the multimodal content list reaches run_conversation as the
first user turn. Docs mention data: URLs and the read guard.
Subagents can now SEE images. Each delegate_task task accepts an optional
images list (max 8; local paths or http(s) URLs). Vision-capable children
receive native image_url content parts on their goal turn (local files as
data URLs, remote URLs verbatim); non-vision children get
[Image attached at: ...] hints plus a vision_analyze pointer. Routing
reuses agent.image_routing (decide_image_input_mode /
build_native_content_parts), so agent.image_input_mode governs delegation
exactly like inbound gateway images.
Best-effort by contract: malformed images arrays fail the call loudly
before any child spawns; unreadable paths are skipped with a log line;
any exception in the forwarding path degrades to the text-only goal.
Adapted from RooCodeInc/Roomote#1796 / #1767 (Fast agent forwards bounded
current-turn attachments to delegated coding tasks).
Drop the no-op and still-running change-detectors (4 invariant tests remain:
pipe closed on finish, PTY closed on finish, poll still serves buffered
output, prune releases handles). The accretion-caps fake session now
carries process/_pty like the real dataclass, since prune reads them.
_release_finished_handles read the dataclass fields through getattr
fallbacks and swallowed every exception; use the real attributes, suppress
only OSError/ValueError on the pipe close (an stdin flush can hit EPIPE),
and rely on ptyprocess/pywinpty close() idempotence for the master fd.
The call-site comment claimed the reader had always drained the pipe;
on kill_process/_reconcile_local_exit the reader may still be reading,
so state what actually happens (its next read raises, the loop exits).
Widen #75162: _prune_if_needed() drops finished sessions (TTL expiry and
oldest-finished eviction at MAX_PROCESSES) — release their Popen/PTY
handles there too, covering sessions inserted into _finished without
passing through _move_to_finished(). The release helper is idempotent,
so double-close on the normal path is a no-op. Adds two tests: prune
releases handles of dropped sessions, and a still-running session's
pipe stays open.
Finished sessions retained their subprocess.Popen pipe objects (and PTY
masters) until the finished-process TTL (FINISHED_TTL_SECONDS, default 30
minutes) elapsed. Under heavy background churn — deployments, archivers,
watchers — finished-but-unpruned sessions accumulated one open pipe FD
each, exhausting the gateway process's file descriptor budget and
surfacing as a 'file descriptor limit' error on new background spawns.
The registry never rejects spawns (it prunes oldest-finished at
MAX_PROCESSES), so the real defect was the retained-handle leak, not a
registry-cap rejection. The fix closes each finished session's Popen
stdout/stderr/stdin streams and PTY master in _move_to_finished(), right
after the reader loop drains EOF. poll()/wait()/read_log() serve output
from the buffered output_buffer — never from the pipe — so the release is
lossless.
Tests: 4 new cases in TestFinishedHandleRelease — Popen pipes closed,
PTY closed, no-handle sessions safe, and poll() still serves buffered
output after the release. All 4 fail on main (reproduction) and pass
with the fix.
Port from nearai/ironclaw#7965: BM25 admits any document scoring above
zero, i.e. sharing ONE term with the query. A long descriptive search
for a capability that does not exist therefore returned a plausible-
looking ranked list, and the model read 'results exist' as 'it is in
here somewhere' and rephrased instead of stopping (IronClaw production
trace: 652 tool calls, 216 of them tool_search, hunting a 'data' tool
that did not exist).
A document must now match at least half the query's ANSWERABLE terms
(terms present anywhere in the index) before it is offered. Coverage
only engages from four answerable terms up, preserving recall on short
queries; exact tool-name matches remain authoritative; the substring
fallback is unchanged.
Docs: relevance-floor bullet added to tool-search.md implementation
details.
Adds fal-ai/kling-image/v3/text-to-image ($0.028/img, native 2K default,
8 aspect ratios) with its image-to-image edit endpoint. The i2i schema
takes a SINGULAR `image_url` string instead of the usual `image_urls`
list, so the catalog gains an `edit_image_param` knob that
_build_fal_edit_payload honors (first source image only); the
edit-contract test now validates whichever image key the entry declares.
Adds kling-v3 (fal-ai/kling-video/v3/standard/*) and kling-v3-pro
(fal-ai/kling-video/v3/pro/*) to FAL_FAMILIES: start_image_url i2v key,
aspect_ratio dropped on i2v, string duration 3-15s, generate_audio and
negative_prompt real, no seed/resolution keys per the published llms.txt
schemas. Payload shapes pinned in tests; docs mention updated.
The ACP schema (agent-client-protocol 0.9.0, ContentChunk.messageId) says
"Both clients and agents MUST use UUID format for message IDs". The
ported allocator emitted hermes-assistant-N strings, which a strict
client may reject or fail to group. A fresh uuid4 per message keeps the
grouping semantics and can never collide with an earlier turn's id, so
the counter/prefix state is gone.
Tests trimmed to three invariants: chunks share one UUID until the None
flush sentinel (empty deltas ignored), thought + text share an id, and
the no-allocator shape stays unchanged.
ACP clients group streamed agent_message_chunk / agent_thought_chunk
updates into one assistant reply by messageId, and use a new id to
start the next reply (root-reply replacement semantics). Hermes' ACP
adapter sent every chunk without a messageId, so clients that replace
'the current assistant message' per chunk collapsed separate
autonomous turns into one bubble.
- AssistantMessageIdAllocator (per ACP session, monotonic across
turns): a contiguous run of reasoning + text deltas shares one
hermes-assistant-N id; the None flush sentinel Hermes core emits
before tool execution / at end of stream closes it.
- make_message_cb / make_thinking_cb stamp update.message_id when an
allocator is provided; legacy no-allocator shape unchanged.
- Unstreamed final responses open their own id; plugin-transformed
responses reuse the streamed message's id (replacement).
- Tests: grouping until flush, thought+text sharing, monotonic ids,
empty-string vs None sentinel, legacy shape.
The rebase moved the inbound attachment loop into SlackAdapter._append_link_unfurls,
so the nested-table hunk now lives there and is asserted directly. Drop the source-
provenance references and duplicate cell-level cases; one ragged/malformed-row test
covers raw_text, rich_text, None and unknown cell types.
Port from qwibitai/nanoclaw#3666: Slack represents a pasted table as
'table' blocks — usually nested in attachments[].blocks[], sometimes
top-level. They appear in neither the message text nor the file list,
so the agent received the sentence before the table and nothing else.
- _render_slack_table_block(): projects rows as 'cell | cell' lines,
collecting text leaves from raw_text/rich_text cell subtrees; capped
at 20k chars with a visible '[table truncated]' marker.
- Wired into all three ingestion paths: _extract_text_from_slack_blocks
(thread history + attachment-nested blocks), the live inbound
attachment loop, and _extract_additional_text_from_slack_blocks
(top-level blocks on live messages).
- _serialize_slack_blocks_for_agent skips 'table' blocks — the
allowlist drops 'rows', so it only emitted an empty husk.
Parametrize the reject case over send_video/send_document/send_voice
(each asserts no file upload, base fallback never runs, error names the
file, size and limit), fold the all-oversized batch case into the batch
test, and drop the duplicated fixture setup.
send_voice built its own discord.File from the audio bytes and never ran
the size preflight, so an oversized audio attachment still burned the
doomed 413 round-trip that #50846 is about. Route it through the same
_reject_oversized_upload helper as _send_file_attachment.
The limit constant and _discord_upload_limit_bytes lived on the adapter
facade while every consumer is in adapter_media.py; the facade+sibling
layout puts topic code in the sibling, so they move there.
Discord raised the default file upload limit from 10 MiB to 20 MiB for
users, bots, webhooks and interaction responses (developer changelog,
Sep 3 2026). The 25 MiB constant here predates the preflight salvage and
never matched the platform; more importantly discord.py 2.7.1 still
reports 10 MiB via guild.filesize_limit for unboosted guilds, so the
guild-aware path under-reported the cap and rejected 10-20 MiB files
Discord now accepts. Floor the guild value at the platform default so a
stale library constant can only widen, never shrink, the preflight.