…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

* chore(release): add Zhekinmaksim to AUTHOR_MAP (#22449)

Maps zhekinmaksim@gmail.com to GitHub login Zhekinmaksim so
contributor_audit.py recognizes their authored commit in the
upcoming #21930 salvage PR.

* fix(async): replace get_event_loop() with get_running_loop() in async contexts

Follow-up to PR #21293 (cli.py), which fixed the same anti-pattern.
`asyncio.get_event_loop()` is documented as effectively "always returns
the running loop when called from a coroutine" and emits
DeprecationWarning/RuntimeWarning in some interpreter configurations.
The Python docs explicitly recommend get_running_loop() inside coroutines.

Replaces the remaining 9 call sites that are unconditionally inside
async def bodies:

- tools/browser_cdp_tool.py — _cdp_call() (4 sites): deadline + remaining
  computations inside the async websockets.connect context manager.
- hermes_cli/web_server.py — get_status, _start_device_code_flow,
  submit_oauth_code (3 sites): all FastAPI async endpoints offloading
  blocking httpx / PKCE work to run_in_executor.
- environments/agent_loop.py — HermesAgentLoop (1 site): tool dispatch
  inside the async rollout loop.
- environments/benchmarks/terminalbench_2/terminalbench2_env.py —
  rollout_and_score_eval (1 site): test verification thread offload.

All 9 sites are unconditionally inside async def bodies, so a running
loop is guaranteed and no try/except RuntimeError fallback is needed
(unlike the cli.py case in #21293, which ran from a background thread).

Behavior is identical on supported Python versions; aligns the codebase
with the post-#21293 idiom and avoids future warnings as the deprecation
hardens.

Salvaged from PR #21930 by @Zhekinmaksim onto current main (the
original branch was 109 commits behind and carried unintended
stale-branch reverts of unrelated landed changes — _tail_lines
encoding=utf-8 and the Windows PTY bridge guard). Only the 9 swaps
from the PR's intended scope are applied here.

* chore(release): add KvnGz to AUTHOR_MAP (#22458)

Maps obafemiferanmi1999@gmail.com (the commit-author email used on
PR #21473's branch) to GitHub login KvnGz (the PR/branch owner) so
contributor_audit.py recognizes the authored commit in the upcoming
salvage PR.

* fix(tests): pin UTF-8 encoding when reading source files on Windows

Three tests in tests/agent/test_auxiliary_config_bridge.py read
in-tree source files (gateway/run.py and cli.py) via
Path.read_text() with no encoding argument.  The default falls
back to the system locale, which on Western Windows installs is
cp1252, and the read fails as soon as the source contains any
byte that isn't valid cp1252 (e.g. an em-dash in a comment):

    UnicodeDecodeError: 'charmap' codec can't decode byte 0x8f
    in position 41190: character maps to <undefined>

Linux CI doesn't catch this because the default Linux locale is
UTF-8.  Windows contributors hit it on every run of the test suite.

Pin encoding="utf-8" on the three call sites that read repo
source files.  This matches the existing precedent in
hermes_cli/doctor.py:363, where the same pattern (with an
explanatory comment) was applied to fix the .env read on
non-UTF-8 Windows locales.

Affected tests now pass on Windows + Python 3.12:
  - TestGatewayBridgeCodeParity.test_gateway_has_auxiliary_bridge
  - TestGatewayBridgeCodeParity.test_gateway_no_compression_env_bridge
  - TestCLIDefaultsHaveAuxiliaryKeys.test_cli_defaults_can_merge_auxiliary

* feat(plugins): add standalone_sender_fn for out-of-process cron delivery

Plugin platforms (IRC, Teams, Google Chat) currently fail with
`No live adapter for platform '<name>'` when a `deliver=<plugin>` cron
job runs in a separate process from the gateway, even though the
platforms are eligible cron targets via `cron_deliver_env_var` (added
in #21306). Built-in platforms (Telegram, Discord, Slack, etc.) use
direct REST helpers in `tools/send_message_tool.py` so cron can deliver
without holding the gateway in the same process; plugin platforms
historically depended on `_gateway_runner_ref()` which returns `None`
out of process.

This change adds an optional `standalone_sender_fn` field to
`PlatformEntry` so plugins can register an ephemeral send path that
opens its own connection, sends, and closes without needing the live
adapter. The dispatch site in `_send_via_adapter` falls through to the
hook when the gateway runner is unavailable, with a descriptive error
when neither path applies. The hook is optional, so existing plugins
are unaffected.

Reference migrations land in the same change for IRC, Teams, and
Google Chat, exercising the hook across stdlib (asyncio + IRC protocol),
Bot Framework OAuth client_credentials, and Google service-account
flows respectively.

Security hardening on the new code paths:
* IRC: control-character stripping on chat_id and message body to
  block CRLF command injection; bounded nick-collision retries; JOIN
  before PRIVMSG so channels with the default `+n` mode accept the
  delivery.
* Teams: TEAMS_SERVICE_URL validated against an allowlist of known
  Bot Framework hosts (`smba.trafficmanager.net`,
  `smba.infra.gov.teams.microsoft.us`) to block SSRF; chat_id and
  tenant_id constrained to the documented Bot Framework character set;
  per-request timeouts so a slow STS endpoint cannot starve the
  activity POST.
* Google Chat: chat_id and thread_id validated against strict
  resource-name regexes; service-account refresh wrapped in
  `asyncio.wait_for` so a hung token endpoint cannot stall the
  scheduler.

Test coverage: 20 new tests covering happy path, missing-config errors,
network failure modes, and each defensive validation. Existing tests
unchanged. `bash scripts/run_tests.sh tests/tools/test_send_message_tool.py
tests/gateway/test_irc_adapter.py tests/gateway/test_teams.py
tests/gateway/test_google_chat.py` reports 341 passed, 0 regressions.

Documentation: new "Out-of-process cron delivery" section in
website/docs/developer-guide/adding-platform-adapters.md and an entry
in gateway/platforms/ADDING_A_PLATFORM.md naming the hook.

* fix(profiles): exclude infrastructure artifacts when cloning with --clone-all

When the source profile is the default (~/.hermes), shutil.copytree()
was copying multi-GB infrastructure alongside the ~40 MB of actual
profile data: hermes-agent/ (repo checkout + 3 GB venv), .worktrees/,
profiles/ (sibling profiles — recursive!), bin/ (installed binaries),
node_modules/ (hundreds of MB).

Add _CLONE_ALL_DEFAULT_EXCLUDE_ROOT frozenset with these five entries
and pass an ignore callback to copytree().  Exclusions are gated on
the source actually being the default profile (is_default_source) so
named-profile sources are never affected.

Also exclude at any depth: __pycache__/, *.pyc, *.pyo, *.sock, *.tmp.
Profile data (config.yaml, .env, auth.json, state.db, sessions/,
skills/, logs/) is preserved intact — clone-all means 'complete
snapshot minus infrastructure'.

Mirrors the approach already used by _default_export_ignore() and
_DEFAULT_EXPORT_EXCLUDE_ROOT (the export-side exclusion set which is
broader because it produces a portable archive, not a live clone).

Co-authored-by: MustafaKara7 <karamusti912@gmail.com>
Co-authored-by: fahdad <30740087+fahdad@users.noreply.github.com>
Fixes #5022
Based on PRs #5025, #5026, and #21728

* fix(banner): resolve update-check repo from running code, not profile-scoped path

check_for_updates() and _resolve_repo_dir() were preferring
$HERMES_HOME/hermes-agent/ over Path(__file__).parent.parent.resolve()
when looking for a .git checkout.  For profiles created with
--clone-all, $HERMES_HOME/hermes-agent/ points to a stale copy
with a frozen HEAD, causing persistent "N commits behind" banners
that never resolved.

Flip the resolution order: prefer the running code's location first,
fall back to $HERMES_HOME/hermes-agent/ only when the live checkout
doesn't have a .git (system-wide pip installs, distro packages).

The embedded-rev branch (HERMES_REVISION env var, set by nix builds)
is unaffected — it uses git ls-remote against upstream, never reads
the local checkout's HEAD.

Based on PR #21728 by @fahdad

* fix(gateway): clear slash-confirm state during session boundary cleanup

* feat(gateway): stream Telegram edits safely

* chore: add nik1t7n to AUTHOR_MAP

Nikita Nosov (nik1t7n, PR #22264) — first-time contributor email
and noreply alias.

* fix(telegram): exclude row-label column from bullet items in table rendering

When a GFM table has a row-label column (first column with no header),
_render_table_block_for_telegram incorrectly included the row-label cell
in the bullet zip alongside the data cells, producing a spurious bullet
like '• 維度: 核心賣點' before the real data rows.

Detect the row-label column by comparing the first data row cell count
against the header count (has_row_label_col = len(first_data_row) ==
len(headers) + 1). When present, use cells[0] as the heading and
zip headers against cells[1:] only, correctly excluding the row-label
from the bullet list.

Fixes #22604

* feat: confirm prompt for destructive slash commands (#4069) (#22687)

/clear, /new, /reset, and /undo now ask the user to confirm before
discarding conversation state — three-option prompt routed through the
existing tools.slash_confirm primitive.

Native yes/no buttons render on Telegram, Discord, and Slack (their
adapters already implement send_slash_confirm); other platforms get a
text-fallback prompt and reply with /approve, /always, or /cancel.

The classic prompt_toolkit CLI uses the same three-option flow via the
established _prompt_text_input pattern (see _confirm_and_reload_mcp).
TUI keeps its existing modal overlay (#12312).

Gated by new config key approvals.destructive_slash_confirm (default
true). Picking 'Always Approve' flips the gate to false so subsequent
destructive commands run silently — matches the established
mcp_reload_confirm UX.

Out of scope: /cron remove (separate domain — scheduled jobs, not
session history). Existing TUI overlay env-var (HERMES_TUI_NO_CONFIRM)
left unchanged; cosmetic unification can come later.

Closes #4069.

* fix(kanban): call recompute_ready after unlink_tasks removes a dependency

Problem:
unlink_tasks() removes a parent→child dependency edge but does not trigger
recompute_ready().  A child whose last blocking parent is unlinked stays
stuck in 'todo' indefinitely — it only promotes to 'ready' on the next
dispatcher tick or a manual 'hermes kanban recompute'.  For CLI-only users
without a dispatcher, the child is permanently stuck.

Root cause:
complete_task() and unblock_task() both call recompute_ready() after their
write transaction so downstream children are evaluated immediately.
unlink_tasks() was missing this call — removing a dependency is
semantically equivalent to completing one, so the same recompute is needed.

Fix:
Capture the rowcount result before the write_txn exits, then call
recompute_ready(conn) outside the transaction when a row was actually
deleted (so the child sees the updated task_links state).

Tests:
Added test_unlink_tasks_triggers_recompute_ready in
tests/hermes_cli/test_kanban_db.py: creates parent A (done) + parent C
(running), child B with both parents (todo), unlinks C→B, asserts B is
ready immediately.  Stash-verified: FAILS without fix (child stays todo),
PASSES with fix.
62/62 tests green in tests/hermes_cli/test_kanban_db.py.

Closes #22459.

* chore: add wesleysimplicio to AUTHOR_MAP

* perf(google_chat): defer heavy google-cloud imports to first adapter use (#22681)

Plugin discovery imports every bundled platform plugin at model_tools
import time. The google_chat adapter unconditionally pulled in
google.cloud.pubsub_v1, googleapiclient, grpc, httplib2, and friends at
module top — about 33 MB RSS and 110 ms wall on every CLI invocation,
even ones that never construct a gateway adapter.

Wrap the heavy imports in _load_google_modules(): an idempotent loader
that rebinds the module-level globals (pubsub_v1, service_account,
HttpError, MediaFileUpload, …) on first call and is invoked from
GoogleChatAdapter.__init__, connect(), and check_google_chat_requirements().

The HttpError = Exception placeholder is preserved for the brief window
before the loader runs, so 'except HttpError as exc:' clauses stay
correct (Python looks up the name at try/except evaluation time, not
at function definition time).

Measured impact on a 9950X3D, 7-run medians:
  import cli:              895 → 787 ms  (-108 ms / -12%)
                           133 → 110 MB  ( -23 MB / -17%)
  import model_tools:      491 → 400 ms  ( -91 ms / -19%)
                            95 →  66 MB  ( -29 MB / -31%)
  google_chat alone:       244 → 132 ms  (-112 ms / -46%)
                            83 →  50 MB  ( -33 MB / -40%)
  hermes chat -q (cold):   177 → 145 MB  ( -32 MB / -18%)

Real-world win lands on every path that imports cli.py: hermes chat,
hermes gateway, cron jobs, batch runs, subagents. Long-lived gateway
processes save ~30 MB resident.

All 157 google_chat tests pass; full gateway suite (5050 tests) green.

* feat(plugins): HERMES_PLUGINS_DEBUG=1 surfaces plugin discovery logs (#22684)

Plugin authors had no easy way to figure out why their plugin wasn't
loading — failures were buried in agent.log at WARNING and skip reasons
(disabled, not enabled, depth cap, exclusive) were DEBUG-only and
invisible by default.

Set HERMES_PLUGINS_DEBUG=1 to attach a stderr handler at DEBUG to the
hermes_cli.plugins logger only. Surfaces:

  - which directories were scanned + manifest counts per source
  - per manifest: resolved key, name, kind, source, on-disk path
  - skip reasons (disabled, not enabled, exclusive, depth cap, no register)
  - per load: tools/hooks/slash/CLI commands the plugin registered
  - full traceback on YAML parse failure (exc_info on the existing warning)
  - full traceback on register() exceptions, pointing at the plugin author's line

Env var off (default) → zero new stderr output, same as before.

Touches only hermes_cli/plugins.py + a doc section in the plugin-build
guide + an entry in the env-vars reference. 3 new tests lock the
attach/idempotent/no-attach behavior.

* fix(kanban): gate claim + unblock on parent completion

Enforce the parent-completion invariant at claim_task (the single
ready->running chokepoint) and re-gate unblock_task so blocked->ready
only fires when parents are done. Prevents child tasks from running
ahead of in-progress parents under the create-then-link race.

Also adds a stress test that races concurrent create+link against
hammered claim_task and asserts no child runs while any parent is undone.

Ref: kanban/boards/cookai/workspaces/t_a6acd07d/root-cause.md
Refs: t_8d6af9d6

* chore: add SiliconID to AUTHOR_MAP

* feat(delegate): show user's actual concurrency / spawn-depth limits in tool description (#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

* fix(cli): expand composite toolset when mixed with configurables in platform_toolsets

When platform_toolsets[<platform>] contains both a composite (e.g.
hermes-cli) and at least one configurable opt-in (e.g. spotify), the
has_explicit_config branch in _get_platform_tools silently dropped the
composite, leaving sessions with only the configurable + plugin tools
and no native tools (terminal, file, web, browser, memory, etc.).

Mirror the else-branch's subset inference for composites that sit
alongside the configurables, but apply _DEFAULT_OFF_TOOLSETS only to the
implicit expansion so user-listed default-off toolsets (spotify,
discord) survive.

* fix(gateway): refresh runtime argv metadata

* fix(plugins): resolve Git binary for installs under minimal PATH

Resolve git via shutil.which with POSIX and Git-for-Windows fallbacks before clone and pull so Dashboard/API installs do not misreport Git as missing.

Add regression tests for the resolver and pull subprocess invocation.

* chore: add xieNniu to AUTHOR_MAP

* fix(telegram): honor message.quote for partial-quote reply context

When a Telegram user replies using the native quote feature to select
only part of a prior message, _build_message_event was injecting the
ENTIRE replied-to message into reply_to_text via
message.reply_to_message.text/caption. python-telegram-bot exposes
the user-selected substring as message.quote (TextQuote.text); we now
prefer that and fall back to the full replied-to text only when no
native quote is present.

The agent-visible "[Replying to: \"...\"]" prefix can otherwise expand
the user's narrow quote into the full prior message, causing the agent
to act on unrelated actionable-looking text the user did not select
(e.g. multi-item briefings where the user quotes one bullet but the
prefix injects every bullet). Falls back cleanly when message.quote
is absent (PTB <21 or replies that don't quote a substring).

Fixes #22619

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(profiles): honour active_profile when HERMES_HOME points to hermes root

Problem:
After `hermes profile use NAME`, the gateway (started via systemd with
HERMES_HOME=/root/.hermes hardcoded) ignores the active profile and
always runs as the Default profile.  WebUI, Telegram, and all non-CLI
platforms are affected.

Root cause:
_apply_profile_override() contained an early-return guard:

    if profile_name is None and os.environ.get("HERMES_HOME"):
        return   # trust the inherited value

The intent was to let child processes inherit their parent's profile via
HERMES_HOME without redundantly re-reading active_profile.  But
systemd also sets HERMES_HOME — to the hermes root (/root/.hermes),
not a profile directory — so the guard fired and silently skipped the
active_profile check.  The user's `hermes profile use NAME` write to
~/.hermes/active_profile was never seen by the gateway process.

Fix:
Only skip the active_profile check when HERMES_HOME is already a
profile directory, identified by its immediate parent directory being
named "profiles" (e.g. ~/.hermes/profiles/coder or
/opt/data/profiles/coder).  When HERMES_HOME points to a root
directory (parent name != "profiles"), continue to read active_profile.

Tests:
- test_hermes_home_at_root_with_active_profile_is_redirected: the
  bug scenario — HERMES_HOME=/root/.hermes + active_profile=coder →
  HERMES_HOME must be redirected to .../profiles/coder.
  Stash-verified: FAILS without fix, PASSES with fix.
- test_hermes_home_already_profile_dir_is_trusted: child-process
  inheritance contract unchanged — .../profiles/coder is trusted as-is.
- test_hermes_home_unset_reads_active_profile: classic path unchanged.
- test_hermes_home_unset_default_profile_no_redirect: "default" still
  produces no redirect.
4/4 tests green.

Closes #22502.

* fix(dingtalk): clarify webhook media behavior

* fix(dingtalk): align override signatures with base + guard Optional[error] in tests

* feat(mcp): add codex preset for built-in MCP server discovery

Adds 'codex' to the _MCP_PRESETS registry so users can add it via

  Connecting to 'codex'...

  ✓ Connected! Found 2 tool(s) from 'codex':

    codex                                    Run a Codex session. Accepts configuration parameters matchi...
    codex-reply                              Continue a Codex conversation by providing the thread id and...

  Enable all 2 tools? [Y/n/select]:
  Cancelled. without manually specifying
the command and args.

Enables: codex mcp-server → Hermes native MCP client → Codex tools
available as first-class Hermes tools.

* fix(cron): avoid github skill false positives in scanner

* fix(cron): keep auth-header exfiltration blocked

* fix(cron): allow quoted URL in github auth-header allowlist

The github-pr-workflow skill wraps the URL in double-quotes
('curl -H ... "https://api.github.com/..."'), which the original
allowlist regex (\s+https://api...) did not match. Without this,
the bundled github-pr-workflow skill is still blocked at every
cron tick despite #22605's fix landing for the bare-URL form.

Make the leading quote optional and add a regression test pinning
both single- and double-quoted forms.

* tests: add Windows skip guards for UNIX-only stdlib imports

* fix: move pytest.importorskip below pytest import in skip-guarded tests

The original PR placed 'pwd = pytest.importorskip("pwd")' on line 4
but 'import pytest' on line 9 — NameError on module load. Same for
test_file_sync_back.py. Plus, the in-function 'pwd = pytest.importorskip'
calls in test_auto_detected_root_is_rejected confused Python's scope
analysis (later 'import pytest' made pytest local everywhere in the
function) and caused UnboundLocalError. Drop the now-redundant
in-function importorskip calls and rely on the module-level guard.

* chore: add wali-reheman to AUTHOR_MAP

* feat(gateway): add Telegram guest mention mode

* fix: follow-up for salvaged PR #22263

- Restore allowed_chats gate before thread_id check so ignored_threads
  applies universally (even to guest mentions).
- Compute _message_mentions_bot once in _should_process_message to
  eliminate redundant second entity scan when guest_mode=true and the
  message does not mention the bot.
- Remove redundant _is_group_chat from _is_guest_mention (caller already
  verified the message is a group chat).
- Update _telegram_allowed_chats docstring to note guest_mode exception.
- Add test coverage: bot_command entity, text_mention entity,
  caption_entities, and ignored_threads + guest_mode interaction.
- Add nik1t7n to AUTHOR_MAP.

* fix(agent): notify context engine on commit_memory_session (#22764)

When session_id rotates (e.g. /new), commit_memory_session was firing
MemoryManager.on_session_end but skipping ContextEngine.on_session_end.
Engines that accumulate per-session state (LCM-style DAGs, summary
stores) leaked that state from the rotated-out session into whatever
continued under the same compressor instance.

Mirror the call shutdown_memory_provider already makes — same
lifecycle moment, same hook contract ("real session boundaries (CLI
exit, /reset, gateway expiry)"). /new is a real boundary for the old
session_id; providers keep their state but the rotated-out session_id
is done.

6 regression tests covering both-hooks-fire, no-memory-manager,
no-context-engine, both failure-tolerant paths.

Closes #22394.

* fix(tests): harden run_tests.sh — uv-aware bootstrap + scrub HERMES_CRON_SESSION (#22767)

Two unrelated but co-located fixes to scripts/run_tests.sh:

1. pytest-split bootstrap (#22401): the script tried '$PYTHON -m pip
   install pytest-split' on first run, but uv-created venvs ship without
   pip. Result: 'No module named pip' before any test ran. Add a uv
   fallback (uv pip install --python $PYTHON), keep pip as a secondary
   path, and emit a clear error pointing at 'uv pip install -e ".[dev]"'
   when neither is available. Also declare pytest-split in
   pyproject.toml dev extra so a normal '.[dev]' install provisions it.

2. HERMES_CRON_SESSION leak (#22400): the hermetic env scrub already
   unsets HERMES_GATEWAY_SESSION and HERMES_INTERACTIVE but missed the
   sibling HERMES_CRON_SESSION. When run_tests.sh is invoked from a
   Hermes cron job, that variable leaks into pytest, flipping
   tools/approval.py into cron-deny mode and breaking
   tests/acp/test_approval_isolation.py and friends.

Closes #22400.
Closes #22401.

* fix(kanban): sanitize comment author rendering in build_worker_context (#22769)

Operator-controlled HERMES_PROFILE values were rendered as
'**${author}** (${ts}):' — markdown bold with no provenance prefix.
Worker comment bodies render directly underneath. A misleading
profile name like 'hermes-system' or 'operator' could be misread by
the next worker as a system directive above attacker-influenced
content (confused-deputy primitive gated on operator misconfig).

The LLM-controlled author-forgery surface was already closed in
#22435 (author removed from KANBAN_COMMENT_SCHEMA). This is
defense-in-depth: render with an explicit 'comment from worker
`<author>` at <ts>:' prefix so even 'hermes-system' resolves to
'comment from worker `hermes-system` at ...' — parseable as
worker-comment metadata, not a system directive. Strip backticks
from author so they can't break out of the fence.

Update test_build_worker_context_caps_comments to count by body
regex since the rendered author line now also starts with
'comment '.

Closes #22452.

* fix(agent): hydrate memory-nudge counters from conversation_history (#22774)

Gateway creates a fresh AIAgent per inbound message in several common
scenarios: cache miss, idle eviction (1h TTL), config-signature
mismatch, process restart. A freshly-built AIAgent has
_turns_since_memory=0 and _user_turn_count=0, so the
memory.nudge_interval trigger ('_turns_since_memory >=
_memory_nudge_interval') can never be reached when these reconstructions
happen on roughly the cadence of the interval. A user can chat for hours
on Telegram without ever seeing a self-improvement review fire.

Reconstruct the counters from conversation_history at the top of
run_conversation(), right after the existing _hydrate_todo_store call.
Idempotent guard ('if self._user_turn_count == 0') means a cached agent
that already accumulated counters keeps them; only freshly-built agents
hydrate. Modulo arithmetic preserves the original 1-in-N cadence rather
than firing a review immediately on resume.

7 regression tests pinning the contract (mid-cycle history, modulo wrap,
idempotency, zero-interval skip, role==user filtering, production-code
anchor).

Closes #22357.

* fix(api-server): emit length/error finish_reason for truncation/failure (#22775)

Non-streaming /v1/chat/completions wrapped any AIAgent result \u2014 including
partial/failed runs \u2014 as a successful 200 with finish_reason='stop' and
the internal failure string substituted into message.content. API
clients had no way to distinguish 'agent answered: X' from
'agent crashed and the X you see is its error message'.

After the fix:
  - completed: True             \u2192 200 finish_reason='stop' (unchanged)
  - partial + truncated text    \u2192 200 finish_reason='length' + hermes extras
  - partial + no text / failed  \u2192 502 OpenAI error envelope (SDKs raise)
  - other failures              \u2192 200 finish_reason='error' + hermes extras

Adds X-Hermes-Completed / X-Hermes-Partial / X-Hermes-Error headers
plus a 'hermes' extras object on partial responses for clients that
want the full picture.

Closes #22496.

* fix(cli): make Ctrl+Enter insert newline on WSL/SSH/Windows Terminal (#22777)

Native Windows, WSL, SSH sessions, and Windows Terminal all send
Ctrl+Enter as bare LF (c-j). Hermes was binding c-j as submit on
every POSIX platform, so Ctrl+Enter submitted instead of inserting
a newline on those terminals. Reported in #22379.

Add _preserve_ctrl_enter_newline() predicate that detects the
environments where Ctrl+Enter must produce a newline (sys.platform
== 'win32', SSH_CONNECTION/SSH_CLIENT/SSH_TTY env, WT_SESSION,
WSL_DISTRO_NAME, /proc/version 'microsoft' marker). Gate the
c-j-as-submit binding off in those environments and gate the
c-j-as-newline handler on. Local POSIX TTYs without those markers
(docker exec, plain ssh from a Mac) keep c-j as submit so plain
Enter still works on thin PTYs.

Add install_ctrl_enter_alias() in hermes_cli/pt_input_extras.py
mapping the three CSI-u / modifyOtherKeys variants of Ctrl+Enter
('\x1b[13;5u', '\x1b[27;5;13~', '\x1b[27;5;13u') to the
(Escape, ControlM) tuple Alt+Enter produces. This lets Kitty /
mintty / xterm-with-modifyOtherKeys users over SSH get a Ctrl+Enter
newline through the existing Alt+Enter handler.

9 new tests + extended existing test_lf_enter_binds_to_submit_handler_posix
to cover bare-local vs SSH branches.

Closes #22379.

* fix(fallback): skip chain entries matching current provider/model/base_url (#22780)

_try_activate_fallback() walked the chain by index without comparing
the candidate entry against the currently-failing backend. So a
misconfigured chain that listed the same provider+model as the primary,
or two custom_providers entries pointing at the same shim URL, would
loop the same failure 3x for the same backend.

After the fix, advance() skips:
  - entries where (provider, model) match the current agent's
  - entries with a base_url + model matching the current backend
    (catches two custom_providers names pointing at the same shim)

Recursing through self._try_activate_fallback() continues to the next
chain entry; if everything matches, returns False and the caller
moves on without retrying the same broken path.

3 regression tests covering same-provider-same-model skip, same-base_url-
same-model skip, and the all-self-matching-returns-False exhaustion path.

Closes #22548 (the Hermes-side portion). The 120s timeout itself in
the downstream claude-cli shim is a deployment concern documented in
that issue's wherewolf87 comment.

* fix(memory): tighten MEMORY_GUIDANCE against ephemeral PR/issue/SHA notes (#22781)

The model regularly writes session-outcome facts to MEMORY.md despite
the existing 'Do NOT save task progress' line — entries like
'Submitted PR #22577 for the kanban dedup fix' or 'Fixed bug X in
file Y'. These are stale within days, pollute the system prompt,
and crowd out durable user preferences (the issue #22563 reporter
saw 9 sections of bug-fix notes injected on a brand-new task).

Add explicit examples of what NOT to save (PR numbers, issue
numbers, commit SHAs, 'fixed/submitted/Phase N done', file counts)
plus the 7-day-staleness heuristic so the model has a concrete
calibration target rather than guessing what counts as 'task progress'.

Closes #22563 (the prompt-side, low-risk portion). The bigger
relevance-based-injection / vector-retrieval feature requested in
#22563 is tracked under #2184 (Richer local memory). Per skill rule
on prompt caching, dynamic memory injection breaks the frozen-snapshot
invariant and needs a separate design call.

* fix(tools): install cua-driver when Computer Use is enabled via 'hermes tools' (#22765)

Returning users who enabled '🖱️ Computer Use (macOS)' via 'hermes tools'
saw '✓ Saved configuration' but no install — cua-driver was never on
PATH and the toolset failed at first use. Two compounding causes:

1. _toolset_needs_configuration_prompt fell through to _toolset_has_keys,
   which returned True for any provider with empty env_vars. cua-driver
   has no env vars, so the gate skipped _configure_toolset entirely and
   _run_post_setup('cua_driver') never ran.

2. No stable CLI entry-point existed for re-running the install when
   the picker no-op'd it (e.g. when toggling the toolset off+on inside
   one picker session, where 'added' is empty).

Changes:

- hermes_cli/tools_config.py: add _POST_SETUP_INSTALLED registry
  mapping post_setup keys to installed-state predicates. The gate
  now returns True when any visible provider has a registered
  post_setup whose predicate fails. cua_driver is the only opt-in
  for now; other post_setup hooks keep their existing behaviour.
- hermes_cli/main.py: add 'hermes computer-use install' and
  'hermes computer-use status' as a stable docs target. install
  reuses the same _run_post_setup('cua_driver') path that the
  picker invokes; status reports whether cua-driver is on PATH.
- tools/computer_use/cua_backend.py: install hint now points users
  at 'hermes computer-use install' first.
- website/docs/user-guide/features/computer-use.md: document the
  new command as the primary install path.
- website/docs/reference/cli-commands.md: catalog 'hermes
  computer-use' alongside 'hermes tools'.
- tests/hermes_cli/test_post_setup_gating.py: regression coverage
  for the gate predicate (missing -> setup forced, installed ->
  setup skipped, broken predicate -> non-blocking, unregistered
  keys -> behaviour unchanged).

Fixes #22737. Reported by @f-trycua.

* perf(doctor): parallelize API connectivity checks and disable IMDS (#22766)

`hermes doctor` ran every connectivity probe sequentially and on a typical
developer laptop spent ~2s of its ~5s wall time inside boto3's EC2
instance-metadata-service lookup (169.254.169.254) — the default
AWS credential chain probes IMDS even when AWS_BEARER_TOKEN_BEDROCK
or AWS_ACCESS_KEY_ID is the only legitimate source.

Refactor the API Connectivity section so every probe (OpenRouter,
Anthropic, ~16 static API-key providers + dynamic profiles, AWS
Bedrock) is a pure function returning a structured result, then
fan them out through a ThreadPoolExecutor(max_workers=8). Output
order, glyphs, colours, padding, and issue strings stay byte-for-byte
identical to the sequential implementation; results are gathered
in submission order.

Also disable IMDS for the parallel block by setting
AWS_EC2_METADATA_DISABLED=true on the parent thread before submitting
work (and restoring its prior value in a finally block). Bedrock's
real-API call gets a Config(connect_timeout=5, read_timeout=10,
retries={max_attempts:1}) so a transient regional failure can't pad
the run by 30+ seconds.

Measured impact (5-run medians, 9950X3D):
  hermes doctor:           5.07 → 2.16 s  (-57%)

Doctor tests: 48 passed (test_doctor.py + test_doctor_command_install.py).

The remaining ~2s of wall is import overhead + a couple of one-off
network calls outside the API Connectivity section (`fetch_models_dev`
provider catalog refresh, Nous OAuth refresh in `Auth Providers`).
Those are next-tier targets, not part of this change.

* fix: send correct resolution param to xAI image generation API

The xAI /v1/images/generations endpoint expects resolution as a
literal string ('1k' or '2k'), not the numeric value ('1024').

- Change _XAI_RESOLUTIONS from a dict mapping to a validation set
- Use the resolution key directly instead of the mapped value
- Fall back to DEFAULT_RESOLUTION on invalid config values

Fixes 422 Unprocessable Entity errors when resolution was sent.

* chore: add A-kamal to AUTHOR_MAP for PR #18678

* test(xai-image): regression-guard literal '1k'/'2k' resolution payload

The xAI image-gen provider was DOA from PR #14765 onward — every request
422'd because the resolution param was being mapped to '1024'/'2048' but
xAI's API expects the literal strings '1k'/'2k'. PR #18678 fixed the
mapping; this test asserts the wire payload carries the literal so the
regression cannot recur silently.

* perf(gateway): defer QQAdapter and YuanbaoAdapter imports via PEP 562 (#22790)

`gateway/platforms/__init__.py` eagerly imported `QQAdapter` and
`YuanbaoAdapter` at package-init time, which transitively pulled in
qqbot's chunked-upload + keyboards + onboard machinery and yuanbao's
websocket stack. About 84 ms wall and 23 MB RSS on every fresh process
that touched anything under `gateway.platforms` — including `hermes
chat` (via run_agent → cli's plugin discovery transitive import).

Nothing in the codebase actually consumes these symbols from the
package root; every real call site uses the long-form path
(`from gateway.platforms.qqbot import QQAdapter`,
`from gateway.platforms.yuanbao import YuanbaoAdapter` in gateway/run.py).
The eager re-export was only there for convenience.

Replace with a PEP 562 module-level `__getattr__` that lazily imports
on first attribute access. Public API stays identical:
`from gateway.platforms import QQAdapter` keeps working but only
pays the import cost when the symbol is actually touched. `__dir__`
preserves help() / autocomplete behavior.

Measured impact (7-run medians, 9950X3D):
  import gateway.platforms        127 →  43 ms  (-66%)
                                   50 →  27 MB  (-46%)
  import gateway.platforms.base   127 →  44 ms  (-65%)
                                   50 →  27 MB  (-46%)
  import cli (full chat path)     745 → 710 ms  ( -5%)
                                   96 →  90 MB  ( -6%)
  hermes chat -q (cold)                  -5 MB

The per-import win is biggest because qqbot/yuanbao deps don't overlap
with anything on the gateway-platforms path — full `import cli`
already loads aiohttp/websockets transitively from other places, so
the marginal CLI win is smaller than the isolated import benchmark.
The `gateway.platforms.base` win is what matters most for long-lived
gateway processes: every gateway boot saves 23 MB resident.

All 144 qqbot tests pass; broader gateway suite (5132 tests) passes
modulo 4 pre-existing flakes also failing on main without this change.

* docs: deep audit — fix stale config keys, missing commands, and registry drift (#22784)

* docs: deep audit — fix stale config keys, missing commands, and registry drift

Cross-checked ~80 high-impact docs pages (getting-started, reference, top-level
user-guide, user-guide/features) against the live registries:

  hermes_cli/commands.py    COMMAND_REGISTRY (slash commands)
  hermes_cli/auth.py        PROVIDER_REGISTRY (providers)
  hermes_cli/config.py      DEFAULT_CONFIG (config keys)
  toolsets.py               TOOLSETS (toolsets)
  tools/registry.py         get_all_tool_names() (tools)
  python -m hermes_cli.main <subcmd> --help (CLI args)

reference/
- cli-commands.md: drop duplicate hermes fallback row + duplicate section,
  add stepfun/lmstudio to --provider enum, expand auth/mcp/curator subcommand
  lists to match --help output (status/logout/spotify, login, archive/prune/
  list-archived).
- slash-commands.md: add missing /sessions and /reload-skills entries +
  correct the cross-platform Notes line.
- tools-reference.md: drop bogus '68 tools' headline, drop fictional
  'browser-cdp toolset' (these tools live in 'browser' and are runtime-gated),
  add missing 'kanban' and 'video' toolset sections, fix MCP example to use
  the real mcp_<server>_<tool> prefix.
- toolsets-reference.md: list browser_cdp/browser_dialog inside the 'browser'
  row, add missing 'kanban' and 'video' toolset rows, drop the stale
  '38 tools' count for hermes-cli.
- profile-commands.md: add missing install/update/info subcommands, document
  fish completion.
- environment-variables.md: dedupe GMI_API_KEY/GMI_BASE_URL rows (kept the
  one with the correct gmi-serving.com default).
- faq.md: Anthropic/Google/OpenAI examples — direct providers exist (not just
  via OpenRouter), refresh the OpenAI model list.

getting-started/
- installation.md: PortableGit (not MinGit) is what the Windows installer
  fetches; document the 32-bit MinGit fallback.
- installation.md / termux.md: installer prefers .[termux-all] then falls
  back to .[termux].
- nix-setup.md: Python 3.12 (not 3.11), Node.js 22 (not 20); fix invalid
  'nix flake update --flake' invocation.
- updating.md: 'hermes backup restore --state pre-update' doesn't exist —
  point at the snapshot/quick-snapshot flow; correct config key
  'updates.pre_update_backup' (was 'update.backup').

user-guide/
- configuration.md: api_max_retries default 3 (not 2); display.runtime_footer
  is the real key (not display.runtime_metadata_footer); checkpoints defaults
  enabled=false / max_snapshots=20 (not true / 50).
- configuring-models.md: 'hermes model list' / 'hermes model set ...' don't
  exist — hermes model is interactive only.
- tui.md: busy_indicator -> tui_status_indicator with values
  kaomoji|emoji|unicode|ascii (not kawaii|minimal|dots|wings|none).
- security.md: SSH backend keys (TERMINAL_SSH_HOST/USER/KEY) live in .env,
  not config.yaml.
- windows-wsl-quickstart.md: there is no 'hermes api' subcommand — the
  OpenAI-compatible API server runs inside hermes gateway.

user-guide/features/
- computer-use.md: approvals.mode (not security.approval_level); fix broken
  ./browser-use.md link to ./browser.md.
- fallback-providers.md: top-level fallback_providers (not
  model.fallback_providers); the picker is subcommand-based, not modal.
- api-server.md: API_SERVER_* are env vars — write to per-profile .env,
  not 'hermes config set' which targets YAML.
- web-search.md: drop web_crawl as a registered tool (it isn't); deep-crawl
  modes are exposed through web_extract.
- kanban.md: failure_limit default is 2, not '~5'.
- plugins.md: drop hard-coded '33 providers' count.
- honcho.md: fix unclosed quote in echo HONCHO_API_KEY snippet; document
  that 'hermes honcho' subcommand is gated on memory.provider=honcho;
  reconcile subcommand list with actual --help output.
- memory-providers.md: legacy 'hermes honcho setup' redirect documented.

Verified via 'npm run build' — site builds cleanly; broken-link count went
from 149 to 146 (no regressions, fixed a few in passing).

* docs: round 2 audit fixes + regenerate skill catalogs

Follow-up to the previous commit on this branch:

Round 2 manual fixes:
- quickstart.md: KIMI_CODING_API_KEY mentioned alongside KIMI_API_KEY;
  voice-mode and ACP install commands rewritten — bare 'pip install ...'
  doesn't work for curl-installed setups (no pip on PATH, not in repo
  dir); replaced with 'cd ~/.hermes/hermes-agent && uv pip install -e
  ".[voice]"'. ACP already ships in [all] so the curl install includes it.
- cli.md / configuration.md: 'auxiliary.compression.model' shown as
  'google/gemini-3-flash-preview' (the doc's own claimed default);
  actual default is empty (= use main model). Reworded as 'leave empty
  (default) or pin a cheap model'.
- built-in-plugins.md: added the bundled 'kanban/dashboard' plugin row
  that was missing from the table.

Regenerated skill catalogs:
- ran website/scripts/generate-skill-docs.py to refresh all 163 per-skill
  pages and both reference catalogs (skills-catalog.md,
  optional-skills-catalog.md). This adds the entries that were genuinely
  missing — productivity/teams-meeting-pipeline (bundled),
  optional/finance/* (entire category — 7 skills:
  3-statement-model, comps-analysis, dcf-model, excel-author, lbo-model,
  merger-model, pptx-author), creative/hyperframes,
  creative/kanban-video-orchestrator, devops/watchers,
  productivity/shop-app, research/searxng-search,
  apple/macos-computer-use — and rewrites every other per-skill page from
  the current SKILL.md. Most diffs are tiny (one line of refreshed
  metadata).

Validation:
- 'npm run build' succeeded.
- Broken-link count moved 146 -> 155 — the +9 are zh-Hans translation
  shells that lag every newly-added skill page (pre-existing pattern).
  No regressions on any en/ page.

* feat(transports/codex): pass reasoning.effort to xAI Responses API

The is_xai_responses branch only sent include=[reasoning.encrypted_content]
without forwarding the resolved reasoning_effort. Other Responses providers
(OpenAI, GitHub) already get effort forwarded — this aligns the xAI path.

Without this, agent.reasoning_effort is silently dropped on the xAI direct
path, making Hermes unable to control reasoning depth on grok-4.x via
api.x.ai. Tests added to TestCodexBuildKwargs cover effort passthrough,
disabled state, and minimal-clamp parity with non-xAI.

* perf(models_dev): cache-first lookup, skip network when disk cache is fresh (#22808)

`fetch_models_dev()` is on the hot path of every `AIAgent.__init__`
(via `context_compressor → get_model_context_length`). The previous
policy was "always try network first, only fall back to disk if
network fails," so every fresh `hermes chat` / `hermes gateway` /
batch / cron process paid 250-500 ms re-fetching a 2 MB JSON registry
that was already on disk from earlier runs.

Add a stage 2 between in-mem and network: if
`models_dev_cache.json` exists and its mtime is younger than the
existing `_MODELS_DEV_CACHE_TTL` (1 hour, same TTL the in-mem cache
already uses), load from disk and skip the network call.

The in-mem TTL is anchored to the disk file's age, so a 50-min-old
cache stays in-memory for only 10 more minutes — no surprise
extension of staleness window.

Invariants preserved:
- `force_refresh=True` still always hits the network and only falls
  back to disk on failure (`hermes config refresh` semantics).
- Missing disk cache → fall through to network (first-ever run).
- Stale disk cache (mtime > TTL) → fall through to network.
- Negative file age (clock skew) → fall through to network.
- Network failure → existing stage-4 stale-disk fallback unchanged.

Measured impact (3-run medians, 9950X3D, fresh process per run):
  fetch_models_dev cold:  256 → 17 ms  (-93%)
  hermes chat -q wall:   4.00 → 3.73 s (-7% median)
                         3.99 → 3.60 s (-10% min)

The chat-end-to-end win is bounded below by API latency variance, but
the fetch_models_dev microbenchmark is the cleanest signal: 239 ms
shaved off every fresh-process agent construction.

Win compounds with the previous perf PRs:
  #22681 google_chat lazy-load
  #22766 doctor parallel + IMDS off
  #22790 gateway.platforms PEP 562

Tests: all 30 `tests/agent/test_models_dev.py` pass (added 4 new ones
covering the new disk-cache-first path, force_refresh override, stale
disk fallback, and missing-disk-cache fall-through). Full `tests/agent/`
suite: 2560 passed, 0 failed.

* fix(auxiliary): rotate pooled auth after quota failures

* chore: add Qwinty to AUTHOR_MAP

* fix(browser_tool): do not cache transient None cloud provider resolution

Problem: `_get_cloud_provider()` set `_cloud_provider_resolved = True`
before resolution. If credentials were briefly unavailable on the first
call (e.g. a managed Nous Portal token mid-refresh), the resolver pinned
the entire process to local mode forever, even after credentials
self-healed seconds later.

Root cause: bookkeeping was set up-front, so any code path that fell
through to `return _cached_cloud_provider` (config read failure, no
credentials yet, explicit-provider instantiation failure) committed the
transient `None` to the cache permanently.

Fix: invert the bookkeeping. `_cloud_provider_resolved = True` is now
set only when (a) the user explicitly chose `cloud_provider: local`, or
(b) a provider was successfully resolved. All transient `None` paths
return without poisoning the cache, so the next call retries. Explicit
provider instantiation failures now log at warning level with stack
trace so operators can diagnose them.

Tests: 5 new cases in tests/tools/test_browser_cloud_provider_cache.py
covering explicit local, successful resolution, no-credentials-yet,
config read failure, and explicit provider instantiation failure.
Stash-verify confirmed the 3 transient-None tests fail without the fix.
All 320 existing browser tests still green.

Closes #22324

* fix(browser_tool): fall through to autodetect on config read failure

* fix(email): send IMAP ID extension to support 163/NetEase mailbox

163/NetEase IMAP servers reject every UID SEARCH/FETCH with `BYE Unsafe
Login` unless the client first identifies itself via the RFC 2971 ID
command after LOGIN.  Without this, the email gateway logs in OK but
then fails on the very first poll and the connection is torn down.

Send the ID payload best-effort after both `imap.login()` sites
(`EmailAdapter.connect` and `_fetch_new_messages`).  Failures are
swallowed at debug level so non-supporting IMAP servers (Gmail,
Outlook, Fastmail, Yahoo, etc.) keep working unchanged.

Closes #22271

* fix(email): use real hermes version in IMAP ID command

* fix(deps): declare youtube-transcript-api in pyproject.toml [youtube] extra

skills/media/youtube-content/scripts/fetch_transcript.py and
optional-skills/productivity/memento-flashcards/scripts/youtube_quiz.py
both import youtube-transcript-api at runtime, but the package was not
listed in pyproject.toml.  A fresh `uv sync` therefore omits it, and
both skills fail on first invocation with:

    ModuleNotFoundError: No module named 'youtube_transcript_api'

Add a new [youtube] optional-dependency group with
youtube-transcript-api>=1.2.0 (the v1.x API surface the scripts already
use) and include it in [all] so standard installs pick it up.

Regression tests: TestPyprojectDeclaresYoutubeExtra verifies the extra
is present in pyproject.toml and included in [all].

Closes #22243

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* fix(agent): extract thinking from content-list blocks for DeepSeek V4 Pro

DeepSeek V4 Pro returns thinking content as typed blocks inside the
content array rather than as a top-level reasoning_content field:

  [{"type": "thinking", "thinking": "..."}, {"type": "output", ...}]

_extract_reasoning only handled content as a plain string, so the
thinking text was silently dropped.  On the next turn the session was
replayed without the thinking block, causing:

  HTTP 400: The content[].thinking in the thinking mode must be
  passed back to the API.

Fix: when content is a list and no structured reasoning field was
found, scan for items with type=='thinking' and accumulate their
'thinking' (or 'text') value into reasoning_parts.  Structured fields
(reasoning, reasoning_content, reasoning_details) still take priority
so existing provider behaviour is unchanged.

Closes #21944

* fix(kanban): make _migrate_add_optional_columns idempotent on concurrent open

ALTER TABLE calls inside _migrate_add_optional_columns were guarded by a
snapshot of PRAGMA table_info taken at function entry.  When the gateway
dispatcher opens the kanban DB twice per tick (once in _tick_once_for_board
and once via init_db's discard-and-reconnect path), a second connection can
run the same migration before the first one commits, causing:

  sqlite3.OperationalError: duplicate column name: consecutive_failures

This crashed the dispatcher on every first tick after a gateway restart
(subsequent ticks succeeded because the columns were then present).

Fix: introduce _add_column_if_missing() which wraps ALTER TABLE in a
try/except that swallows OperationalError whose message contains
'duplicate column name'.  All ALTER TABLE calls in
_migrate_add_optional_columns are routed through this helper.

Closes #21708

* fix(doctor): skip pluggable provider profiles when a dedicated check exists (#22346)

Problem
-------
`hermes doctor` ran two health checks for Anthropic: a dedicated one
with the correct `x-api-key` + `anthropic-version` headers, and a
generic Bearer-auth one driven by the pluggable `ProviderProfile` for
"anthropic". The generic check called `https://api.anthropic.com/v1/models`
with `Authorization: Bearer ...`, which Anthropic answers with HTTP 404,
producing a noisy duplicate warning even when the dedicated check passed.

Root cause
----------
`hermes_cli/doctor.py:_build_apikey_providers_list` deduplicated profiles
against a `_known_canonical` set built from the static list (Z.AI/GLM,
Kimi, DeepSeek, …). Providers with their own dedicated check above the
generic loop (Anthropic, OpenRouter, Bedrock) were not in that set, so
their profiles were appended and ran a second, broken check.

Fix
---
Add `{"anthropic", "openrouter", "bedrock"}` to the skip set, and
also skip profiles whose aliases match any of those names (e.g.
`claude`, `claude-oauth` → anthropic).

Tests
-----
tests/hermes_cli/test_doctor_dedicated_provider_skip.py:
  - test_build_apikey_providers_list_skips_dedicated_check_providers:
    asserts the assembled list does not contain anthropic, openrouter,
    or bedrock entries.
  - test_build_apikey_providers_list_includes_non_dedicated_providers:
    sanity guard that legitimate providers (DeepSeek, Z.AI/GLM) survive.
Both confirmed via stash-verify (fail pre-fix with anthropic/openrouter
leaking, pass post-fix).

Fixes #22346

* fix(doctor): normalize provider name and aliases before dedicated-skip check

* fix(completion): use valid zsh _arguments exclusion-group syntax

The generated zsh completion script used `(-h --help)` as the exclusion
group for `_arguments`, which zsh rejects with:

  _arguments:comparguments: invalid argument: (-h --help){-h,--help}[...]

Exclusion groups in `_arguments` cannot contain long options. Use the
canonical `(-)` form (exclude all other options) which correctly
handles flag pairs like `-h`/`--help`.

Fixes NousResearch/hermes-agent#22686

* fix(model-metadata): align hy3-preview static fallback + delete change-detector test (#22805)

Two co-located fixes:

1. agent/model_metadata.py: bump hy3-preview static fallback from
   256000 to 262144 (256 * 1024) to match OpenRouter live metadata
   so cache and offline both agree (issue #22268).

2. tests/hermes_cli/test_tencent_tokenhub_provider.py: replace the
   exact-value change-detector (assert ctx == 256000) with an
   invariant assertion (registered + >= 4096). Per AGENTS.md
   'Don't write change-detector tests': pinning the upstream-controlled
   context length is exactly the test class the rule forbids — it
   breaks every time the provider bumps the published value, with
   zero behavioral coverage gained.

Salvage of #22574 with a redirect on the test approach. The
contributor's diff bumped the integer and added a SECOND
change-detector pinning DEFAULT_CONTEXT_LENGTHS[hy3-preview] == 262144,
which would re-break on the next published bump. We instead delete
the change-detector entirely and assert the relationship.

Closes #22268.

* fix(delegate): add explicit do-not-use guidance to acp_command/acp_args schema (carve-out of #22680)

acp_command / acp_args descriptions previously primed the model to
populate them — "Per-task ACP command override (e.g. 'copilot')" —
even when no ACP CLI was installed. Models with weaker schema-following
discipline would set them and the spawn would fail.

Add explicit "Do NOT set unless the user has explicitly told you"
guidance at both the top-level acp_command and the per-task override.
Strengthen acp_args to mention it's empty unless acp_command is set.
Adds 2 tests pinning the descriptions.

Note: this is a cosmetic prompt-engineering fix — the params remain
exposed in the schema. The fully-correct fix is to gate them behind
a config flag or runtime ACP-CLI detection so the schema only emits
them when an ACP harness is available. Tracked as a follow-up; this
PR ships the low-cost stopgap.

Salvage of #22680 (delegate schema only). The original PR also
bundled unrelated fixes for #22548, #21944, #22150 — those
need separate PRs since #22548 and #21944 are already addressed
on main (#22780 + #22798 in flight) and #22150 deserves its own
review.

Closes #22013.

* feat(gateway): add Telegram notification mode to suppress intermediate push notifications

Add a configurable notifications mode for the Telegram platform adapter
that controls which messages trigger push notifications.

- display.platforms.telegram.notifications: "all" (default) | "important"
- HERMES_TELEGRAM_NOTIFICATIONS env var override
- In "important" mode, all sends use disable_notification=True except:
  - Approvals (send_exec_approval) and slash confirmations
  - Final response messages (metadata["notify"]=True)
- Zero overhead in default "all" mode
- Zero impact on non-Telegram platforms

Closes #22771

* chore: add CalmProton to AUTHOR_MAP

* fix(telegram): default notifications to 'important' (silence intermediate)

Per-tool-call push notifications on Telegram are noisy enough that
'all' is the wrong default — long agent runs spam the user's notification
shade with status messages they didn't ask to be pinged about. Final
responses, approval prompts, and slash confirmations still notify;
intermediate progress, streaming, and tool-progress messages now
deliver silently via disable_notification.

Users who want the legacy behavior can opt back in with:
  display:
    platforms:
      telegram:
        notifications: all
or HERMES_TELEGRAM_NOTIFICATIONS=all.

* fix(gateway): adopt unit's HERMES_HOME for --system CLI ops

When systemd_restart / systemd_status / systemd_stop run under sudo,
HERMES_HOME is stripped and HOME=/root, so get_hermes_home() resolves
to /root/.hermes instead of the unit's pinned home. read_runtime_status
and get_running_pid then look at the wrong gateway_state.json — the
60s status poll never sees "running", times out, and forces another
systemctl restart that SIGTERMs the in-progress new gateway.

Read the unit's pinned HERMES_HOME from `systemctl show -p Environment`
and mirror it into os.environ before any HERMES_HOME-derived read.
Early-out when system=False (user-scope inherits naturally). Errors
swallowed so a transient systemctl failure doesn't break unrelated
CLI ops.

Closes #22035.

* chore: add mbac to AUTHOR_MAP

* fix(openrouter): add x-grok-conv-id header for Grok models to improve prompt cache hit rates (carve-out of #22708)

Pass session_id through to provider profile build_api_kwargs_extras so
the OpenRouter profile can attach an xAI cache-affinity header
(x-grok-conv-id: <session-id>) for x-ai/grok-* models. xAI prompt
cache requires server affinity via this header — without it the cache
is poisoned and Grok prompt-cache hit rates drop dramatically on
multi-turn sessions.

Carve-out of #22708 by Ninso112. The original PR bundled a /diff
slash command, a zsh completion fix (already on main via #22802),
and holographic memory null-guards. This salvage keeps just the
Grok header work — small, targeted, and well-tested. Other
contributors and changes preserved for separate review.

Closes #22705.

* chore: add Ninso112 to AUTHOR_MAP

* fix(gateway): preserve Ctrl+C for Windows foreground runs

* fix: make session search initialize session db

* perf(teams): defer httpx import to first webhook call (#22831)

Same pattern as the google_chat lazy-load (PR #22681), applied to the
Teams plugin. The bundled `plugins/platforms/teams/adapter.py` did
`import httpx` at module top, which dragged the entire httpx +
httpcore stack into every process that triggered plugin discovery —
including `hermes` invocations that never instantiate the Teams
adapter.

`httpx` is only needed inside one method
(`TeamsMeetingPipeline._write_summary_via_incoming_webhook`), and the
`httpx.AsyncBaseTransport` parameter annotation is already string-only
thanks to the existing `from __future__ import annotations`. Move the
runtime import inside the method.

Measured impact (7-run medians, 9950X3D):
  teams plugin alone:    118 → 89 ms  (-25%)
                         46 → 38 MB   (-17%)
  import cli (full):     unchanged
  import model_tools:    unchanged

The full-CLI numbers are flat because httpx is loaded transitively
from many other modules on that path. The microbench win is the real
signal: 29 ms / 8 MB shaved off any process that touches the teams
plugin without otherwise pulling httpx — primarily future workflows
where the gateway is enabled but Teams is not configured.

Tests: 44/44 `tests/gateway/test_teams.py` pass; 345 across all
plugin-platform suites (teams + qqbot + google_chat). The test file
imports `httpx` itself for the `MockTransport` fixture, which is
correct — tests legitimately use httpx, only the plugin's module-level
import was the issue.

* fix(acp): honor task cwd for foreground terminal commands

* feat(openrouter): wire Pareto Code router with min_coding_score knob (#22838)

Pick openrouter/pareto-code as your model and OpenRouter auto-routes each
request to the cheapest model meeting your coding-quality bar (ranked by
Artificial Analysis). The new openrouter.min_coding_score config key (0.0-1.0,
default 0.65) tunes the floor.

- hermes_cli/models.py: add openrouter/pareto-code to OPENROUTER_MODELS so
  it shows up in the picker with a description
- hermes_cli/config.py: add openrouter.min_coding_score (default 0.65 — lands
  on a mid-tier coder on the current Pareto frontier)
- plugins/model-providers/openrouter: emit extra_body.plugins =
  [{id: pareto-router, min_coding_score: X}] when model is openrouter/pareto-code
  AND the score is a valid float in [0.0, 1.0]
- agent/transports/chat_completions.py: same emission on the legacy flag
  path (when no provider profile is loaded)
- run_agent.py: openrouter_min_coding_score kwarg + storage; plumbed into
  both build_kwargs() invocations and the context-summary extra_body path
- cli.py: read openrouter.min_coding_score once at init, validate float in
  [0,1], pass to AIAgent constructions (CLI + background-task paths)
- cron/scheduler.py, batch_runner.py, tools/delegate_tool.py,
  tui_gateway/server.py: propagate the kwarg (mirrors providers_order
  plumbing — subagents inherit, cron/batch read from config)
- tests: profile-level + transport-level coverage of the model gating,
  unset/empty/out-of-range handling, and the legacy flag path
- docs: new 'OpenRouter Pareto Code Router' section in providers.md

Verified end-to-end against api.openrouter.ai: at score=0.65 we land on a
mid-tier coder, at omission we get the strongest. Score is silently dropped
on any model other than openrouter/pareto-code, so it's safe to leave set.

* fix(gateway): preserve reasoning_content, codex_message_items, finish_reason on transcript replay (#22839)

PR #2974 whitelisted three reasoning fields (reasoning, reasoning_details,
codex_reasoning_items) for the gateway's simple-text replay branch. Three
more fields were added to the DB later but the whitelist was never updated:

  - reasoning_content: provider-facing thinking text. _copy_reasoning_content_for_api
    promotes 'reasoning' -> 'reasoning_content' at send time only when the
    strings happen to match. Carrying the original verbatim avoids loss
    for providers that return them as distinct fields (DeepSeek/Kimi/
    Moonshot thinking modes), and preserves the empty-string sentinel
    that DeepSeek V4 Pro requires for thinking-mode replay.
  - codex_message_items: exact assistant message items with 'phase'.
    OpenAI docs: 'preserve and resend phase on all assistant messages —
    dropping it can degrade performance.' Required for prefix cache hits.
    No recovery path exists — once dropped, gone.
  - finish_reason: informational; cheap to keep so transcripts replay
    identically across CLI and gateway.

The CLI is unaffected because cli.py keeps the live in-memory message list
across turns (cli.py:10046 'self.conversation_history = result["messages"]').
The gateway rebuilds agent_history from the SQLite transcript on every turn,
so any field stripped during replay is silently lost.

Refactors the inline whitelist into a module-level _build_replay_entry()
helper so the contract can be unit-tested. 16 new tests pin the field set
and falsy-value handling.

Verified end-to-end: DB stores all 8 fields, replay now preserves all 8
(was preserving only 5 for assistant text turns).

* docs(openrouter): document auxiliary.<task>.extra_body for OR routing and Pareto (#22844)

The plumbing for setting OpenRouter provider preferences and the Pareto Code
router on auxiliary tasks already exists — auxiliary.<task>.extra_body is
forwarded verbatim by call_llm() / async_call_llm(). It just wasn't documented,
so users who wanted (e.g.) Pareto Code routing for compression but the strongest
coder for the main agent had no way to discover the escape hatch.

- hermes_cli/config.py: expand the auxiliary section header with a YAML
  example showing provider routing plus plugins under extra_body, and an
  explicit note that main-agent provider_routing / openrouter.min_coding_score
  do NOT propagate to aux calls (each task is independent by design)
- website/docs/user-guide/configuration.md: new 'OpenRouter routing and
  Pareto Code for auxiliary tasks' subsection with worked example
- website/docs/integrations/providers.md: cross-link from the Pareto Code
  Router section to the aux-side doc

E2E verified that auxiliary.<task>.extra_body reaches the OpenRouter API with
the configured provider routing and plugins blocks intact.

* docs: round 2 audit — messaging, developer-guide, guides, integrations (#22858)

Cross-checked 75 docs pages under user-guide/messaging/, developer-guide/,
guides/, and integrations/ against the live registries and gateway code.

messaging/
- index.md: API Server toolset is hermes-api-server (was 'hermes (default)');
  Google Chat slug is hermes-google_chat (underscore — plugin name uses _).
- google_chat.md: drop bogus 'pip install hermes-agent[google…

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR #22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

…backends (#25126)

* feat(video_gen): unified video_generate tool with pluggable provider backends

One core video_generate tool, every backend a plugin. Mirrors the
image_gen + memory_provider + context_engine architecture: ABC, registry,
plugin-context registration hook, and per-plugin model catalogs surfaced
through hermes tools.

Surface (one schema, every backend):
- operation: generate / edit / extend
- modalities: text-to-video (prompt only), image-to-video (prompt +
  image_url), video edit (prompt + video_url), video extend (video_url)
- reference_image_urls, duration, aspect_ratio, resolution,
  negative_prompt, audio, seed, model override
- Providers ignore unknown kwargs and declare what they support via
  VideoGenProvider.capabilities() — backend-specific quirks stay in the
  backend, the agent learns one tool

Backends shipped:
- plugins/video_gen/xai/  — Grok-Imagine, full generate/edit/extend +
  image-to-video + reference images (salvaged from PR #10600 by
  @Jaaneek, reshaped into the plugin interface)
- plugins/video_gen/fal/  — Veo 3.1 (t2v + i2v), Kling O3 i2v,
  Pixverse v6 i2v with model-aware payload building that drops keys a
  model doesn't declare

Wiring:
- agent/video_gen_provider.py — VideoGenProvider ABC, normalize_operation,
  success_response / error_response, save_b64_video / save_bytes_video,
  $HERMES_HOME/cache/videos/
- agent/video_gen_registry.py — thread-safe register/get/list +
  get_active_provider() reading video_gen.provider from config.yaml
- hermes_cli/plugins.py — PluginContext.register_video_gen_provider()
- hermes_cli/tools_config.py — Video Generation category in
  hermes tools, plugin-only providers list, model picker per plugin,
  config write to video_gen.{provider,model}
- toolsets.py — new video_gen toolset
- tests: 31 new tests covering ABC, registry, tool dispatch, both plugins
- docs: developer-guide/video-gen-provider-plugin.md (parallel to the
  image-gen guide), sidebar + toolsets-reference + plugin guides updated

Supersedes: #25035 (FAL), #17972 (FAL), #14543 (xAI), #13847 (HappyHorse),
#10458 (provider categories), #10786 (xAI media+search bundle), #2984
(FAL duplicate), #19086 (Google Veo standalone — easy port to plugin
interface).

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): dynamic schema reflects active backend's capabilities

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR #22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

* test(video_gen/xai): full surface integration tests + cleaner schema

Verified end-to-end that the xAI plugin handles every documented mode
from PR #10600's surface: text-to-video, image-to-video,
reference-images-to-video, video edit, video extend (with and without
prompt). All five modes route to the correct xAI endpoint
(/videos/generations, /videos/edits, /videos/extensions) with the right
payload shape (image / reference_images / video keys), and all five
client-side rejections fire before the network: edit-without-prompt,
extend-without-video_url, image+refs conflict, >7 references, and
duration/aspect_ratio clamping.

15 new integration tests grouped into four classes (endpoint routing,
modalities, validation, clamping). httpx is stubbed via a small fake
AsyncClient that records POSTs so the tests assert the actual payload
the plugin would send to xAI — not just the success/error envelope.

Also cleaned up a description redundancy: when a model's operations
match the backend's overall set, we no longer print the duplicate
'operations supported by this model' line. xAI's description now reads:

    Active backend: xAI . model: grok-imagine-video
    - operations supported by this backend: edit, extend, generate
    - modalities supported by this backend: image, reference_images, text
    - aspect_ratio choices: 16:9, 1:1, 2:3, 3:2, 3:4, 4:3, 9:16
    - resolution choices: 480p, 720p
    - duration range: 1-15s
    - reference_image_urls: up to 7 images

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): collapse surface to t2v + i2v, family-based auto-routing

Two design changes per Teknium:

1) Drop edit/extend from the tool surface entirely. Only text-to-video
and image-to-video remain. The agent sees a clean tool with two
modalities; backend-specific quirks like xAI's edit/extend endpoints
stay out of the unified schema.

2) FAL: pick a model FAMILY once, the plugin routes between the
family's text-to-video and image-to-video endpoints based on whether
image_url was passed. Users no longer pick 'fal-ai/veo3.1' AND
'fal-ai/veo3.1/image-to-video' as separate options — they pick
'veo3.1', and the plugin handles the rest.

Catalog rewritten as families:

    veo3.1            fal-ai/veo3.1                                /  fal-ai/veo3.1/image-to-video
    pixverse-v6       fal-ai/pixverse/v6/text-to-video             /  fal-ai/pixverse/v6/image-to-video
    kling-o3-standard fal-ai/kling-video/o3/standard/text-to-video /  fal-ai/kling-video/o3/standard/image-to-video

xAI uses a single endpoint (/videos/generations) for both modes,
routed by the presence of the 'image' field in the payload — no
edit/extend exposure.

Schema changes:
- VIDEO_GENERATE_SCHEMA: drop operation, drop video_url. Final params:
  prompt (required), image_url, reference_image_urls, duration,
  aspect_ratio, resolution, negative_prompt, audio, seed, model.
- VideoGenProvider ABC: drop normalize_operation, VALID_OPERATIONS,
  DEFAULT_OPERATION. capabilities() drops 'operations' key.
- success_response: add 'modality' field ('text' | 'image') so the
  agent and logs can see which endpoint was actually hit.

Dynamic schema builder simplified — no operations bullet, no
'switch backends if you need edit/extend' guidance. When the active
backend supports both modalities (the common case), description reads:

    Active backend: FAL . model: pixverse-v6
    - supports both text-to-video (omit image_url) and image-to-video
      (pass image_url) - routes automatically
    - aspect_ratio choices: 16:9, 9:16, 1:1
    - resolution choices: 360p, 540p, 720p, 1080p
    - duration range: 1-15s
    - audio: pass audio=true to enable native audio (pricing tier)
    - negative_prompt: supported

Tests: 51 in the video_gen slice, 216 across the broader image+video
sweep, all passing. New FAL routing tests prove pixverse-v6 + no image
hits text-to-video endpoint, pixverse-v6 + image_url hits
image-to-video endpoint, same for veo3.1 and kling-o3-standard.

Docs updated: developer-guide page rewrites the 'model families' pattern
as a first-class section so external plugin authors know the convention.
toolsets-reference and toolsets.py descriptions match the new surface.

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen/fal): expand catalog to 6 families, cheap + premium tiers

Catalog now covers everything Teknium specced from FAL:

  Cheap tier:
    ltx-2.3        fal-ai/ltx-2.3-22b/text-to-video       / image-to-video
    pixverse-v6    fal-ai/pixverse/v6/text-to-video       / image-to-video

  Premium tier:
    veo3.1         fal-ai/veo3.1                          / fal-ai/veo3.1/image-to-video
    seedance-2.0   bytedance/seedance-2.0/text-to-video   / image-to-video
    kling-v3-4k    fal-ai/kling-video/v3/4k/text-to-video / image-to-video
    happy-horse    fal-ai/happy-horse/text-to-video       / image-to-video

DEFAULT_MODEL moved from veo3.1 (premium) to pixverse-v6 (cheap, sane
defaults, both modalities) — better first-run UX for users who haven't
explicitly picked a model.

New family-entry knob: image_param_key. Kling v3 4K's image-to-video
endpoint expects start_image_url instead of image_url; declaring
image_param_key='start_image_url' on the family lets _build_payload
remap correctly. Other families default to plain image_url.

Per-family capability flags reflect each model's docs:
- LTX 2.3 + Happy Horse: minimal payloads (no duration/aspect/resolution
  enum exposed by FAL — let endpoint apply defaults)
- Seedance: 6 aspect ratios incl 21:9, durations 4-15, audio supported,
  negative prompts NOT supported per docs
- Kling v3 4K: 16:9/9:16/1:1, 3-15s, audio + negative
- Veo 3.1: unchanged, 16:9/9:16, 4/6/8s

Tests: +5 covering the new families (full catalog, Kling 4K
start_image_url remap, Seedance routing, LTX payload minimality, Happy
Horse minimality). 56/56 in the slice green.

Note: I did NOT add the FAL-hosted xAI Grok-Imagine variant. Hermes
already has a direct xAI plugin that talks to xAI's own API; routing
the same model through FAL's wrapper would duplicate the surface
without adding capabilities. Users on FAL who want Grok-Imagine should
use the xAI plugin directly; flag if you want both routes available.

* test(video_gen): tool-surface routing matrix — every model x modality

End-to-end matrix test driven through _handle_video_generate() — the
actual function the agent's video_generate tool call lands in. Writes
config.yaml, invokes the registered handler with a raw args dict, then
asserts the outbound HTTP/SDK call hit the right endpoint with the right
payload shape.

Parametrized over FAL_FAMILIES.keys() so the matrix auto-discovers new
families as they're added (add a family to FAL_FAMILIES and you get
both modalities tested for free).

Coverage:
- All 6 FAL families x {text-only, text+image} = 12 cases
- xAI x {text-only, text+image} = 2 cases
- tool-level model= arg overrides config = 2 cases

For each case, verifies:
- result['success'] is True
- result['modality'] matches input shape ('text' if no image_url, 'image' otherwise)
- outbound endpoint URL matches the family's text_endpoint or image_endpoint
- text-only payloads carry no image-shaped keys
- text+image payloads carry the family's image key (image_url for most,
  start_image_url for kling-v3-4k, wrapped 'image' object for xAI)

All 16 cases passing. Confirms the tool surface routes every
(provider, model, modality) combination correctly with zero leakage.

* feat(video_gen): keep video_gen out of first-run setup, surface in status

Two changes:

1. video_gen joins _DEFAULT_OFF_TOOLSETS, so it is NOT pre-selected in
   the first-run toolset checklist. Video gen is niche, paid, and slow —
   most users don't want it nagging them during initial setup. Anyone
   who wants it opts in via 'hermes tools' -> Video Generation, which
   already routes to the provider+model picker.

2. The 'hermes setup' status panel learns about video_gen — but only
   shows the row when a plugin reports available. Users without
   FAL_KEY/XAI_API_KEY see nothing about video gen; users with one of
   those keys see 'Video Generation (FAL) ✓' as confirmation it's wired.

Verified live:
- Fresh install (no creds): zero video_gen mentions in wizard.
- With FAL_KEY: status row appears with active backend name.
- 160/160 in the setup + tools_config + video_gen test slice.

Rationale: image_gen is on by default because it's a featured creative
tool used in casual chat (telegrams, etc). Video gen is heavier — long
wait, paid per-second pricing. Default-off matches user intent better.

---------

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

…backends (NousResearch#25126)

* feat(video_gen): unified video_generate tool with pluggable provider backends

One core video_generate tool, every backend a plugin. Mirrors the
image_gen + memory_provider + context_engine architecture: ABC, registry,
plugin-context registration hook, and per-plugin model catalogs surfaced
through hermes tools.

Surface (one schema, every backend):
- operation: generate / edit / extend
- modalities: text-to-video (prompt only), image-to-video (prompt +
  image_url), video edit (prompt + video_url), video extend (video_url)
- reference_image_urls, duration, aspect_ratio, resolution,
  negative_prompt, audio, seed, model override
- Providers ignore unknown kwargs and declare what they support via
  VideoGenProvider.capabilities() — backend-specific quirks stay in the
  backend, the agent learns one tool

Backends shipped:
- plugins/video_gen/xai/  — Grok-Imagine, full generate/edit/extend +
  image-to-video + reference images (salvaged from PR NousResearch#10600 by
  @Jaaneek, reshaped into the plugin interface)
- plugins/video_gen/fal/  — Veo 3.1 (t2v + i2v), Kling O3 i2v,
  Pixverse v6 i2v with model-aware payload building that drops keys a
  model doesn't declare

Wiring:
- agent/video_gen_provider.py — VideoGenProvider ABC, normalize_operation,
  success_response / error_response, save_b64_video / save_bytes_video,
  $HERMES_HOME/cache/videos/
- agent/video_gen_registry.py — thread-safe register/get/list +
  get_active_provider() reading video_gen.provider from config.yaml
- hermes_cli/plugins.py — PluginContext.register_video_gen_provider()
- hermes_cli/tools_config.py — Video Generation category in
  hermes tools, plugin-only providers list, model picker per plugin,
  config write to video_gen.{provider,model}
- toolsets.py — new video_gen toolset
- tests: 31 new tests covering ABC, registry, tool dispatch, both plugins
- docs: developer-guide/video-gen-provider-plugin.md (parallel to the
  image-gen guide), sidebar + toolsets-reference + plugin guides updated

Supersedes: NousResearch#25035 (FAL), NousResearch#17972 (FAL), NousResearch#14543 (xAI), NousResearch#13847 (HappyHorse),
NousResearch#10458 (provider categories), NousResearch#10786 (xAI media+search bundle), NousResearch#2984
(FAL duplicate), NousResearch#19086 (Google Veo standalone — easy port to plugin
interface).

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): dynamic schema reflects active backend's capabilities

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR NousResearch#22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

* test(video_gen/xai): full surface integration tests + cleaner schema

Verified end-to-end that the xAI plugin handles every documented mode
from PR NousResearch#10600's surface: text-to-video, image-to-video,
reference-images-to-video, video edit, video extend (with and without
prompt). All five modes route to the correct xAI endpoint
(/videos/generations, /videos/edits, /videos/extensions) with the right
payload shape (image / reference_images / video keys), and all five
client-side rejections fire before the network: edit-without-prompt,
extend-without-video_url, image+refs conflict, >7 references, and
duration/aspect_ratio clamping.

15 new integration tests grouped into four classes (endpoint routing,
modalities, validation, clamping). httpx is stubbed via a small fake
AsyncClient that records POSTs so the tests assert the actual payload
the plugin would send to xAI — not just the success/error envelope.

Also cleaned up a description redundancy: when a model's operations
match the backend's overall set, we no longer print the duplicate
'operations supported by this model' line. xAI's description now reads:

    Active backend: xAI . model: grok-imagine-video
    - operations supported by this backend: edit, extend, generate
    - modalities supported by this backend: image, reference_images, text
    - aspect_ratio choices: 16:9, 1:1, 2:3, 3:2, 3:4, 4:3, 9:16
    - resolution choices: 480p, 720p
    - duration range: 1-15s
    - reference_image_urls: up to 7 images

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): collapse surface to t2v + i2v, family-based auto-routing

Two design changes per Teknium:

1) Drop edit/extend from the tool surface entirely. Only text-to-video
and image-to-video remain. The agent sees a clean tool with two
modalities; backend-specific quirks like xAI's edit/extend endpoints
stay out of the unified schema.

2) FAL: pick a model FAMILY once, the plugin routes between the
family's text-to-video and image-to-video endpoints based on whether
image_url was passed. Users no longer pick 'fal-ai/veo3.1' AND
'fal-ai/veo3.1/image-to-video' as separate options — they pick
'veo3.1', and the plugin handles the rest.

Catalog rewritten as families:

    veo3.1            fal-ai/veo3.1                                /  fal-ai/veo3.1/image-to-video
    pixverse-v6       fal-ai/pixverse/v6/text-to-video             /  fal-ai/pixverse/v6/image-to-video
    kling-o3-standard fal-ai/kling-video/o3/standard/text-to-video /  fal-ai/kling-video/o3/standard/image-to-video

xAI uses a single endpoint (/videos/generations) for both modes,
routed by the presence of the 'image' field in the payload — no
edit/extend exposure.

Schema changes:
- VIDEO_GENERATE_SCHEMA: drop operation, drop video_url. Final params:
  prompt (required), image_url, reference_image_urls, duration,
  aspect_ratio, resolution, negative_prompt, audio, seed, model.
- VideoGenProvider ABC: drop normalize_operation, VALID_OPERATIONS,
  DEFAULT_OPERATION. capabilities() drops 'operations' key.
- success_response: add 'modality' field ('text' | 'image') so the
  agent and logs can see which endpoint was actually hit.

Dynamic schema builder simplified — no operations bullet, no
'switch backends if you need edit/extend' guidance. When the active
backend supports both modalities (the common case), description reads:

    Active backend: FAL . model: pixverse-v6
    - supports both text-to-video (omit image_url) and image-to-video
      (pass image_url) - routes automatically
    - aspect_ratio choices: 16:9, 9:16, 1:1
    - resolution choices: 360p, 540p, 720p, 1080p
    - duration range: 1-15s
    - audio: pass audio=true to enable native audio (pricing tier)
    - negative_prompt: supported

Tests: 51 in the video_gen slice, 216 across the broader image+video
sweep, all passing. New FAL routing tests prove pixverse-v6 + no image
hits text-to-video endpoint, pixverse-v6 + image_url hits
image-to-video endpoint, same for veo3.1 and kling-o3-standard.

Docs updated: developer-guide page rewrites the 'model families' pattern
as a first-class section so external plugin authors know the convention.
toolsets-reference and toolsets.py descriptions match the new surface.

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen/fal): expand catalog to 6 families, cheap + premium tiers

Catalog now covers everything Teknium specced from FAL:

  Cheap tier:
    ltx-2.3        fal-ai/ltx-2.3-22b/text-to-video       / image-to-video
    pixverse-v6    fal-ai/pixverse/v6/text-to-video       / image-to-video

  Premium tier:
    veo3.1         fal-ai/veo3.1                          / fal-ai/veo3.1/image-to-video
    seedance-2.0   bytedance/seedance-2.0/text-to-video   / image-to-video
    kling-v3-4k    fal-ai/kling-video/v3/4k/text-to-video / image-to-video
    happy-horse    fal-ai/happy-horse/text-to-video       / image-to-video

DEFAULT_MODEL moved from veo3.1 (premium) to pixverse-v6 (cheap, sane
defaults, both modalities) — better first-run UX for users who haven't
explicitly picked a model.

New family-entry knob: image_param_key. Kling v3 4K's image-to-video
endpoint expects start_image_url instead of image_url; declaring
image_param_key='start_image_url' on the family lets _build_payload
remap correctly. Other families default to plain image_url.

Per-family capability flags reflect each model's docs:
- LTX 2.3 + Happy Horse: minimal payloads (no duration/aspect/resolution
  enum exposed by FAL — let endpoint apply defaults)
- Seedance: 6 aspect ratios incl 21:9, durations 4-15, audio supported,
  negative prompts NOT supported per docs
- Kling v3 4K: 16:9/9:16/1:1, 3-15s, audio + negative
- Veo 3.1: unchanged, 16:9/9:16, 4/6/8s

Tests: +5 covering the new families (full catalog, Kling 4K
start_image_url remap, Seedance routing, LTX payload minimality, Happy
Horse minimality). 56/56 in the slice green.

Note: I did NOT add the FAL-hosted xAI Grok-Imagine variant. Hermes
already has a direct xAI plugin that talks to xAI's own API; routing
the same model through FAL's wrapper would duplicate the surface
without adding capabilities. Users on FAL who want Grok-Imagine should
use the xAI plugin directly; flag if you want both routes available.

* test(video_gen): tool-surface routing matrix — every model x modality

End-to-end matrix test driven through _handle_video_generate() — the
actual function the agent's video_generate tool call lands in. Writes
config.yaml, invokes the registered handler with a raw args dict, then
asserts the outbound HTTP/SDK call hit the right endpoint with the right
payload shape.

Parametrized over FAL_FAMILIES.keys() so the matrix auto-discovers new
families as they're added (add a family to FAL_FAMILIES and you get
both modalities tested for free).

Coverage:
- All 6 FAL families x {text-only, text+image} = 12 cases
- xAI x {text-only, text+image} = 2 cases
- tool-level model= arg overrides config = 2 cases

For each case, verifies:
- result['success'] is True
- result['modality'] matches input shape ('text' if no image_url, 'image' otherwise)
- outbound endpoint URL matches the family's text_endpoint or image_endpoint
- text-only payloads carry no image-shaped keys
- text+image payloads carry the family's image key (image_url for most,
  start_image_url for kling-v3-4k, wrapped 'image' object for xAI)

All 16 cases passing. Confirms the tool surface routes every
(provider, model, modality) combination correctly with zero leakage.

* feat(video_gen): keep video_gen out of first-run setup, surface in status

Two changes:

1. video_gen joins _DEFAULT_OFF_TOOLSETS, so it is NOT pre-selected in
   the first-run toolset checklist. Video gen is niche, paid, and slow —
   most users don't want it nagging them during initial setup. Anyone
   who wants it opts in via 'hermes tools' -> Video Generation, which
   already routes to the provider+model picker.

2. The 'hermes setup' status panel learns about video_gen — but only
   shows the row when a plugin reports available. Users without
   FAL_KEY/XAI_API_KEY see nothing about video gen; users with one of
   those keys see 'Video Generation (FAL) ✓' as confirmation it's wired.

Verified live:
- Fresh install (no creds): zero video_gen mentions in wizard.
- With FAL_KEY: status row appears with active backend name.
- 160/160 in the setup + tools_config + video_gen test slice.

Rationale: image_gen is on by default because it's a featured creative
tool used in casual chat (telegrams, etc). Video gen is heavier — long
wait, paid per-second pricing. Default-off matches user intent better.

---------

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* chore(release): add AUTHOR_MAP entry for ryptotalent

* fix(auxiliary): pass cfg_base_url and cfg_api_key when resolving task provider

_resolve_task_provider_model drops cfg_base_url and cfg_api_key when
returning a named provider, causing configured API keys and base URLs
to be lost. Pass them through so named providers can use custom
endpoints while still resolving credentials from provider-specific
env vars.

Closes #20139

* chore(release): add AUTHOR_MAP entry for hookinglau

* fix(dashboard): skip browser-open on headless Linux to prevent process exit

Fixes #24127

On headless Linux VPS (no DISPLAY or WAYLAND_DISPLAY), some Python
webbrowser backends register TUI programs such as links, lynx, or
www-browser.  GenericBrowser.open() spawns these without redirecting
stdin/stdout, allowing them to take over the terminal.  This can cause
the process to receive SIGHUP and exit immediately even though uvicorn
bound the port successfully, producing a misleading success message
followed by an empty --status.

Fix: detect headless Linux at startup and skip the auto-open when no
display server is available.  On such systems the URL is still printed
so the user can open it manually or via an SSH tunnel.  The webbrowser
call is also wrapped in a try/except so any unexpected failure on other
platforms is silently absorbed rather than surfacing as an unhandled
exception in the daemon thread.

* chore(release): add AUTHOR_MAP entries for AllynSheep

* fix(dashboard): display real config path on Config page

Replace the hardcoded i18n placeholder "~/.hermes/config.yaml" with the
real config_path returned from api.getStatus(), falling back to the i18n
string while loading or on API failure.

Co-authored-by: aqilaziz <gonzes7@gmail.com>

* chore(release): add AUTHOR_MAP entries for aqilaziz

* fix(dashboard): rescan plugins when cached directory is removed

* fix(tui): clear scrollback buffer on startup to prevent tmux scrollback leakage

When TUI exits, tmux captures some TUI output into its scrollback buffer.
On restart, stale scrollback content appears at the top of screen before
AlternateScreen takes over.

Add ANSI escape sequences at startup:
- ESC[2J  clear visible screen
- ESC[H   cursor home
- ESC[3J  clear scrollback buffer

* chore(release): add AUTHOR_MAP entries for laoli-no1

* fix(cli): use display-width for response box header label to support CJK

Replace `len(label)` with `HermesCLI._status_bar_display_width(label)`
in two places where the response box top border is rendered.

`len()` counts characters, not terminal columns. CJK characters like
`测` and `试` each occupy 2 columns, causing the top border
`╭─ 测试 ───╮` to render 2 columns wider than the bottom border
`╰─────────╯`.

The `_status_bar_display_width` helper already exists (line 2881) and
uses `prompt_toolkit.utils.get_cwidth` for proper CJK width calculation.

* chore(release): add AUTHOR_MAP entries for NorethSea

* fix(cli): @-file completion crash on Windows when paths aren't cp1252-decodable

The fuzzy @-file completer shells out to 'rg --files' via subprocess.run
with text=True. On Windows, Python 3.13 decodes stdout using the system
ANSI codepage (cp1252), so any filename containing bytes like 0x81/0x8f
crashes the background reader thread with UnicodeDecodeError. The
exception is swallowed inside subprocess, leaving proc.stdout=None, and
the next line ('proc.stdout.strip()') blows up with:

  AttributeError: 'NoneType' object has no attribute 'strip'

This takes down the prompt_toolkit event loop and forces 'Press ENTER to
continue' until the user clears the @-query.

Fix:
- Pass encoding='utf-8', errors='replace' so rg's UTF-8 output is decoded
  consistently across platforms and unmappable bytes don't crash.
- Guard 'proc.stdout' with a None check before .strip(), so a future
  reader-thread failure degrades gracefully instead of breaking input.

* chore(release): add AUTHOR_MAP entries for JamesX88

* fix(telegram): clear in-progress reaction on cancelled processing (#24628)

When the user runs /stop or a session is interrupted mid-flight, the
👀 in-progress reaction lingered on the user's message indefinitely.
Without another agent run to swap it for 👍/👎, the eyes stayed there
forever — visually misleading (looks like the agent is still working).

Fix: on ProcessingOutcome.CANCELLED, call set_message_reaction with
reaction=None to clear all reactions on the message. Documented Bot API
semantics (equivalent to Bot API 10.0's deleteMessageReaction, but works
on PTB 22.6 already without the version bump).

Test changes:
- Renamed test_on_processing_complete_cancelled_keeps_existing_reaction
  → test_on_processing_complete_cancelled_clears_reaction; updated
  assertion to expect set_message_reaction(reaction=None).
- Added test_on_processing_complete_cancelled_skipped_when_disabled
  (TELEGRAM_REACTIONS=false short-circuits).
- Added test_clear_reactions_handles_api_error_gracefully and
  test_clear_reactions_returns_false_without_bot to cover the new
  _clear_reactions helper.

* fix(lsp): typescript SDK install + tsc-missing skip + shellcheck warning (#24630)

Three follow-ups to PR #24168 found during live E2E testing on TS/bash files:

1. typescript-language-server now installs the typescript SDK (tsserver)
   alongside it. Without that sibling install, initialize() failed with
   "Could not find a valid TypeScript installation" and the server was
   marked broken — no diagnostics ever reached the agent. New extra_pkgs
   field on INSTALL_RECIPES makes that explicit and reusable for future
   peer-dep cases.

2. _check_lint now treats "linter command exists on PATH but cannot
   actually run" as skipped instead of error. The motivating case is
   npx tsc when typescript is not in node_modules — npx prints its
   "This is not the tsc command you are looking for" banner and exits
   non-zero, which previously blocked the LSP semantic tier (gated on
   success or skipped). Pattern-matched per base command (npx,
   rustfmt, go) so genuine lint errors still flow through normally.

3. hermes lsp status now surfaces a Backend warnings section when
   bash-language-server is installed but shellcheck is missing. The
   server itself spawns fine but bash-language-server delegates
   diagnostics to shellcheck — without it on PATH the integration
   looks alive but never reports any problems. Same warning is
   logged once at server spawn time.

Validation:

- 12 new tests in tests/agent/lsp/test_install_and_lint_fixes.py:
    * recipe carries typescript SDK
    * _install_npm passes both pkg + extras to npm CLI
    * backwards compat: recipes without extras still work
    * _backend_warnings quiet when bash absent / both present
    * _backend_warnings fires when bash installed without shellcheck
    * status output includes the Backend warnings section
    * _looks_like_linter_unusable catches the npx tsc banner
    * real TS type errors not misclassified as unusable
    * unfamiliar linters fall through normally
    * _check_lint returns skipped on npx tsc unusable
    * _check_lint returns error on real tsc type errors
- Full lsp + file_operations test suite: 245/245 pass
- Live E2E:
    * try_install("typescript-language-server") installs both packages
      into node_modules
    * write_file(bad.ts, ...) returns lint=skipped + lsp_diagnostics
      with two real TS errors (was lint=error, no lsp_diagnostics)
    * hermes lsp status renders the shellcheck warning when bash is
      installed but shellcheck is not on PATH

* docs(gateway): mention Weixin in gateway help and docstrings

Salvage of #21063 — adds 'Weixin, and more' to module-level docstrings
in gateway/__init__.py, gateway/config.py, gateway/platforms/base.py
and the 'hermes gateway' subparser description.

Co-authored-by: wuwuzhijing <chuang.guo@hopechart.com>

* fix(ci): install ripgrep in e2e job

Closes #22003

* fix(ci): bump e2e job timeout to 15 minutes

Closes #22006

* fix(gateway): reduce systemd restart delay

* fix(send_message): recognize XMPP JIDs as explicit targets

_parse_target_ref() has no handler for XMPP JIDs (user@server or
room@conference.server), so they fall through to the final
`return None, None, False`. This causes send_message to fail when
targeting an XMPP chat by JID, since the JID is not numeric and
doesn't match any other platform pattern.

Add an explicit check for XMPP targets containing '@', matching the
existing Matrix pattern above it.

* fix(doctor): skip /models health check for providers that don't support it

Xiaomi MiMo's /v1/models endpoint returns 401 even with a valid API key,
causing hermes doctor to falsely report 'invalid API key'.

Add a `supports_health_check` field to ProviderProfile (default True).
Providers whose /models endpoint doesn't support auth verification can
set it to False. The doctor's dynamic provider discovery now reads this
field instead of hardcoding True.

The xiaomi provider plugin sets supports_health_check=False.

* fix(web): add Bearer auth header for Tavily /crawl endpoint

Tavily's /crawl endpoint requires Authorization: Bearer <key> in the header,
unlike /search and /extract which accept api_key in the JSON body.
Without the header, crawl returns 401 Unauthorized.

* fix(cron): include whatsapp in _HOME_TARGET_ENV_VARS

Cron jobs using `deliver: whatsapp` were silently dropped because the
resolver's home-channel env var dict in cron/scheduler.py listed every
messaging platform except whatsapp. _resolve_delivery_targets() returned
[] and no message was sent — but jobs.json marked the run successful and
no log line surfaced the failure.

The gateway adapter and the send_message tool path both honored
WHATSAPP_HOME_CHANNEL correctly; only the cron path missed.

Adds 'whatsapp' -> 'WHATSAPP_HOME_CHANNEL' to _HOME_TARGET_ENV_VARS.
Verified end-to-end with multiple cron pings landing in WhatsApp
self-chat after the fix.

Fixes #22997

* fix(cache): drop ttl=1h on Portal Qwen — Alibaba upstream is 5m-only (#24702)

PR #24151 routed Portal Qwen (qwen3.6-plus) through the prefix_and_2
long-lived cache layout, attaching {"type":"ephemeral","ttl":"1h"}
markers to the tools[-1] entry and the stable system-prefix block.
That layout works for Portal Claude because Anthropic / OpenRouter on
Anthropic routes honour 1h TTL — but Portal Qwen ultimately proxies to
Alibaba DashScope, which documents a single "ephemeral" TTL of 5
minutes on its Context Cache. The ttl="1h" qualifier is silently
dropped upstream, so the two highest-value breakpoints (tools array +
system prefix) never land. Only the rolling-window 5m markers on the
last 2 messages cache, which matches the observed ~25% read rate.

Fix: keep Portal Qwen on cache_control via _anthropic_prompt_cache_policy
returning (True, False), but drop it from _supports_long_lived_anthropic_cache
so it rides the standard system_and_3 5m layout (system + last 3 messages,
all at 5m). Same 4 breakpoints, all in a TTL the upstream actually honours.

Refs: https://www.alibabacloud.com/help/en/model-studio/context-cache
      https://openrouter.ai/docs/features/prompt-caching (Alibaba Qwen
      section: "TTL: 5 minutes")

- _supports_long_lived_anthropic_cache: Portal scope narrowed back to Claude
- tests: flip the two qwen long-lived expectations to False, retitle
  non_claude_non_qwen_rejected -> non_claude_rejected

* fix(retry): use float() for Retry-After header to handle sub-second values

* fix(signal): handle group messages from linked devices in syncMessage path

Closes #23064

When Hermes connects to Signal via signal-cli in daemon mode (linked
device setup), group messages sent from the user's phone were silently
dropped. The syncMessage handler only processed events where
destinationNumber equals the bot's own number (Note to Self).

Group messages from linked devices carry a groupInfo.groupId instead of a
destinationNumber. Extend the condition to also pass through sync messages
that have a groupId, so group messages are promoted to dataMessage and
reach the agent.

* fix(voice_mode): detect audio in WSL when sd.query_devices() returns empty list but PULSE_SERVER is set

In WSL2, sounddevice.query_devices() returns [] even when the
PulseAudio bridge is functional. The existing code already handled
the case where the query itself raises an exception, but it missed
the empty-list case.

This change treats an empty device list as non-fatal in WSL when
PULSE_SERVER is configured, matching the existing exception-handler
behavior.

Fixes: WSL users seeing 'No audio input/output devices detected'
even though paplay/arecord work fine.

* fix(tui): use TERMINAL_CWD in _session_info for accurate status line path

_session_info() used os.getcwd() which reflects the gateway process
working directory, not the user's actual working directory. This caused
the TUI status line to display incorrect paths (e.g. D:\HermesWork
instead of D:\Hermes\HermesWork) after agent turns that changed the
process cwd.

Align with session.create which already correctly reads TERMINAL_CWD
env var set by the CLI launcher.

* docs(lsp): document follow-up fixes from #24630 (#24709)

- Note that typescript-language-server pulls in the typescript SDK
  automatically (peer-dep relationship was previously implicit and
  caused initialize failures when the SDK was absent).
- Add a Troubleshooting entry for the new Backend warnings section
  in hermes lsp status, with the shellcheck install commands across
  apt / brew / scoop.

Reflects what shipped in PR #24630.

* fix(gateway): add chat_id to hook_ctx for message source tracking

* fix(install): use stash@{0} instead of git rev-parse refs/stash for autostash recovery
Autostash creates refs/stash as a pointer to the latest stash commit, but
git stash apply/drop expect the symbolic ref format like stash@{0}, not
the raw commit SHA. Using the commit SHA causes: error: 'X is not a stash reference'

* fix(telegram): use thread fallback helper in slash-confirm result send

PR #23458 introduced _send_message_with_thread_fallback() and applied it
to all control-style sends (send_update_prompt, send_approval_request,
send_model_picker_prompt), but the slash-confirm result message in
handle_callback_query still called self._bot.send_message directly.

In supergroups with stale message_thread_id on the callback's parent
message, this raises "Message thread not found" and silently swallows
the result text. Replace with the helper so the same retry-without-
thread-id logic applies.

* fix(prompt_builder): inject tool-use enforcement for GLM models

GLM-family models (z-ai/glm-4.5-air, z-ai/glm-4.5-flash, etc.) exhibit
the same "describe-instead-of-call" failure mode that gpt/codex/gemini/
gemma/grok already trigger enforcement for. Without the injection,
free-tier GLM workers spawned by the kanban dispatcher routinely exit
cleanly (rc=0) without invoking kanban_complete or kanban_block,
producing the "protocol violation" error and triggering the dispatcher's
gave_up path.

Observed in real workloads: seven consecutive kanban tasks across three
GLM-tier profiles (shipbackend, frontend-engineer, backend-engineer) all
failed with the identical message:

    worker exited cleanly (rc=0) without calling kanban_complete or
    kanban_block — protocol violation

Re-running the same tasks on Claude Haiku immediately resolved them.
Adding "glm" to TOOL_USE_ENFORCEMENT_MODELS closes the gap so future
GLM-routed work receives the explicit "every response must contain a
tool call or final result" steering that already protects the other
enforcement-gated model families.

One-line change; no behavior change for non-GLM models.

* fix(line): use build_source instead of nonexistent create_source

The LINE adapter calls self.create_source(...) which raises
AttributeError on every inbound message — no such method exists.
The base PlatformAdapter exposes this factory as build_source(),
consistent with the IRC and Teams adapters.

Fixes #23728

* fix(wecom): update connection status after WebSocket reconnection

The WeCom adapter's _listen_loop() automatically reconnects when the
WebSocket drops, but it never called _mark_connected() after a successful
reconnection. This left the runtime status file (gateway_state.json) stuck
in "disconnected" even though the adapter was fully operational again.

Add self._mark_connected() right after _open_connection() succeeds so
that the dashboard and health probes report the correct state.

Tested by forcing a WebSocket close via the heartbeat loop and verifying
that the status file updated from "disconnected" back to "connected".

* fix(tools): forward thread_id via metadata in _send_via_adapter live path

The live adapter path in _send_via_adapter called adapter.send() without
passing thread_id, while the standalone fallback path correctly forwarded
it. For plugin platforms (google_chat, teams, irc, line) running with the
gateway in-process, this caused every threaded reply to land as a new
top-level message instead of continuing the thread.

Matches the pattern already used by _send_matrix_via_adapter and
_send_feishu: build metadata={"thread_id": thread_id} and pass it through.

* fix(gateway): make WhatsApp npm install timeout configurable

Default timeout raised from 60s to 300s (5 minutes) to accommodate
slower systems like Unraid NAS. Configurable via WHATSAPP_NPM_INSTALL_TIMEOUT
environment variable.

* fix(test): use i18n t() for restart drain assertion

The test_restart_command_while_busy_requests_drain_without_interrupt test
was asserting against a hardcoded emoji string that was valid before the
i18n migration. After gateway/run.py switched to t("gateway.draining",
count=N), the test sees the translated output (or the raw key when the
locale catalog isn't resolved in xdist workers).

Fix by asserting against t("gateway.draining", count=1) — this produces
the correct expected value regardless of whether the locale file is
available in the test environment.

* fix(agent): clear stale config context_length on model switch

When switching models via /model, AIAgent._config_context_length was
never cleared, so the new model inherited the previous model's context
window instead of auto-detecting the correct one via
get_model_context_length().

Clear _config_context_length to None before the runtime field swap so
the full resolution chain (custom_providers per-model, endpoint probe,
models.dev, etc.) is re-evaluated for the newly selected model.

Closes #21509

* fix: approval DELETE pattern DOTALL flag allows newline bypass

* fix(cache): kill long-lived prefix layout — system prompt is now byte-static within a session (#24778)

The long-lived prefix-cache layout split the system prompt into stable/
context/volatile blocks and re-derived them on every API call. The
volatile tier (timestamp + memory snapshot + USER profile) ticks per
turn, so the system message bytes mutated mid-conversation and broke
upstream prompt caches (OpenRouter, Nous Portal, Anthropic).

Diagnosed via live wire-format diffing: an 8-turn conversation showed
OLD layout flipping system block[1] sha mid-session at the minute
boundary, dropping cached_tokens to 0 on that turn (cumulative
66.6% vs 83.3% for the single-block layout). Hermes invariant:
history (system + all but the last 1-2 messages) must be static.

Fix: drop the long-lived layout entirely. Single layout everywhere —
system_and_3 with one cached system string built once on first turn,
replayed verbatim on every subsequent turn. Loses cross-session 1h
prefix caching for Claude (the feature that motivated the split), but
within-session caching now actually works on every provider.

Removed:
- run_agent.py: _use_long_lived_prefix_cache flag, _long_lived_cache_ttl,
  _supports_long_lived_anthropic_cache method, the long-lived branch in
  run_conversation, mark_tools_for_long_lived_cache call site
- agent/prompt_caching.py: apply_anthropic_cache_control_long_lived,
  mark_tools_for_long_lived_cache, _mark_system_stable_block helper
- hermes_cli/config.py: prompt_caching.long_lived_prefix and
  prompt_caching.long_lived_ttl config keys
- tests/agent/test_prompt_caching_live.py (entire file)
- tests/agent/test_prompt_caching.py: TestMarkToolsForLongLivedCache,
  TestApplyAnthropicCacheControlLongLived
- tests/run_agent/test_anthropic_prompt_cache_policy.py:
  TestSupportsLongLivedAnthropicCache

Targeted tests: 62/62 pass.

* feat(nous): unified client=hermes-client-v<version> tag on every Portal request (#24779)

* feat(nous): unified client=hermes-client-v<version> tag on every Portal request

Every Hermes request to Nous Portal now carries the same
client=hermes-client-v<__version__> tag (e.g. client=hermes-client-v0.13.0
on this release), sourced live from hermes_cli.__version__. The release
script's regex bump auto-aligns it on every release.

Centralized in agent/portal_tags.py and wired into all four call sites:
- NousProfile.build_extra_body (main agent loop, every chat completion)
- auxiliary_client.NOUS_EXTRA_BODY + _build_call_kwargs (aux client)
- run_agent.py compression-summary fallback path
- tools/web_tools.py web_extract fallback

Replaces the client=aux marker added in #24194 with the unified version
tag. Tests assert against the helper output (invariant) rather than the
literal string, so they don't need updating on every release.

* feat(nous): cover /goal judge and kanban specify aux paths

Two aux-using surfaces bypassed call_llm by invoking
client.chat.completions.create() directly without extra_body, so they
were missing the unified Portal client tag:

- hermes_cli/goals.py — /goal standing-goal judge
- hermes_cli/kanban_specify.py — kanban triage specifier

Both now pass extra_body=get_auxiliary_extra_body() or None so they
inherit the version tag when the aux client points at Nous Portal, and
emit nothing otherwise (no tag leak to OpenRouter/Anthropic auxes).

* feat(providers): rename Alibaba Cloud to Qwen Cloud, reorder picker (#24835)

- Rename 'Alibaba Cloud (DashScope)' display label to 'Qwen Cloud'
  in CANONICAL_PROVIDERS (model picker, /model, hermes model TUI) and
  PROVIDER_REGISTRY (setup wizard prompts, status output).
- Move Qwen Cloud (alibaba) up to position 6 — directly below
  OpenAI Codex and above Xiaomi MiMo.
- Move Qwen OAuth (Portal) (qwen-oauth) to the bottom of the
  canonical provider list.

Provider slug 'alibaba' is unchanged — only the display label
moved. DashScope env var (DASHSCOPE_API_KEY) and base URL are
unchanged. The separate 'alibaba-coding-plan' plugin provider is
not affected.

* fix(docker): chown .venv to hermes so lazy_deps can install platform packages (#24841)

The Dockerfile permissions section made /opt/hermes/.venv readable but not
writable by the hermes runtime user.  Since the 2026-05-12 policy change
moved messaging packages (discord.py, telegram, slack, etc.) out of [all]
and into lazy_deps.py, the Docker image no longer ships with them
pre-installed.  At first gateway boot, lazy_deps.ensure() tries to
`uv pip install` them into the venv but fails with EACCES because
site-packages is root-owned.

The result: every messaging platform adapter silently fails to load inside
Docker containers, producing only a cryptic "discord.py not installed"
warning despite the gateway being correctly configured.

Two-part fix:

1. Dockerfile: add /opt/hermes/.venv to the existing chown -R hermes:hermes
   line so the default (UID 10000) case works out of the box.

2. docker/entrypoint.sh: extend the needs_chown block to also re-chown the
   .venv when HERMES_UID is remapped. Without this, the build-time chown
   becomes stale when someone uses the documented HERMES_UID override in
   docker-compose.yml.

Fixes #21536
Related: #17674, #21543, #21755

* fix(cli): add 'lsp' to _BUILTIN_SUBCOMMANDS so plugin discovery is skipped

`lsp` is registered as a top-level subparser in `main()` (lines 9539-9545)
via `agent.lsp.cli.register_subparser`, so it shows up in `hermes --help`
output alongside the other built-ins. The `_BUILTIN_SUBCOMMANDS` set used
by `_plugin_cli_discovery_needed` to short-circuit the ~500-650ms plugin
import pass did not list it, so every `hermes lsp ...` invocation paid
the full discovery cost despite being a fully-built-in command.

This is also caught by the parity guard added in #22120:
`tests/hermes_cli/test_startup_plugin_gating.py::test_builtin_set_covers_every_registered_subcommand`
has been failing on clean origin/main with:

    AssertionError: _BUILTIN_SUBCOMMANDS is missing these live
    subcommands: ['lsp']. Add them to hermes_cli/main.py::_BUILTIN_SUBCOMMANDS
    so plugin discovery can be skipped when the user targets them.

Fix: add `"lsp"` to the frozenset (alphabetical position between `logs`
and `mcp`). The accompanying `test_builtin_set_has_no_phantom_entries`
guard still passes because `lsp` is genuinely live — registered via the
guarded `try/except Exception` in main() since #24168.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix: classify landed file mutations with diagnostics

* refactor: import FILE_MUTATING_TOOL_NAMES from shared module

Drops the duplicate _FILE_MUTATING_TOOLS frozenset in run_agent.py and
imports the canonical FILE_MUTATING_TOOL_NAMES from
agent/tool_result_classification.py (aliased as _FILE_MUTATING_TOOLS to
avoid renaming the existing call sites). Prevents future drift if
another file-mutating tool is added — only one set needs updating.

No behavior change: same frozenset({'write_file', 'patch'}), and the
117 PR-scoped tests still pass.

* fix(gateway): add lazy_deps.ensure() to slack, matrix, dingtalk, feishu adapters (#25014)

Only Discord and Telegram had lazy-install hooks in their
check_*_requirements() functions. The remaining four platforms that were
moved to lazy_deps (Slack, Matrix, DingTalk, Feishu) would just return
False immediately if their packages weren't pre-installed — no attempt
to install them at runtime.

This means even with the .venv permissions fix (#24841), these four
platforms would still fail to load in Docker (or any fresh install)
unless the user manually ran pip install.

Add the same lazy_deps.ensure() pattern to all four, matching the
existing Discord/Telegram implementation.

* ci(docker): split :latest (releases only) from :main (main HEAD)

Previously :latest tracked the tip of main, which meant pulling :latest
got you whatever was last merged — fine for development, surprising for
users who expect :latest to mean 'the most recent stable release'.

Reshape the publish flow so the floating tags carry their conventional
meaning:

  - :sha-<sha>      every main commit (unchanged, immutable)
  - :main           tip of main (NEW; what :latest used to do)
  - :<release_tag>  every published release, e.g. :v1.2.3 (unchanged)
  - :latest         most recent release (CHANGED; release-only now)

Implementation:

  - Rename the move-latest job to move-main; it still gates on push to
    main, still ancestor-checks the existing :main label before
    retagging, still uses cancel-in-progress: false so queued moves run
    serially.

  - Add a new move-latest job gated on release: published. Reads the
    OCI revision label off the existing :latest and only advances if
    the release commit is a strict descendant. This keeps backport
    releases on older branches (e.g. patching v1.1.5 after v1.2.3 has
    already shipped) from dragging :latest backwards.

  - merge job exposes pushed_release_tag and release_tag outputs so
    move-latest knows when to fire and what to retag from.

* fix(cli): repair broken zsh completion generation

* test(cli): strengthen zsh completion regression coverage

* refactor(profiles): remove dead generate_bash_completion / generate_zsh_completion

These two functions in hermes_cli/profiles.py have no callers — the live
`hermes completion {bash,zsh}` command uses hermes_cli/completion.py's
generate_bash() / generate_zsh() instead. Multiple PRs (incl. #6141) tried
to fix the trailing-`_hermes "$@"` zsh bug here, only to discover the
patch never reached users. Delete the dead code so future contributors
patch the right file.

The actual user-facing fix lives in the preceding cherry-picked commits
to hermes_cli/completion.py.

* chore(release): map anton.kuenzi@gmail.com -> ZeterMordio

For PR #11754 salvage (zsh completion compdef registration + _arguments
syntax tests). CI release script blocks unmapped emails.

* fix(security): reduce unnecessary shell=True in subprocess calls

- memory_setup.py: use shlex.split() for plugin dep checks instead of shell=True
- transcription_tools.py: avoid shell=True for auto-detected whisper commands
  (user-provided templates via env var still use shell=True for compatibility)
- cli.py: add comment clarifying intentional shell=True for user quick_commands
- Add test verifying auto-detected template is shlex-safe

Addresses CONTRIBUTING.md Priority #3 (Security hardening — shell injection).

* chore(release): map iuyup author for PR #6155 salvage

* feat(custom): prompt and persist explicit api_mode for custom providers

Adds an explicit API compatibility mode prompt to the `hermes model -> custom`
flow so Codex-compatible third-party endpoints (and any other non-default
backend whose URL doesn't match the existing heuristics in
`_detect_api_mode_for_url`) can be selected explicitly instead of silently
falling back to chat_completions.

Choices: Auto-detect / chat_completions / codex_responses / anthropic_messages.

Persists `api_mode` to:
  - `model.api_mode` (active session config)
  - the matching `custom_providers[*]` entry (so re-activating the named
    provider next time replays the same transport)

Salvaged from PR #6125 onto current main: kept the new prompt and the
`_save_custom_provider(api_mode=...)` plumbing; the named-custom flow
already extracts and applies `api_mode` from the saved entry on current
main so those changes are preserved as-is. Test fixtures updated for the
new prompt and the existing display-name prompt.

Co-authored-by: littlewwwhite <1095245867@qq.com>

* fix(setup): drop post-setup chat handoff (#25067)

Removes the 'Launch hermes chat now? (Y/n)' prompt at the end of
hermes setup. The summary already prints 'Ready to go! → hermes'
so the auto-launch was redundant, and on macOS 26+ it could crash
in prompt_toolkit when setup was invoked from the curl install
script with stdin redirected from /dev/tty (#5884, #6128).

After setup, users run 'hermes' themselves like every other CLI
tool. Same pattern applies to the Windows installer.

Closes #6128 (narrower env-var-guarded fix superseded by removing
the prompt outright).

* fix(tools): refuse skill_view name collisions instead of guessing

skill_view ran the direct-path strategy across every skill dir before
the recursive strategy, so a top-level skill in an external dir could
silently shadow a same-named nested local skill. /skills correctly
listed the local version (deduped local-first by _find_all_skills) but
skill_view loaded the external one — confusing, and a real bug class
for users with skills.external_dirs registered alongside categorized
local skills.

Pick a louder fix than @polkn's PR #6136 proposed: collect every match
across all dirs (direct path, recursive by parent dir name, legacy
flat <name>.md), and if there's more than one, refuse with an error
that surfaces every matching path plus a hint to load by the
categorized form. Local-first precedence would have replaced silent
external-shadowing with silent same-name collisions between two
externals, or made an externally-shadowed-by-local skill unreachable
by bare name with no signal. Refusing forces the user to disambiguate
once and never wonder which skill ran.

Recovery: pass the full categorized path
("foundations/runtime/explore-codebase" instead of
"explore-codebase"), or rename one of the colliding skills.

Co-authored-by: pol <pol.kuijken@gmail.com>

* fix(cli): preserve startup banner on terminal resize

Recover from SIGWINCH without clearing the physical screen or scrollback
buffer. The startup banner and tool summary are printed before
prompt_toolkit owns the live chrome, so they live in normal terminal
scrollback. Calling erase_screen() + \x1b[3J] on every resize removed
that UI permanently — _replay_output_history cannot reconstruct it
because the banner was never added to _OUTPUT_HISTORY.

Instead, just reset prompt_toolkit's renderer cache and invalidate so
the next incremental redraw starts from a clean slate, then let the
original on_resize handler recalculate layout for the new terminal
size. This matches the behaviour of bash/zsh/fish on SIGWINCH.

Fixes NousResearch/hermes-agent#22999

* tui: make URLs clickable + hover-highlight in any terminal (#25071)

* tui: make URLs clickable + hover-highlight in any terminal

Problem
-------
URLs printed by `hermes --tui` were not clickable in basic macOS Terminal.app.
Cmd+click did nothing, the cursor didn't change shape — like nothing was
detected — even though arrow buttons and other Box onClick handlers worked
fine.

Root cause
----------
Two layers of dead plumbing:

1. `<Link>` only emitted the underlying `<ink-link>` (which carries the
   hyperlink metadata into the screen buffer) when `supportsHyperlinks()`
   said yes. On Apple_Terminal that's false, so the per-cell hyperlink
   field stayed empty, so `Ink.getHyperlinkAt()` had nothing to return on
   click. The visible underline was just decorative.

2. `Ink.openHyperlink()` calls `this.onHyperlinkClick?.(url)`, but
   `onHyperlinkClick` was never assigned anywhere in the codebase. The
   click pipeline (`App.tsx → onOpenHyperlink → Ink.openHyperlink`) ran
   but bailed silently on the optional chain.

Bonus discovery: even when wired up, there was no hover affordance —
terminal apps can't change the system mouse cursor, so users had no
visual signal that a cell was clickable. Arrow buttons in the chrome
worked because they had explicit `<Box onClick>` styling; inline link
URLs didn't.

Fix
---
- `Link.tsx`: always emit `<ink-link>` regardless of terminal capability.
  The renderer's `wrapWithOsc8Link` already gates the actual OSC 8 escape
  on `supportsHyperlinks()` further down — so terminals that don't
  understand OSC 8 still don't see the escape, but the screen-buffer
  metadata (which the click dispatcher reads) is now populated everywhere.

- `ink.tsx + root.ts`: add `onHyperlinkClick?: (url: string) => void` to
  `Options` / `RenderOptions`, wire it to the existing `Ink.onHyperlinkClick`
  field in the constructor.

- `src/lib/openExternalUrl.ts`: small platform-aware opener using
  `child_process.spawn` with arg-array (no shell) — http(s) only, rejects
  `file:`, `javascript:`, `data:`, etc., so a hostile model can't trigger
  arbitrary local handlers via `<Link url="file:///...">`. Detached + stdio
  ignore so closing the TUI doesn't kill the browser and Chrome stderr
  doesn't leak into the alt screen.

- `entry.tsx`: pass `onHyperlinkClick: openExternalUrl` to `ink.render`.

- `hyperlinkHover.ts` + Ink hover wiring: track the URL under the pointer
  in `Ink.hoveredHyperlink`, update it from `dispatchHover`, and inverse-
  highlight every cell of the matching link in the render-pass overlay
  (same pattern as `applySearchHighlight`). This is the cursor-hover
  affordance for clickable links — terminals don't expose cursor shape,
  so we light up the link itself.

- `types/hermes-ink.d.ts`: add `onHyperlinkClick` to the `RenderOptions`
  shim so consumers (`entry.tsx`) type-check against the new option.

Tests
-----
- `src/lib/openExternalUrl.test.ts` (15 cases): http(s) accepted; file/js/
  data/mailto/ftp/ssh rejected; macOS open(1), Windows cmd.exe start with
  empty title slot, Linux xdg-open dispatch; shell-metacharacter URLs
  pass through unmolested as a single argv element; synchronous spawn
  failure returns false.

Verified empirically in Apple Terminal 455.1 (macOS 15.7.3): clicking a
URL opens in default browser, hovering inverts the link cells, and
moving away clears the highlight. Full TUI suite: 713 passing, 0
type errors.

Reverts
-------
The earlier attempt that version-gated Apple_Terminal in
`supports-hyperlinks.ts` was based on a wrong assumption — Terminal.app
silently strips OSC 8 sequences but does not render them as clickable
hyperlinks. Reverted to the original allowlist.

* tui: address Copilot review — explorer.exe on win32 + comment fixes

- openExternalUrl: switch win32 from `cmd.exe /c start` to `explorer.exe`.
  cmd.exe's `start` builtin reparses the URL through cmd's tokenizer, so
  `&`, `|`, `^`, `<`, `>` either split the command or get reinterpreted —
  breaking both the protocol-allowlist safety story AND plain http(s) URLs
  with `&` in query strings. `explorer.exe <url>` invokes the registered
  protocol handler directly with no shell.

- openExternalUrl.test.ts: rename the win32 test to reflect the new
  contract and add two regression tests — one with `&|^<>` metachars,
  one with the common analytics-URL `&` query-param pattern — both pinned
  to single-argv-element delivery via explorer.exe.

- Link.tsx: fix misleading comment. OSC 8 escapes are emitted
  unconditionally by the renderer (`wrapWithOsc8Link` in
  render-node-to-output.ts, `oscLink` in log-update.ts). Non-supporting
  terminals silently strip the sequence, which is why hover/click
  affordance has to come from the in-process overlay rather than the
  terminal's own link rendering.

Verified: 715/715 tests pass, type-check + build clean.

* tui: address Copilot review #2 — async spawn errors + hover scope + docs

1. openExternalUrl: attach a no-op `'error'` listener on the spawned
   child BEFORE unref(). spawn() returns a ChildProcess synchronously
   even when the binary is missing (ENOENT on xdg-open / explorer.exe),
   unreachable, or otherwise unusable; the failure surfaces later as
   an 'error' event. An unhandled 'error' on an EventEmitter crashes
   Node, which would tear down the whole TUI. The listener is a
   deliberate no-op — we already returned `true` synchronously and the
   user just doesn't see the browser pop.

2. openExternalUrl.test.ts: add a regression test using a real
   EventEmitter to simulate the async-error path. Pins both the
   listener-attached contract and the "doesn't throw on emit" behavior.
   Was 17/17, now 18/18.

3. ink.tsx dispatchHover: bypass `getHyperlinkAt()` and read
   `cellAt(...).hyperlink` directly. `getHyperlinkAt` falls back to
   `findPlainTextUrlAt` for cells without an OSC 8 hyperlink, but the
   render-pass overlay (`applyHyperlinkHoverHighlight`) only matches on
   `cell.hyperlink === hoveredUrl` — so plain-text URLs would burn
   re-renders without ever producing the highlight. Hover is now a
   strictly 1:1 fit for what the overlay can paint. Plain-text URLs
   still get the click action via the existing dispatch path.

4. root.ts + ink.tsx doc comments: replace the misleading "typically
   `open` / `xdg-open` / `start` shell" wording with the actual safe
   recipe — argv-array spawn into `open` / `xdg-open` / `explorer.exe`,
   with an explicit warning that `cmd.exe /c start` reparses the URL
   through cmd's tokenizer and is unsafe + breaks `&`-query URLs.

Verified: 716/716 tests pass, type-check + build clean.

* tui: address Copilot review #3 — hover damage, alt-screen cleanup, opener allowlist

1. ink.tsx onRender: stop folding steady-state hover into hlActive.
   hlActive forces a full-screen damage diff so previous-frame inverted
   cells get re-emitted when the highlight set changes. The transition
   IS the trigger — enter / leave / change-to-other-link. While the
   pointer just sits on a link the painted cells don't change and the
   per-cell diff handles the no-op. Folding the steady state in would
   burn a full-screen diff on every frame. Added a
   lastRenderedHoveredHyperlink tracker and gate the hlActive bump on
   `hovered !== lastRendered`.

2. ink.tsx setAltScreenActive: clear hoveredHyperlink (and the tracker)
   when toggling alt-screen state. Hover dispatch is alt-screen-gated,
   so once we leave there's no path to clear it. Without this, remounting
   <AlternateScreen> would paint a phantom hover from the previous
   session until the next mouse-move arrived.

3. openExternalUrl.ts openCommand: allowlist linux + the BSD family for
   xdg-open and return null for everything else (aix, sunos, cygwin,
   haiku, etc.). Previously the default-fallback always returned
   xdg-open, which made the caller's `if (!command) return false` dead
   and yielded a misleading `true` on platforms that probably don't
   have xdg-open. New tests cover the null path AND the
   openExternalUrl-returns-false-without-spawning behavior.

Verified: 718/718 tests pass, type-check + build clean.

* tui: address Copilot review #4 — doc comment accuracy

1. openExternalUrl return-value doc: now lists all three false paths
   (URL rejected / no opener for platform / synchronous spawn throw)
   plus a note that async 'error' events still return true because the
   spawn was attempted.

2. ink.tsx onHyperlinkClick field doc: clarifies the callback receives
   either an OSC 8 hyperlink OR a plain-text URL detected by
   findPlainTextUrlAt — App.tsx routes both into the same callback.

3. hyperlinkHover applyHyperlinkHoverHighlight doc: drops the misleading
   'caller forces full-frame damage' promise. Caller decides; for hover
   the current caller only forces full damage on transitions.

No behavior change. 718/718 tests pass.

* tui: address Copilot review #5 — lint fixes

1. ink.tsx: reorder `./hyperlinkHover.js` import before `./screen.js` to
   satisfy perfectionist/sort-imports.

2. Link.tsx: drop unused `fallback` parameter destructuring + the
   trailing `void (null as ...)` dead-statement (would trip
   no-unused-expressions). Kept `fallback?: ReactNode` on the Props
   interface as a documented compat shim so existing call sites still
   compile, with a comment explaining why it's no longer wired up.

3. openExternalUrl.test.ts: replace `typeof import('node:child_process').spawn`
   inline annotations (forbidden by @typescript-eslint/consistent-type-imports)
   with a `SpawnLike` type alias backed by a real `import type { spawn as SpawnFn }`.

No behavior change. 718/718 tests pass, type-check clean, lint clean on
all modified files.

* fix(gateway): preserve queued follow-up transcript history

Keep the outer history_offset when _run_agent drains queued follow-ups recursively so transcript persistence includes every queued turn in the chain instead of only the last one.

* test(gateway): make queued follow-up regression generic

Replace tenant-specific example text in the transcript offset regression with generic follow-up turns so the upstream test documents the bug without customer-specific wording.

* chore(release): map mgongzai author for PR #25183 salvage

* feat(video_gen): unified video_generate tool with pluggable provider backends (#25126)

* feat(video_gen): unified video_generate tool with pluggable provider backends

One core video_generate tool, every backend a plugin. Mirrors the
image_gen + memory_provider + context_engine architecture: ABC, registry,
plugin-context registration hook, and per-plugin model catalogs surfaced
through hermes tools.

Surface (one schema, every backend):
- operation: generate / edit / extend
- modalities: text-to-video (prompt only), image-to-video (prompt +
  image_url), video edit (prompt + video_url), video extend (video_url)
- reference_image_urls, duration, aspect_ratio, resolution,
  negative_prompt, audio, seed, model override
- Providers ignore unknown kwargs and declare what they support via
  VideoGenProvider.capabilities() — backend-specific quirks stay in the
  backend, the agent learns one tool

Backends shipped:
- plugins/video_gen/xai/  — Grok-Imagine, full generate/edit/extend +
  image-to-video + reference images (salvaged from PR #10600 by
  @Jaaneek, reshaped into the plugin interface)
- plugins/video_gen/fal/  — Veo 3.1 (t2v + i2v), Kling O3 i2v,
  Pixverse v6 i2v with model-aware payload building that drops keys a
  model doesn't declare

Wiring:
- agent/video_gen_provider.py — VideoGenProvider ABC, normalize_operation,
  success_response / error_response, save_b64_video / save_bytes_video,
  $HERMES_HOME/cache/videos/
- agent/video_gen_registry.py — thread-safe register/get/list +
  get_active_provider() reading video_gen.provider from config.yaml
- hermes_cli/plugins.py — PluginContext.register_video_gen_provider()
- hermes_cli/tools_config.py — Video Generation category in
  hermes tools, plugin-only providers list, model picker per plugin,
  config write to video_gen.{provider,model}
- toolsets.py — new video_gen toolset
- tests: 31 new tests covering ABC, registry, tool dispatch, both plugins
- docs: developer-guide/video-gen-provider-plugin.md (parallel to the
  image-gen guide), sidebar + toolsets-reference + plugin guides updated

Supersedes: #25035 (FAL), #17972 (FAL), #14543 (xAI), #13847 (HappyHorse),
#10458 (provider categories), #10786 (xAI media+search bundle), #2984
(FAL duplicate), #19086 (Google Veo standalone — easy port to plugin
interface).

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): dynamic schema reflects active backend's capabilities

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR #22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

* test(video_gen/xai): full surface integration tests + cleaner schema

Verified end-to-end that the xAI plugin handles every documented mode
from PR #10600's surface: text-to-video, image-to-video,
reference-images-to-video, video edit, video extend (with and without
prompt). All five modes route to the correct xAI endpoint
(/videos/generations, /videos/edits, /videos/extensions) with the right
payload shape (image / reference_images / video keys), and all five
client-side rejections fire before the network: edit-without-prompt,
extend-without-video_url, image+refs conflict, >7 references, and
duration/aspect_ratio clamping.

15 new integration tests grouped into four classes (endpoint routing,
modalities, validation, clamping). httpx is stubbed via a small fake
AsyncClient that records POSTs so the tests assert the actual payload
the plugin would send to xAI — not just the success/error envelope.

Also cleaned up a description redundancy: when a model's operations
match the backend's overall set, we no longer print the duplicate
'operations supported by this model' line. xAI's description now reads:

    Active backend: xAI . model: grok-imagine-video
    - operations supported by this backend: edit, extend, generate
    - modalities supported by this backend: image, reference_images, text
    - aspect_ratio choices: 16:9, 1:1, 2:3, 3:2, 3:4, 4:3, 9:16
    - resolution choices: 480p, 720p
    - duration range: 1-15s
    - reference_image_urls: up to 7 images

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): collapse surface to t2v + i2v, family-based auto-routing

Two design changes per Teknium:

1) Drop edit/extend from the tool surface entirely. Only text-to-video
and image-to-video remain. The agent sees a clean tool with two
modalities; backend-specific quirks like xAI's edit/extend endpoints
stay out of the unified schema.

2) FAL: pick a model FAMILY once, the plugin routes between the
family's text-to-video and image-to-video endpoints based on whether
image_url was passed. Users no longer pick 'fal-ai/veo3.1' AND
'fal-ai/veo3.1/image-to-video' as separate options — they pick
'veo3.1', and the plugin handles the rest.

Catalog rewritten as families:

    veo3.1            fal-ai/veo3.1                                /  fal-ai/veo3.1/image-to-video
    pixverse-v6       fal-ai/pixverse/v6/text-to-video             /  fal-ai/pixverse/v6/image-to-video
    kling-o3-standard fal-ai/kling-video/o3/standard/text-to-video /  fal-ai/kling-video/o3/standard/image-to-video

xAI uses a single endpoint (/videos/generations) for both modes,
routed by the presence of the 'image' field in the payload — no
edit/extend exposure.

Schema changes:
- VIDEO_GENERATE_SCHEMA: drop operation, drop video_url. Final params:
  prompt (required), image_url, reference_image_urls, duration,
  aspect_ratio, resolution, negative_prompt, audio, seed, model.
- VideoGenProvider ABC: drop normalize_operation, VALID_OPERATIONS,
  DEFAULT_OPERATION. capabilities() drops 'operations' key.
- success_response: add 'modality' field ('text' | 'image') so the
  agent and logs can see which endpoint was actually hit.

Dynamic schema builder simplified — no operations bullet, no
'switch backends if you need edit/extend' guidance. When the active
backend supports both modalities (the common case), description reads:

    Active backend: FAL . model: pixverse-v6
    - supports both text-to-video (omit image_url) and image-to-video
      (pass image_url) - routes automatically
    - aspect_ratio choices: 16:9, 9:16, 1:1
    - resolution choices: 360p, 540p, 720p, 1080p
    - duration range: 1-15s
    - audio: pass audio=true to enable native audio (pricing tier)
    - negative_prompt: supported

Tests: 51 in the video_gen slice, 216 across the broader image+video
sweep, all passing. New FAL routing tests prove pixverse-v6 + no image
hits text-to-video endpoint, pixverse-v6 + image_url hits
image-to-video endpoint, same for veo3.1 and kling-o3-standard.

Docs updated: developer-guide page rewrites the 'model families' pattern
as a first-class section so external plugin authors know the convention.
toolsets-reference and toolsets.py descriptions match the new surface.

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen/fal): expand catalog to 6 families, cheap + premium tiers

Catalog now covers everything Teknium specced from FAL:

  Cheap tier:
    ltx-2.3        fal-ai/ltx-2.3-22b/text-to-video       / image-to-video
    pixverse-v6    fal-ai/pixverse/v6/text-to-video       / image-to-video

  Premium tier:
    veo3.1         fal-ai/veo3.1                          / fal-ai/veo3.1/image-to-video
    seedance-2.0   bytedance/seedance-2.0/text-to-video   / image-to-video
    kling-v3-4k    fal-ai/kling-video/v3/4k/text-to-video / image-to-video
    happy-horse    fal-ai/happy-horse/text-to-video       / image-to-video

DEFAULT_MODEL moved from veo3.1 (premium) to pixverse-v6 (cheap, sane
defaults, both modalities) — better first-run UX for users who haven't
explicitly picked a model.

New family-entry knob: image_param_key. Kling v3 4K's image-to-video
endpoint expects start_image_url instead of image_url; declaring
image_param_key='start_image_url' on the family lets _build_payload
remap correctly. Other families default to plain image_url.

Per-family capability flags reflect each model's docs:
- LTX 2.3 + Happy Horse: minimal payloads (no duration/aspect/resolution
  enum exposed by FAL — let endpoint apply defaults)
- Seedance: 6 aspect ratios incl 21:9, durations 4-15, audio supported,
  negative prompts NOT supported per docs
- Kling v3 4K: 16:9/9:16/1:1, 3-15s, audio + negative
- Veo 3.1: unchanged, 16:9/9:16, 4/6/8s

Tests: +5 covering the new families (full catalog, Kling 4K
start_image_url remap, Seedance routing, LTX payload minimality, Happy
Horse minimality). 56/56 in the slice green.

Note: I did NOT add the FAL-hosted xAI Grok-Imagine variant. Hermes
already has a direct xAI plugin that talks to xAI's own API; routing
the same model through FAL's wrapper would duplicate the surface
without adding capabilities. Users on FAL who want Grok-Imagine should
use the xAI plugin directly; flag if you want both routes available.

* test(video_gen): tool-surface routing matrix — every model x modality

End-to-end matrix test driven through _handle_video_generate() — the
actual function the agent's video_generate tool call lands in. Writes
config.yaml, invokes the registered handler with a raw args dict, then
asserts the outbound HTTP/SDK call hit the right endpoint with the right
payload shape.

Parametrized over FAL_FAMILIES.keys() so the matrix auto-discovers new
families as they're added (add a family to FAL_FAMILIES and you get
both modalities tested for free).

Coverage:
- All 6 FAL families x {text-only, text+image} = 12 cases
- xAI x {text-only, text+image} = 2 cases
- tool-level model= arg overrides config = 2 cases

For each case, verifies:
- result['success'] is True
- result['modality'] matches input shape ('text' if no image_url, 'image' otherwise)
- outbound endpoint URL matches the family's text_endpoint or image_endpoint
- text-only payloads carry no image-shaped keys
- text+image payloads carry the family's image key (image_url for most,
  start_image_url for kling-v3-4k, wrapped 'image' object for xAI)

All 16 cases passing. Confirms the tool surface routes every
(provider, model, modality) combination correctly with zero leakage.

* feat(video_gen): keep video_gen out of first-run setup, surface in status

Two changes:

1. video_gen joins _DEFAULT_OFF_TOOLSETS, so it is NOT pre-selected in
   the first-run toolset checklist. Video gen is niche, paid, and slow —
   most users don't want it nagging them during initial setup. Anyone
   who wants it opts in via 'hermes tools' -> Video Generation, which
   already routes to the provider+model picker.

2. The 'hermes setup' status panel learns about video_gen — but only
   shows the row when a plugin reports available. Users without
   FAL_KEY/XAI_API_KEY see nothing about video gen; users with one of
   those keys see 'Video Generation (FAL) ✓' as confirmation it's wired.

Verified live:
- Fresh install (no creds): zero video_gen mentions in wizard.
- With FAL_KEY: status row appears with active backend name.
- 160/160 in the setup + tools_config + video_gen test slice.

Rationale: image_gen is on by default because it's a featured creative
tool used in casual chat (telegrams, etc). Video gen is heavier — long
wait, paid per-second pricing. Default-off matches user intent better.

---------

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(codex-runtime): optional codex app-server runtime for OpenAI/Codex models (#24182)

* feat(codex-runtime): scaffold optional codex app-server runtime

Foundational commit for an opt-in alternate runtime that hands OpenAI/Codex
turns to a 'codex app-server' subprocess instead of Hermes' tool dispatch.
Default behavior is unchanged.

Lands in three pieces:

1. agent/transports/codex_app_server.py — JSON-RPC 2.0 over stdio speaker
   for codex's app-server protocol (codex-rs/app-server). Spawn, init
   handshake, request/response, notification queue, server-initiated
   request queue (for approval round-trips), interrupt-friendly blocking
   reads. Tested against real codex 0.130.0 binary end-to-end during
   development.

2. hermes_cli/runtime_provider.py:
   - Adds 'codex_app_server' to _VALID_API_MODES.
   - Adds _maybe_apply_codex_app_server_runtime() helper, called at the
     end of _resolve_runtime_from_pool_entry(). Inert unless
     'model.openai_runtime: codex_app_server' is set in config.yaml AND
     provider in {openai, openai-codex}. Other providers cannot be
     rerouted (anthropic, openrouter, etc. preserved).

3. tests/agent/transports/test_codex_app_server_runtime.py — 24 tests
   covering api_mode registration, the rewriter helper (default-off,
   case-insensitive, opt-in, non-eligible providers preserved), version
   parser, missing-binary handling, error class. Does NOT require codex
   CLI installed.

This commit is wire-only: the api_mode is recognized but AIAgent does
not yet branch on it. Followup commits add the session adapter, event
projector, approval bridge, transcript projection (so memory/skill
review still works), plugin migration, and slash command.

Existing tests remain green:
- tests/cli/test_cli_provider_resolution.py (29 passed)
- tests/agent/test_credential_pool_routing.py (included above)

* feat(codex-runtime): add codex item projector for memory/skill review

The translator that lets Hermes' self-improvement loop keep working under the
Codex runtime: converts codex 'item/*' notifications into Hermes' standard
{role, content, tool_calls, tool_call_id} message shape that
agent/curator.py already knows how to read.

Item taxonomy (matches codex-rs/app-server-protocol/src/protocol/v2/item.rs):
  - userMessage          → {role: user, content}
  - agentMessage         → {role: assistant, content: text}
  - reasoning            → stashed in next assistant's 'reasoning' field
  - commandExecution     → assistant tool_call(name='exec_command') + tool result
  - fileChange           → assistant tool_call(name='apply_patch') + tool result
  - mcpToolCall          → assistant tool_call(name='mcp.<server>.<tool>') + tool result
  - dynamicToolCall      → assistant tool_call(name=<tool>) + tool result
  - plan/hookPrompt/etc  → opaque assistant note, no fabricated tool_calls

Invariants preserved:
  - Message role alternation never violated: each tool item produces at most
    one assistant + one tool message in that order, correlated by call_id.
  - Streaming deltas (item/<type>/outputDelta, item/agentMessage/delta)
    don't materialize messages — only item/completed does. Mirrors how
    Hermes already only writes the assistant message after streaming ends.
  - Tool call ids are deterministic (codex item id-based) so replays produce
    identical messages and prefix caches stay valid (AGENTS.md pitfall #16).
  - JSON args use sorted_keys for the same reason.

Real wire formats verified against codex 0.130.0 by capturing live
notifications from thread/shellCommand and including one as a fixture
(COMMAND_EXEC_COMPLETED).

23 new tests, all green:
  - Streaming deltas don't materialize (3 paths)
  - Turn/thread frame events are silent
  - commandExecution: 5 tests including non-zero exit annotation +
    deterministic id stability across replays
  - agentMessage + reasoning attachment + reasoning consumption
  - fileChange: summary without inlined content
  - mcpToolCall: namespaced naming + error surfacing
  - userMessage: text fragments only (drops images/etc)
  - opaque items: no fabricated tool_calls
  - Helpers: deterministic id stability + sorted JSON args
  - Role alternation invariant across all four tool-shaped item types

This commit is a pure addition. AIAgent integration (the wire that uses the
projector) is the next commit.

* feat(codex-runtime): add session adapter + approval bridge

The third self-contained module: CodexAppServerSession owns one Codex
thread per Hermes session, drives turn/start, consumes streaming
notifications via CodexEventProjector, handles server-initiated approval
requests, and translates cancellation into turn/interrupt.

The adapter has a single public per-turn method:

    result = session.run_turn(user_input='...', turn_timeout=600)
    # result.final_text          → assistant text for the caller
    # result.projected_messages  → list ready to splice into AIAgent.messages
    # result.tool_iterations     → tick count for _iters_since_skill nudge
    # result.interrupted         → True on Ctrl+C / deadline / interrupt
    # result.error               → error string when the turn cannot complete
    # result.turn_id, thread_id  → for sessions DB / resume

Behavior:

  - ensure_started() spawns codex, does the initialize handshake, and
    issues thread/start with cwd + permissions profile. Idempotent.
  - run_turn() blocks until turn/completed, drains server-initiated
    requests (approvals) before reading notifications so codex never
    deadlocks waiting for us, projects every item/completed via the
    projector, and increments tool_iterations for the skill nudge gate.
  - request_interrupt() is thread-safe (threading.Event); the next loop
    iteration issues turn/interrupt and unwinds.
  - turn_timeout deadlock guard issues turn/interrupt and records an
    error if the turn never completes.
  - close() escalates terminate → kill via the underlying client.

Approval bridge:

  Codex emits server-initiated requests for execCommandApproval and
  applyPatchApproval. The adapter translates Hermes' approval choice
  vocabulary onto codex's decision vocabulary:

    Hermes 'once'                → codex 'approved'
    Hermes 'session' or 'always' → codex 'approvedForSession'
    Hermes 'deny' / anything else → codex 'denied'

  Routing precedence:
    1. _ServerRequestRouting.auto_approve_* flags (cron / non-interactive)
    2. approval_callback wired by the CLI (defers to
       tools.approval.prompt_dangerous_approval())
    3. Fail-closed denial when neither is wired

  Unknown server-request methods are answered with JSON-RPC error -32601
  so codex doesn't hang waiting for us.

Permission profile mapping mirrors AGENTS.md:
    Hermes 'auto'              → codex 'workspace-write'
    Hermes 'approval-required' → codex 'read-only-with-approval'
    Hermes 'unrestricted/yolo' → codex 'full-access'

20 new tests, all green. Combined with prior commits this PR now has
67 tests across three modules:
  - test_codex_app_server_runtime.py: 24 (api_mode + transport surface)
  - test_codex_event_projector.py: 23 (item taxonomy projections)
  - test_codex_app_server_session.py: 20 (turn loop + approvals + interrupts)

Full tests/agent/transports/ directory: 249/249 pass — no regressions
to existing transport tests.

Still no wire into AIAgent.run_conversation(); that integration commit
is small and goes next.

* feat(codex-runtime): wire codex_app_server runtime into AIAgent

The integration commit. AIAgent.run_conversation() now early-returns to a
new helper _run_codex_app_server_turn() when self.api_mode ==
'codex_app_server', bypassing the chat_completions tool loop entirely.

Three small surgical edits to run_agent.py (~105 LOC total):

1. Line ~1204 (constructor api_mode validation set):
   Add 'codex_app_server' so an explicit api_mode='codex_app_server'
   passed to AIAgent() isn't silently rewritten to 'chat_completions'.

2. Line ~12048 (run_conversation, just before the while loop):
   Early-return to _run_codex_app_server_turn() when self.api_mode is
   'codex_app_server'. Placed AFTER all standard pre-loop setup —
   logging context, session DB, surrogate sanitization, _user_turn_count
   and _turns_since_memory increments, _ext_prefetch_cache, memory
   manager on_turn_start — so behavior outside the model-call loop is
   identical between paths. Default Hermes flow is unchanged when the
   flag is off.

3. End-of-class (line ~15497):
   New method _run_codex_app_server_turn(). Lazy-instantiates one
   CodexAppServerSession per AIAgent (reused across turns), runs the
   turn, splices projected_messages into messages, increments
   _iters_since_skill by tool_iterations (since the chat_completions
   loop normally does that per iteration), fires
   _spawn_background_review on the same cadence as the default path.

Counter accounting:

  _turns_since_memory  ← already incremented at run_conversation:11817
                         (gated on memory store configured) — codex
                         helper does NOT touch it (would double-count).
  _user_turn_count     ← already incremented at run_conversation:11793
                         — codex helper does NOT touch it.
  _iters_since_skill   ← incremented in the chat_completions loop per
                         tool iteration. Codex helper increments by
                         turn.tool_iterations since the loop is bypassed.

User message:

  ALREADY appended to messages by run_conversation pre-loop (line 11823)
  before the early-return reaches us…

…backends (NousResearch#25126)

* feat(video_gen): unified video_generate tool with pluggable provider backends

One core video_generate tool, every backend a plugin. Mirrors the
image_gen + memory_provider + context_engine architecture: ABC, registry,
plugin-context registration hook, and per-plugin model catalogs surfaced
through hermes tools.

Surface (one schema, every backend):
- operation: generate / edit / extend
- modalities: text-to-video (prompt only), image-to-video (prompt +
  image_url), video edit (prompt + video_url), video extend (video_url)
- reference_image_urls, duration, aspect_ratio, resolution,
  negative_prompt, audio, seed, model override
- Providers ignore unknown kwargs and declare what they support via
  VideoGenProvider.capabilities() — backend-specific quirks stay in the
  backend, the agent learns one tool

Backends shipped:
- plugins/video_gen/xai/  — Grok-Imagine, full generate/edit/extend +
  image-to-video + reference images (salvaged from PR NousResearch#10600 by
  @Jaaneek, reshaped into the plugin interface)
- plugins/video_gen/fal/  — Veo 3.1 (t2v + i2v), Kling O3 i2v,
  Pixverse v6 i2v with model-aware payload building that drops keys a
  model doesn't declare

Wiring:
- agent/video_gen_provider.py — VideoGenProvider ABC, normalize_operation,
  success_response / error_response, save_b64_video / save_bytes_video,
  $HERMES_HOME/cache/videos/
- agent/video_gen_registry.py — thread-safe register/get/list +
  get_active_provider() reading video_gen.provider from config.yaml
- hermes_cli/plugins.py — PluginContext.register_video_gen_provider()
- hermes_cli/tools_config.py — Video Generation category in
  hermes tools, plugin-only providers list, model picker per plugin,
  config write to video_gen.{provider,model}
- toolsets.py — new video_gen toolset
- tests: 31 new tests covering ABC, registry, tool dispatch, both plugins
- docs: developer-guide/video-gen-provider-plugin.md (parallel to the
  image-gen guide), sidebar + toolsets-reference + plugin guides updated

Supersedes: NousResearch#25035 (FAL), NousResearch#17972 (FAL), NousResearch#14543 (xAI), NousResearch#13847 (HappyHorse),
NousResearch#10458 (provider categories), NousResearch#10786 (xAI media+search bundle), NousResearch#2984
(FAL duplicate), NousResearch#19086 (Google Veo standalone — easy port to plugin
interface).

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): dynamic schema reflects active backend's capabilities

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR NousResearch#22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

* test(video_gen/xai): full surface integration tests + cleaner schema

Verified end-to-end that the xAI plugin handles every documented mode
from PR NousResearch#10600's surface: text-to-video, image-to-video,
reference-images-to-video, video edit, video extend (with and without
prompt). All five modes route to the correct xAI endpoint
(/videos/generations, /videos/edits, /videos/extensions) with the right
payload shape (image / reference_images / video keys), and all five
client-side rejections fire before the network: edit-without-prompt,
extend-without-video_url, image+refs conflict, >7 references, and
duration/aspect_ratio clamping.

15 new integration tests grouped into four classes (endpoint routing,
modalities, validation, clamping). httpx is stubbed via a small fake
AsyncClient that records POSTs so the tests assert the actual payload
the plugin would send to xAI — not just the success/error envelope.

Also cleaned up a description redundancy: when a model's operations
match the backend's overall set, we no longer print the duplicate
'operations supported by this model' line. xAI's description now reads:

    Active backend: xAI . model: grok-imagine-video
    - operations supported by this backend: edit, extend, generate
    - modalities supported by this backend: image, reference_images, text
    - aspect_ratio choices: 16:9, 1:1, 2:3, 3:2, 3:4, 4:3, 9:16
    - resolution choices: 480p, 720p
    - duration range: 1-15s
    - reference_image_urls: up to 7 images

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): collapse surface to t2v + i2v, family-based auto-routing

Two design changes per Teknium:

1) Drop edit/extend from the tool surface entirely. Only text-to-video
and image-to-video remain. The agent sees a clean tool with two
modalities; backend-specific quirks like xAI's edit/extend endpoints
stay out of the unified schema.

2) FAL: pick a model FAMILY once, the plugin routes between the
family's text-to-video and image-to-video endpoints based on whether
image_url was passed. Users no longer pick 'fal-ai/veo3.1' AND
'fal-ai/veo3.1/image-to-video' as separate options — they pick
'veo3.1', and the plugin handles the rest.

Catalog rewritten as families:

    veo3.1            fal-ai/veo3.1                                /  fal-ai/veo3.1/image-to-video
    pixverse-v6       fal-ai/pixverse/v6/text-to-video             /  fal-ai/pixverse/v6/image-to-video
    kling-o3-standard fal-ai/kling-video/o3/standard/text-to-video /  fal-ai/kling-video/o3/standard/image-to-video

xAI uses a single endpoint (/videos/generations) for both modes,
routed by the presence of the 'image' field in the payload — no
edit/extend exposure.

Schema changes:
- VIDEO_GENERATE_SCHEMA: drop operation, drop video_url. Final params:
  prompt (required), image_url, reference_image_urls, duration,
  aspect_ratio, resolution, negative_prompt, audio, seed, model.
- VideoGenProvider ABC: drop normalize_operation, VALID_OPERATIONS,
  DEFAULT_OPERATION. capabilities() drops 'operations' key.
- success_response: add 'modality' field ('text' | 'image') so the
  agent and logs can see which endpoint was actually hit.

Dynamic schema builder simplified — no operations bullet, no
'switch backends if you need edit/extend' guidance. When the active
backend supports both modalities (the common case), description reads:

    Active backend: FAL . model: pixverse-v6
    - supports both text-to-video (omit image_url) and image-to-video
      (pass image_url) - routes automatically
    - aspect_ratio choices: 16:9, 9:16, 1:1
    - resolution choices: 360p, 540p, 720p, 1080p
    - duration range: 1-15s
    - audio: pass audio=true to enable native audio (pricing tier)
    - negative_prompt: supported

Tests: 51 in the video_gen slice, 216 across the broader image+video
sweep, all passing. New FAL routing tests prove pixverse-v6 + no image
hits text-to-video endpoint, pixverse-v6 + image_url hits
image-to-video endpoint, same for veo3.1 and kling-o3-standard.

Docs updated: developer-guide page rewrites the 'model families' pattern
as a first-class section so external plugin authors know the convention.
toolsets-reference and toolsets.py descriptions match the new surface.

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen/fal): expand catalog to 6 families, cheap + premium tiers

Catalog now covers everything Teknium specced from FAL:

  Cheap tier:
    ltx-2.3        fal-ai/ltx-2.3-22b/text-to-video       / image-to-video
    pixverse-v6    fal-ai/pixverse/v6/text-to-video       / image-to-video

  Premium tier:
    veo3.1         fal-ai/veo3.1                          / fal-ai/veo3.1/image-to-video
    seedance-2.0   bytedance/seedance-2.0/text-to-video   / image-to-video
    kling-v3-4k    fal-ai/kling-video/v3/4k/text-to-video / image-to-video
    happy-horse    fal-ai/happy-horse/text-to-video       / image-to-video

DEFAULT_MODEL moved from veo3.1 (premium) to pixverse-v6 (cheap, sane
defaults, both modalities) — better first-run UX for users who haven't
explicitly picked a model.

New family-entry knob: image_param_key. Kling v3 4K's image-to-video
endpoint expects start_image_url instead of image_url; declaring
image_param_key='start_image_url' on the family lets _build_payload
remap correctly. Other families default to plain image_url.

Per-family capability flags reflect each model's docs:
- LTX 2.3 + Happy Horse: minimal payloads (no duration/aspect/resolution
  enum exposed by FAL — let endpoint apply defaults)
- Seedance: 6 aspect ratios incl 21:9, durations 4-15, audio supported,
  negative prompts NOT supported per docs
- Kling v3 4K: 16:9/9:16/1:1, 3-15s, audio + negative
- Veo 3.1: unchanged, 16:9/9:16, 4/6/8s

Tests: +5 covering the new families (full catalog, Kling 4K
start_image_url remap, Seedance routing, LTX payload minimality, Happy
Horse minimality). 56/56 in the slice green.

Note: I did NOT add the FAL-hosted xAI Grok-Imagine variant. Hermes
already has a direct xAI plugin that talks to xAI's own API; routing
the same model through FAL's wrapper would duplicate the surface
without adding capabilities. Users on FAL who want Grok-Imagine should
use the xAI plugin directly; flag if you want both routes available.

* test(video_gen): tool-surface routing matrix — every model x modality

End-to-end matrix test driven through _handle_video_generate() — the
actual function the agent's video_generate tool call lands in. Writes
config.yaml, invokes the registered handler with a raw args dict, then
asserts the outbound HTTP/SDK call hit the right endpoint with the right
payload shape.

Parametrized over FAL_FAMILIES.keys() so the matrix auto-discovers new
families as they're added (add a family to FAL_FAMILIES and you get
both modalities tested for free).

Coverage:
- All 6 FAL families x {text-only, text+image} = 12 cases
- xAI x {text-only, text+image} = 2 cases
- tool-level model= arg overrides config = 2 cases

For each case, verifies:
- result['success'] is True
- result['modality'] matches input shape ('text' if no image_url, 'image' otherwise)
- outbound endpoint URL matches the family's text_endpoint or image_endpoint
- text-only payloads carry no image-shaped keys
- text+image payloads carry the family's image key (image_url for most,
  start_image_url for kling-v3-4k, wrapped 'image' object for xAI)

All 16 cases passing. Confirms the tool surface routes every
(provider, model, modality) combination correctly with zero leakage.

* feat(video_gen): keep video_gen out of first-run setup, surface in status

Two changes:

1. video_gen joins _DEFAULT_OFF_TOOLSETS, so it is NOT pre-selected in
   the first-run toolset checklist. Video gen is niche, paid, and slow —
   most users don't want it nagging them during initial setup. Anyone
   who wants it opts in via 'hermes tools' -> Video Generation, which
   already routes to the provider+model picker.

2. The 'hermes setup' status panel learns about video_gen — but only
   shows the row when a plugin reports available. Users without
   FAL_KEY/XAI_API_KEY see nothing about video gen; users with one of
   those keys see 'Video Generation (FAL) ✓' as confirmation it's wired.

Verified live:
- Fresh install (no creds): zero video_gen mentions in wizard.
- With FAL_KEY: status row appears with active backend name.
- 160/160 in the setup + tools_config + video_gen test slice.

Rationale: image_gen is on by default because it's a featured creative
tool used in casual chat (telegrams, etc). Video gen is heavier — long
wait, paid per-second pricing. Default-off matches user intent better.

---------

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.

…backends (NousResearch#25126)

* feat(video_gen): unified video_generate tool with pluggable provider backends

One core video_generate tool, every backend a plugin. Mirrors the
image_gen + memory_provider + context_engine architecture: ABC, registry,
plugin-context registration hook, and per-plugin model catalogs surfaced
through hermes tools.

Surface (one schema, every backend):
- operation: generate / edit / extend
- modalities: text-to-video (prompt only), image-to-video (prompt +
  image_url), video edit (prompt + video_url), video extend (video_url)
- reference_image_urls, duration, aspect_ratio, resolution,
  negative_prompt, audio, seed, model override
- Providers ignore unknown kwargs and declare what they support via
  VideoGenProvider.capabilities() — backend-specific quirks stay in the
  backend, the agent learns one tool

Backends shipped:
- plugins/video_gen/xai/  — Grok-Imagine, full generate/edit/extend +
  image-to-video + reference images (salvaged from PR NousResearch#10600 by
  @Jaaneek, reshaped into the plugin interface)
- plugins/video_gen/fal/  — Veo 3.1 (t2v + i2v), Kling O3 i2v,
  Pixverse v6 i2v with model-aware payload building that drops keys a
  model doesn't declare

Wiring:
- agent/video_gen_provider.py — VideoGenProvider ABC, normalize_operation,
  success_response / error_response, save_b64_video / save_bytes_video,
  $HERMES_HOME/cache/videos/
- agent/video_gen_registry.py — thread-safe register/get/list +
  get_active_provider() reading video_gen.provider from config.yaml
- hermes_cli/plugins.py — PluginContext.register_video_gen_provider()
- hermes_cli/tools_config.py — Video Generation category in
  hermes tools, plugin-only providers list, model picker per plugin,
  config write to video_gen.{provider,model}
- toolsets.py — new video_gen toolset
- tests: 31 new tests covering ABC, registry, tool dispatch, both plugins
- docs: developer-guide/video-gen-provider-plugin.md (parallel to the
  image-gen guide), sidebar + toolsets-reference + plugin guides updated

Supersedes: NousResearch#25035 (FAL), NousResearch#17972 (FAL), NousResearch#14543 (xAI), NousResearch#13847 (HappyHorse),
NousResearch#10458 (provider categories), NousResearch#10786 (xAI media+search bundle), NousResearch#2984
(FAL duplicate), NousResearch#19086 (Google Veo standalone — easy port to plugin
interface).

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): dynamic schema reflects active backend's capabilities

Address the 'capability variance' question — instead of one tool with a
static schema that lies about what every backend supports, the
video_generate tool now rebuilds its description at get_definitions()
time based on the configured video_gen.provider and video_gen.model.

The agent sees backend-specific guidance up-front:
- 'fal-ai/veo3.1/image-to-video': 'image-to-video only — image_url is
  REQUIRED; text-only prompts will be rejected'
- 'fal-ai/veo3.1' (t2v): no image_url restriction shown
- xAI grok-imagine-video: 'operations: generate, edit, extend; up to 7
  reference_image_urls'
- Backends without edit/extend: 'not supported on this backend — surface
  that they need to switch backends via hermes tools'

This is the same pattern PR NousResearch#22694 used for delegate_task self-capping —
documented in the dynamic-tool-schemas skill. Cache invalidation is
free: get_tool_definitions() already memoizes on config.yaml mtime, so a
mid-session backend swap rebuilds the schema automatically.

Tested:
- Empirical FAL OpenAPI schema check confirms image-to-video models
  require image_url (FAL returns HTTP 422 otherwise) — client-side
  rejection in FALVideoGenProvider.generate() now prevents the wasted
  round-trip
- Live E2E: fal-ai/veo3.1/image-to-video + prompt-only → clean
  missing_image_url error; fal-ai/veo3.1 + prompt-only → dispatches
- 6 new tests cover the builder (no config / image-only / full-surface /
  text-only / unknown provider / registry wiring), all passing
- 37/37 in the slice, 134/134 in the broader regression set

* test(video_gen/xai): full surface integration tests + cleaner schema

Verified end-to-end that the xAI plugin handles every documented mode
from PR NousResearch#10600's surface: text-to-video, image-to-video,
reference-images-to-video, video edit, video extend (with and without
prompt). All five modes route to the correct xAI endpoint
(/videos/generations, /videos/edits, /videos/extensions) with the right
payload shape (image / reference_images / video keys), and all five
client-side rejections fire before the network: edit-without-prompt,
extend-without-video_url, image+refs conflict, >7 references, and
duration/aspect_ratio clamping.

15 new integration tests grouped into four classes (endpoint routing,
modalities, validation, clamping). httpx is stubbed via a small fake
AsyncClient that records POSTs so the tests assert the actual payload
the plugin would send to xAI — not just the success/error envelope.

Also cleaned up a description redundancy: when a model's operations
match the backend's overall set, we no longer print the duplicate
'operations supported by this model' line. xAI's description now reads:

    Active backend: xAI . model: grok-imagine-video
    - operations supported by this backend: edit, extend, generate
    - modalities supported by this backend: image, reference_images, text
    - aspect_ratio choices: 16:9, 1:1, 2:3, 3:2, 3:4, 4:3, 9:16
    - resolution choices: 480p, 720p
    - duration range: 1-15s
    - reference_image_urls: up to 7 images

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen): collapse surface to t2v + i2v, family-based auto-routing

Two design changes per Teknium:

1) Drop edit/extend from the tool surface entirely. Only text-to-video
and image-to-video remain. The agent sees a clean tool with two
modalities; backend-specific quirks like xAI's edit/extend endpoints
stay out of the unified schema.

2) FAL: pick a model FAMILY once, the plugin routes between the
family's text-to-video and image-to-video endpoints based on whether
image_url was passed. Users no longer pick 'fal-ai/veo3.1' AND
'fal-ai/veo3.1/image-to-video' as separate options — they pick
'veo3.1', and the plugin handles the rest.

Catalog rewritten as families:

    veo3.1            fal-ai/veo3.1                                /  fal-ai/veo3.1/image-to-video
    pixverse-v6       fal-ai/pixverse/v6/text-to-video             /  fal-ai/pixverse/v6/image-to-video
    kling-o3-standard fal-ai/kling-video/o3/standard/text-to-video /  fal-ai/kling-video/o3/standard/image-to-video

xAI uses a single endpoint (/videos/generations) for both modes,
routed by the presence of the 'image' field in the payload — no
edit/extend exposure.

Schema changes:
- VIDEO_GENERATE_SCHEMA: drop operation, drop video_url. Final params:
  prompt (required), image_url, reference_image_urls, duration,
  aspect_ratio, resolution, negative_prompt, audio, seed, model.
- VideoGenProvider ABC: drop normalize_operation, VALID_OPERATIONS,
  DEFAULT_OPERATION. capabilities() drops 'operations' key.
- success_response: add 'modality' field ('text' | 'image') so the
  agent and logs can see which endpoint was actually hit.

Dynamic schema builder simplified — no operations bullet, no
'switch backends if you need edit/extend' guidance. When the active
backend supports both modalities (the common case), description reads:

    Active backend: FAL . model: pixverse-v6
    - supports both text-to-video (omit image_url) and image-to-video
      (pass image_url) - routes automatically
    - aspect_ratio choices: 16:9, 9:16, 1:1
    - resolution choices: 360p, 540p, 720p, 1080p
    - duration range: 1-15s
    - audio: pass audio=true to enable native audio (pricing tier)
    - negative_prompt: supported

Tests: 51 in the video_gen slice, 216 across the broader image+video
sweep, all passing. New FAL routing tests prove pixverse-v6 + no image
hits text-to-video endpoint, pixverse-v6 + image_url hits
image-to-video endpoint, same for veo3.1 and kling-o3-standard.

Docs updated: developer-guide page rewrites the 'model families' pattern
as a first-class section so external plugin authors know the convention.
toolsets-reference and toolsets.py descriptions match the new surface.

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

* feat(video_gen/fal): expand catalog to 6 families, cheap + premium tiers

Catalog now covers everything Teknium specced from FAL:

  Cheap tier:
    ltx-2.3        fal-ai/ltx-2.3-22b/text-to-video       / image-to-video
    pixverse-v6    fal-ai/pixverse/v6/text-to-video       / image-to-video

  Premium tier:
    veo3.1         fal-ai/veo3.1                          / fal-ai/veo3.1/image-to-video
    seedance-2.0   bytedance/seedance-2.0/text-to-video   / image-to-video
    kling-v3-4k    fal-ai/kling-video/v3/4k/text-to-video / image-to-video
    happy-horse    fal-ai/happy-horse/text-to-video       / image-to-video

DEFAULT_MODEL moved from veo3.1 (premium) to pixverse-v6 (cheap, sane
defaults, both modalities) — better first-run UX for users who haven't
explicitly picked a model.

New family-entry knob: image_param_key. Kling v3 4K's image-to-video
endpoint expects start_image_url instead of image_url; declaring
image_param_key='start_image_url' on the family lets _build_payload
remap correctly. Other families default to plain image_url.

Per-family capability flags reflect each model's docs:
- LTX 2.3 + Happy Horse: minimal payloads (no duration/aspect/resolution
  enum exposed by FAL — let endpoint apply defaults)
- Seedance: 6 aspect ratios incl 21:9, durations 4-15, audio supported,
  negative prompts NOT supported per docs
- Kling v3 4K: 16:9/9:16/1:1, 3-15s, audio + negative
- Veo 3.1: unchanged, 16:9/9:16, 4/6/8s

Tests: +5 covering the new families (full catalog, Kling 4K
start_image_url remap, Seedance routing, LTX payload minimality, Happy
Horse minimality). 56/56 in the slice green.

Note: I did NOT add the FAL-hosted xAI Grok-Imagine variant. Hermes
already has a direct xAI plugin that talks to xAI's own API; routing
the same model through FAL's wrapper would duplicate the surface
without adding capabilities. Users on FAL who want Grok-Imagine should
use the xAI plugin directly; flag if you want both routes available.

* test(video_gen): tool-surface routing matrix — every model x modality

End-to-end matrix test driven through _handle_video_generate() — the
actual function the agent's video_generate tool call lands in. Writes
config.yaml, invokes the registered handler with a raw args dict, then
asserts the outbound HTTP/SDK call hit the right endpoint with the right
payload shape.

Parametrized over FAL_FAMILIES.keys() so the matrix auto-discovers new
families as they're added (add a family to FAL_FAMILIES and you get
both modalities tested for free).

Coverage:
- All 6 FAL families x {text-only, text+image} = 12 cases
- xAI x {text-only, text+image} = 2 cases
- tool-level model= arg overrides config = 2 cases

For each case, verifies:
- result['success'] is True
- result['modality'] matches input shape ('text' if no image_url, 'image' otherwise)
- outbound endpoint URL matches the family's text_endpoint or image_endpoint
- text-only payloads carry no image-shaped keys
- text+image payloads carry the family's image key (image_url for most,
  start_image_url for kling-v3-4k, wrapped 'image' object for xAI)

All 16 cases passing. Confirms the tool surface routes every
(provider, model, modality) combination correctly with zero leakage.

* feat(video_gen): keep video_gen out of first-run setup, surface in status

Two changes:

1. video_gen joins _DEFAULT_OFF_TOOLSETS, so it is NOT pre-selected in
   the first-run toolset checklist. Video gen is niche, paid, and slow —
   most users don't want it nagging them during initial setup. Anyone
   who wants it opts in via 'hermes tools' -> Video Generation, which
   already routes to the provider+model picker.

2. The 'hermes setup' status panel learns about video_gen — but only
   shows the row when a plugin reports available. Users without
   FAL_KEY/XAI_API_KEY see nothing about video gen; users with one of
   those keys see 'Video Generation (FAL) ✓' as confirmation it's wired.

Verified live:
- Fresh install (no creds): zero video_gen mentions in wizard.
- With FAL_KEY: status row appears with active backend name.
- 160/160 in the setup + tools_config + video_gen test slice.

Rationale: image_gen is on by default because it's a featured creative
tool used in casual chat (telegrams, etc). Video gen is heavier — long
wait, paid per-second pricing. Default-off matches user intent better.

---------

Co-authored-by: Jaaneek <Jaaneek@users.noreply.github.com>

…n tool description (NousResearch#22694)

The delegate_task tool description hardcoded 'default 3' / 'default 2' for
max_concurrent_children / max_spawn_depth, which misled the model on any
install that raised these limits — the schema text said 'default 3' even
when the user had set max_concurrent_children=15 / max_spawn_depth=3, so
the model would self-cap at 3 and never use the headroom.

Make the description dynamic. ToolEntry gains an optional
dynamic_schema_overrides callable; registry.get_definitions() merges its
output on top of the static schema before returning it. delegate_tool
registers a builder that reads the current delegation.* config and emits:

- 'up to N items concurrently for this user' (N = max_concurrent_children)
- 'Nested delegation IS enabled / OFF for this user (max_spawn_depth=N)'
- 'orchestrator children can themselves delegate up to M more level(s)'
- 'orchestrator_enabled=false' when the kill switch is set

The model_tools cache key already includes config.yaml mtime+size, so
edits to delegation.* in config invalidate the cached tool definitions
without an explicit hook. CLI_CONFIG staleness within a process is a
pre-existing limitation of _load_config and out of scope here.

Static description / tasks.description / role.description in
DELEGATE_TASK_SCHEMA are placeholders so module import doesn't trigger
cli.CLI_CONFIG load before the test conftest can redirect HERMES_HOME.