Adds Vercel Sandbox as a supported Hermes terminal backend alongside
existing providers (Local, Docker, Modal, SSH, Daytona, Singularity).
Uses the Vercel Python SDK to create/manage cloud microVMs, supports
snapshot-based filesystem persistence keyed by task_id, and integrates
with the existing BaseEnvironment shell contract and FileSyncManager
for credential/skill syncing.
Based on #17127 by @scotttrinh, cherry-picked onto current main.
The cron schema contracts deliver as a string ("local", "origin",
"telegram", "telegram:chat_id[:thread_id]", or comma-separated combos),
but MCP clients and scripts sometimes pass an array like ['telegram'].
Before this change, the list was written to jobs.json verbatim, and
the scheduler's str(deliver).split(',') then tried to resolve the
literal string "['telegram']" as a platform — returning None and
logging 'no delivery target resolved for deliver=[\'telegram\']'.
Fix on both ends:
- tools/cronjob_tools.py: normalize deliver at the API boundary on
create and update, so storage is always a string.
- cron/scheduler.py: normalize deliver in _resolve_delivery_targets,
so existing jobs.json entries with list-form deliver are handled
gracefully without requiring users to edit the file.
Closes#17139
Widen #17163 to the sibling file tools/transcription_tools.py, which had
the same class of bug. STT provider call sites and the _get_provider
selection gate called os.getenv(...) directly and missed keys that only
lived in ~/.hermes/.env.
Same pattern as tts_tool.py: one guarded top-level import of
get_env_value (falls back to os.getenv on ImportError), then every
API-key and paired-base-URL lookup swapped over.
Call sites migrated:
- _transcribe_groq — GROQ_API_KEY
- _transcribe_mistral — MISTRAL_API_KEY
- _transcribe_xai — XAI_API_KEY, XAI_STT_BASE_URL
- _get_provider — GROQ/MISTRAL/XAI_API_KEY in explicit + auto branches
Module-level defaults (DEFAULT_STT_MODEL, GROQ_BASE_URL, etc.) stay on
os.getenv — they're import-time constants, not runtime config, and the
dotenv fallback would add no value there.
New regression tests in tests/tools/test_transcription_dotenv_fallback.py
(8 cases) mirror briandevans' TTS tests: per-provider dotenv-key
forwarding, selection-gate dotenv visibility, and an end-to-end probe
that patches hermes_cli.config.load_env to simulate ~/.hermes/.env
carrying the key while os.environ does not.
Wrap the new top-level `from hermes_cli.config import get_env_value`
in try/except ImportError and fall back to a thin os.getenv shim, so
importing tools.tts_tool keeps working in environments where
hermes_cli.config is unavailable. This matches the existing tolerance
in `_load_tts_config()` (tools/tts_tool.py) and the same
import-fallback pattern in tools/tool_backend_helpers.py::fal_key_is_configured.
Also update the TestDotenvFallbackPerProvider docstring to accurately
describe the mocking strategy: per-provider tests patch
`tools.tts_tool.get_env_value` directly, while the regression-guard
tests cover the lower-level `hermes_cli.config.load_env` integration.
Addresses Copilot review on #17163.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
TTS provider tools (elevenlabs, xai, minimax, mistral, gemini) called
os.getenv("X_API_KEY") directly, which bypassed Hermes's dotenv bridge in
hermes_cli.config. Users who keep their TTS keys only in ~/.hermes/.env saw
"X_API_KEY not set" errors even though the rest of the stack
(agent/credential_pool, hermes_cli/auth) already resolves keys through
get_env_value() — same class of bug as #15914 fixed for those modules.
Switch every TTS env-var lookup (API keys, base URLs, and
check_tts_requirements gates) to get_env_value, which checks os.environ
first and then ~/.hermes/.env. Behaviour for users with keys exported in
the shell is unchanged; users with dotenv-only keys now succeed. The two
diagnostics prints in __main__ are migrated for consistency.
Regression test (tests/tools/test_tts_dotenv_fallback.py):
- per-provider: each backend reads the dotenv key when only
~/.hermes/.env carries it (5 providers).
- end-to-end: with hermes_cli.config.load_env returning the key and
os.environ empty, _generate_minimax_tts and check_tts_requirements
both succeed; reverting tools/tts_tool.py back to os.getenv makes all
7 tests fail with "MINIMAX_API_KEY not set" / similar.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
_update_cwd() uses a bare open(self._cwd_file).read() that never
closes the file descriptor. This method runs on every terminal
command execution, so the fd leaks accumulate in long sessions.
Use a with statement so the fd is released promptly.
Fixes#15552 (standalone resubmission)
When a background terminal process spawns a descendant daemon that
inherits the stdout pipe (e.g. 'hermes update' triggering a gateway
systemctl restart), the reader thread's stdout.read() never returns EOF
and its finally: block never runs. session.exited stays False forever,
so process(action='poll') returns 'running' indefinitely even though
the direct child exited long ago.
Issue #17327: Feishu user polled 74 times over 7 minutes before killing
the gateway manually.
Fix: add _reconcile_local_exit() that checks the direct Popen.poll()
before trusting session.exited. If the direct child has exited, drain
any immediately-readable bytes non-blocking and flip session.exited.
Called from poll() and wait(). The stuck reader thread remains blocked
but is a daemon thread and gets reaped with the process.
Safe no-op for env/PTY sessions, already-exited sessions, and live
children (returns None from Popen.poll()).
_send_yuanbao() already supported media_files= and the user-facing
error strings already advertised yuanbao support, but there was no
dispatch branch in _send_to_platform() actually routing to it. Target
yuanbao in send_message previously fell through to
"Direct sending not yet implemented".
- Add yuanbao media-chunk branch (mirrors Signal/Matrix: media on
final chunk only).
- Add yuanbao elif in the non-media loop.
Salvage of #17411; SKILL.md description change and redundant
sidebars.ts entry dropped, indentation/trailing-whitespace cleaned up.
The "cfg.get('X', {}).get('Y', default)" pattern appears 50+ times
across tools/, gateway/, and plugins/. Each call site manually handles
the same three gotchas:
1. Missing intermediate key → empty dict → chain works
2. Non-dict value at intermediate position → AttributeError
(uncaught in most sites, so a misconfigured YAML crashes the tool)
3. cfg is None → AttributeError
Introduces cfg_get(cfg, *keys, default=None) in hermes_cli/config.py
as the canonical helper. Handles all three uniformly, returns default
only when the final key is *absent* (matches dict.get semantics —
explicit None values are preserved, falsy values like 0 / False / ''
are preserved).
Named cfg_get rather than cfg_path to avoid shadowing the existing
'cfg_path = _hermes_home / "config.yaml"' local variable that appears
in gateway/run.py, cron/scheduler.py, hermes_cli/main.py, etc.
Migrated 20 call sites as the first-batch proof-of-value:
gateway/run.py 10 sites (agent/display subtrees)
tools/browser_tool.py 3 sites
tools/vision_tools.py 2 sites
tools/browser_camofox.py 1 site
tools/approval.py 1 site
tools/skills_tool.py 1 site
tools/skill_manager_tool.py 1 site
tools/credential_files.py 1 site
tools/env_passthrough.py 1 site
The remaining ~30 sites across plugins/ and smaller tool files can be
migrated opportunistically — the helper is now available and the
pattern is established.
Fixed a latent bug along the way: tools/vision_tools.py had its
cfg_get usage at line 560 inside a function that locally re-imports
'from hermes_cli.config import load_config', but the AST-based
migration script wrote the top-level cfg_get import to a different
function scope, leaving line 560's cfg_get as a NameError silently
swallowed by the surrounding try/except. Test
test_vision_uses_configured_temperature_and_timeout caught it. Fixed
by including cfg_get in the function-local import.
Verified:
- 7880/7893 tests/tools/ + tests/gateway/ + tests/hermes_cli/test_config
tests pass; all 13 failures pre-existing on main (MCP, delegate,
session_split_brain — verified earlier in the sweep).
- All 20 migrated sites AST-verified to have cfg_get in scope (either
module-level or function-local).
- Live 'hermes chat' smoke: 2 turns + /model switch + tool calls +
/quit, zero errors. Agent correctly counted 20 cfg_get hits across
8 tool files — matching the migration.
Semantic parity verified against the original pattern across 8 edge
cases (missing keys, None values, falsy values, empty strings, string
instead of dict, None cfg, nested levels).
Add opt-in terminal.docker_run_as_host_user config flag that passes
--user $(id -u):$(id -g) to the Docker backend so files written into
bind-mounted directories (/workspace, /root, docker_volumes entries) are
owned by the host user instead of root.
When enabled on POSIX platforms, also drops SETUID/SETGID caps since the
container no longer needs gosu/su to switch users. Falls back cleanly on
platforms without os.getuid (e.g. native Windows Docker) with a warning.
Wired through all three config.yaml -> TERMINAL_* env-var bridges:
- cli.py env_mappings (CLI + TUI startup)
- gateway/run.py _terminal_env_map (gateway / messaging platforms)
- hermes_cli/config.py _config_to_env_sync (`hermes config set`)
Also fixes docker_mount_cwd_to_workspace silently failing in gateway
mode -- it was missing from gateway/run.py's _terminal_env_map.
Adds tests/tools/test_terminal_config_env_sync.py to guard against
future drift between the three bridges (same bug class shipped twice
in one month).
Bundled Hermes image won't work with this flag since its entrypoint
expects to start as root for the usermod/gosu hermes flow; works with
the default nikolaik/python-nodejs image and plain Debian/Ubuntu.
Previous invariants only gated the primary entry points
(apply_automatic_transitions, archive_skill, CLI pin). Several paths
were unprotected:
- bump_view / bump_use / bump_patch / set_state / set_pinned wrote
usage records unconditionally, which is confusing noise in
.usage.json even though the review list filtered them out
- restore_skill did not check whether a bundled skill now shadows
the archived name
- CLI unpin was asymmetric with CLI pin — it had no gate
Fixes:
- _mutate() (the shared counter / state writer) now drops silently
when the skill is not agent-created. .usage.json never gains a
record for a bundled or hub-installed skill.
- restore_skill() refuses to restore under a name that is now
bundled or hub-installed (would shadow upstream).
- CLI unpin gate matches CLI pin.
New tests:
- 5 provenance-guard tests on skill_usage (one per mutator)
- 1 end-to-end test that hammers every mutator at a bundled skill
and a hub skill, asserts both are untouched on disk, and asserts
the sidecar stays clean
- 2 CLI tests proving pin/unpin refuse bundled skills symmetrically
64/64 tests passing (29 skill_usage + 27 curator + 8 new guards).
Adds the Curator — an auxiliary-model background task that periodically
reviews AGENT-CREATED skills and keeps the collection tidy: tracks usage,
transitions unused skills through active → stale → archived, and spawns
a forked AIAgent to consolidate overlaps and patch drift.
Default: enabled, inactivity-triggered (no cron daemon). Runs on CLI
startup and gateway boot when the last run is older than interval_hours
(default 24) AND the agent has been idle for min_idle_hours (default 2).
Invariants (all load-bearing):
- Never touches bundled or hub-installed skills (.bundled_manifest +
.hub/lock.json double-filter)
- Never auto-deletes — archive only. Archives are recoverable
via `hermes curator restore <skill>`
- Pinned skills bypass all auto-transitions
- Uses the aux client; never touches the main session's prompt cache
New files:
- tools/skill_usage.py — sidecar .usage.json telemetry, atomic writes,
provenance filter
- agent/curator.py — orchestrator: config, idle gating, state-machine
transitions (pure, no LLM), forked-agent review prompt
- hermes_cli/curator.py — `hermes curator {status,run,pause,resume,
pin,unpin,restore}` subcommand
- tests/tools/test_skill_usage.py — 29 tests
- tests/agent/test_curator.py — 25 tests
Modified files (surgical patches):
- tools/skills_tool.py — bump view_count on successful skill_view
- tools/skill_manager_tool.py — bump patch_count on skill_manage
patch/edit/write_file/remove_file; forget record on delete
- hermes_cli/config.py — add curator: section to DEFAULT_CONFIG
- hermes_cli/commands.py — add /curator CommandDef with subcommands
- hermes_cli/main.py — register `hermes curator` subparser via
register_cli() from hermes_cli.curator
- cli.py — /curator slash-command dispatch + startup hook
- gateway/run.py — gateway-boot hook (mirrors CLI)
Validation:
- 54 new tests across skill_usage + curator, all passing in 3s
- 346 tests across all touched files' neighbors green
- 2783 tests across hermes_cli/ + gateway/test_run_progress_topics.py green
- CLI smoke: `hermes curator status/pause/resume` work end-to-end
Companion to PR #16026 (class-first skill review prompt) — together
they form a loop: the review prompt stops near-duplicate skill creation
at the source, and the curator prunes/consolidates what still accumulates.
Refs #7816.
Commit 3c42064e made config.yaml the single source of truth for
TERMINAL_CWD, but the config bridge passes cwd values verbatim to
os.environ. When a user sets terminal.cwd: ~/ in config.yaml, the
literal string '~/'' reaches subprocess.Popen, which the kernel
rejects because it does not expand shell tilde syntax.
This patch adds three defensive layers:
1. gateway/run.py — expanduser at config bridge time so TERMINAL_CWD
is always an absolute path.
2. tools/terminal_tool.py — expanduser when reading TERMINAL_CWD in
_get_env_config(), guarding against stale or manually-set env vars.
3. tools/environments/local.py — expanduser in LocalEnvironment before
passing cwd to subprocess.Popen, the final safety net.
Includes regression tests in test_config_cwd_bridge.py for nested
terminal.cwd, top-level cwd alias, and precedence ordering.
Refs: 3c42064e
init_session() runs a login shell bootstrap that sources profile scripts
(.bashrc, .bash_profile, etc.) before capturing pwd. If any profile
script changes the working directory, the captured cwd overwrites the
configured terminal.cwd value — so terminal commands run in the wrong
directory despite the TUI banner showing the configured path.
Add an explicit 'builtin cd' to the configured cwd in the bootstrap
script, after profile sourcing but before pwd capture, ensuring the
configured terminal.cwd is always what gets recorded.
Fixes#14044
TUI session readiness was still laggy after the gateway-ready fixes. Profiling
session.create -> session.info showed the slow phase is background AIAgent
construction (~1.1s). A cProfile run of tui_gateway.server::_make_agent showed
model_tools/tool discovery importing tools.code_execution_tool, whose
module-level EXECUTE_CODE_SCHEMA calls _get_execution_mode(), which imported
cli.CLI_CONFIG.
That pulled the classic interactive CLI stack (prompt_toolkit/Rich and REPL
setup) into every agent startup path, including hermes --tui where it is not
used. Replace that with hermes_cli.config.read_raw_config(), which is cached and
reads only the raw code_execution section. Existing defaults still apply when
the key is absent.
Measurements on macOS Terminal.app:
- import run_agent: ~466ms -> ~347ms
- model_tools import: ~418ms -> ~272ms
- _make_agent: ~1452ms -> ~1239ms
- session.create -> session.info: ~1167ms -> ~999ms
- full hermes --tui ready p50: ~1655ms -> ~1537ms
Tests:
- scripts/run_tests.sh tests/tools/test_code_execution_modes.py tests/tools/test_code_execution.py
detect_dangerous_command() and detect_hardline_command() were calling
re.search(pattern, text, re.IGNORECASE | re.DOTALL) inline — Python's
re._cache (512 patterns) amortizes compile cost on the warm path, but:
1. The first terminal() call per process pays the full compile fan-out
for all 59 patterns (12 HARDLINE + 47 DANGEROUS). Measured at
~2.6 ms per detect_dangerous_command() call after re.purge().
2. The re._cache is LRU — unrelated regex work elsewhere in the agent
(response parsing, text normalization, etc.) can evict our patterns
and silently re-compile them on the next terminal() call.
Precompiling at module load eliminates both costs:
detect_dangerous_command:
cold 2.613 ms → 0.298 ms (-88%)
warm 0.042 ms → 0.004 ms (-90%)
detect_hardline_command:
cold ~0.6 ms → 0.006 ms
warm 0.011 ms → 0.002 ms
Savings are per terminal() call. Agents with heavy terminal use see
compound savings; the bigger value is the stability guarantee (no
re._cache eviction can silently re-introduce the 2.6 ms cold cost
mid-session).
Implementation:
- HARDLINE_PATTERNS_COMPILED and DANGEROUS_PATTERNS_COMPILED built at
module load from the existing (pattern, description) tuples, using
shared _RE_FLAGS = re.IGNORECASE | re.DOTALL.
- detect_* functions now iterate the compiled list and call pattern_re.search(text).
- Original HARDLINE_PATTERNS and DANGEROUS_PATTERNS lists kept as-is
(other code in the file uses them for key derivation /
_PATTERN_KEY_ALIASES).
Verified:
- 160/161 tests/tools/test_approval*.py pass (1 pre-existing heartbeat
test flake on main).
- 349/349 tests/tools/ 'approval or terminal or dangerous' pass.
- Live hermes chat smoke: 3 benign terminal commands + 1 rm -rf /tmp/
(clarify prompt fired — approval path still works) + 1 sudo (sudo
password prompt fired — DANGEROUS pattern match still works). 23
log lines in the smoke window, zero errors.
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
Two amplifying optimizations to per-turn overhead in the gateway:
1. get_tool_definitions() memoization (model_tools.py)
Keyed on (frozenset(enabled), frozenset(disabled),
registry._generation, config.yaml mtime+size). Only active when
quiet_mode=True (which is every hot-path caller — gateway,
AIAgent.__init__); quiet_mode=False keeps the existing print side
effects. Cached path returns a shallow-copy list sharing read-only
schema dicts.
Measured: 7.5 ms → 0.01 ms per call (~750× speedup). Gateway
constructs fresh AIAgent per message, so this saves ~7 ms/turn before
any LLM work.
2. check_fn() TTL cache (tools/registry.py)
check_fn callables like check_terminal_requirements probe external
state (Docker daemon, Modal SDK, playwright binary). For a long-lived
process, hitting them on every get_definitions() pass was pure waste
— external state changes on human timescales. 30 s TTL so env-var
flips (hermes tools enable X) propagate within a turn or two without
explicit invalidation.
Measured: first call 7.5ms → 1.6ms (check_fn probes now dominate);
subsequent calls ~0.01ms via the upstream memoization.
Invalidation surface:
- registry._generation bumps on register/deregister/register_toolset_alias,
invalidating the memoized definitions automatically.
- config.yaml mtime in the cache key captures user-visible config edits
affecting dynamic schemas (execute_code mode, discord allowlist).
- invalidate_check_fn_cache() exposed for explicit flushes (e.g. after
hermes tools enable/disable).
- tests/conftest.py autouse fixture clears both caches before every test
so env-var monkeypatches don't see stale results.
Also fixes a regression from PR #17046 that I missed:
- tools/web_tools.py — Firecrawl was removed from module scope by the
lazy import, breaking 8 tests that patch 'tools.web_tools.Firecrawl'.
Applied the same _FirecrawlProxy pattern used in auxiliary_client/
run_agent for OpenAI (module-level proxy that looks like the class
but imports the SDK on first call/isinstance; patch() replaces the
attribute as usual).
Verified:
- 49/49 tests/tools/test_web_tools_config.py pass (was 8 failing on main)
- 68/68 tests/tools/test_homeassistant_tool.py pass (was 1 failing in
the full suite due to check_fn TTL cross-test pollution; fixed by
the autouse fixture)
- 3887/3895 tests/tools/ (8 pre-existing fails: 2 delegate, 1 mcp
dynamic discovery, 5 mcp structured content — all confirmed on main)
- 2973/2976 tests/agent/ + tests/run_agent/ (3 pre-existing fails)
- 868/868 tests/run_agent/ (excluding test_run_agent.py which has
pre-existing suite-level issues)
- Live smoke: 2 turns + /model switch + tool calls, zero errors in
agent.log session window.
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
* perf(startup): lazy-import OpenAI, Anthropic, Firecrawl, account_usage
Four heavy SDK/module imports are now deferred off the hot startup path.
Net savings on cold module imports:
cli 1200 → 958 ms (-242)
run_agent 1220 → 901 ms (-319)
tools.web_tools 711 → 423 ms (-288)
agent.anthropic_adapter 230 → 15 ms (-215)
agent.auxiliary_client 253 → 68 ms (-185)
Four independent changes in one PR since they all use the same pattern
and share the same risk profile (heavy SDK import → lazy proxy or
function-local import):
1. tools/web_tools.py:
'from firecrawl import Firecrawl' moved into _get_firecrawl_client(),
which is only called when backend='firecrawl'. Users on Exa/Tavily/
Parallel pay zero firecrawl cost.
2. cli.py + gateway/run.py:
'from agent.account_usage import ...' moved into the /limits handlers.
account_usage transitively pulls the OpenAI SDK chain; only needed
when the user runs /limits.
3. agent/anthropic_adapter.py:
'try: import anthropic as _anthropic_sdk' replaced with a cached
'_get_anthropic_sdk()' accessor. The three usage sites
(build_anthropic_client, build_anthropic_bedrock_client,
read_claude_code_credentials_from_keychain) now resolve via the
accessor. All pre-existing test patches of
'agent.anthropic_adapter._anthropic_sdk' keep working because the
accessor respects any value already in module globals.
4. agent/auxiliary_client.py AND run_agent.py:
'from openai import OpenAI' replaced with an '_OpenAIProxy()' module-
level object that looks like the OpenAI class but imports the SDK on
first call/isinstance check. This preserves:
- 15+ in-module OpenAI(...) construction sites in auxiliary_client
and the single site in run_agent's _create_openai_client (Python's
function-scope name lookup finds the proxy, forwards the call);
- 'patch("agent.auxiliary_client.OpenAI", ...)' and
'patch("run_agent.OpenAI", ...)' test patterns used by 28+ test
files (patch replaces the module attribute as usual).
Tried two alternatives first:
- 'from openai._client import OpenAI' — doesn't skip openai/__init__.py
(the audit's hypothesis here was wrong).
- Module-level __getattr__ — works for external access but Python
function-scope name resolution skips __getattr__, so in-module
OpenAI(...) calls NameError.
Note: 'openai' still loads on 'import cli' because
cli.py -> neuter_async_httpx_del() -> openai._base_client, and
run_agent.py -> code_execution_tool.py (module-level
build_execute_code_schema) -> _load_config() -> 'from cli import
CLI_CONFIG'. Deferring those is a separate, larger change — out of scope
for this PR. The savings above all come from avoiding the openai/*,
anthropic/*, and firecrawl/* top-level type-tree imports on paths that
don't need them.
Verified:
- 302/302 tests in tests/agent/{test_anthropic_adapter,
test_bedrock_1m_context, test_minimax_provider, test_anthropic_keychain}
pass. Two pre-existing failures on main unchanged.
- 106/106 tests/agent/test_auxiliary_client.py pass (1 pre-existing fail).
- 97/97 tests/run_agent/test_create_openai_client_kwargs_isolation.py,
test_plugin_context_engine_init.py, test_invalid_context_length_warning.py,
test_api_max_retries_config.py,
tests/hermes_cli/test_gemini_provider.py, test_ollama_cloud_provider.py
pass (1 pre-existing fail).
- Live hermes chat smoke: 2 turns + /model switch + tool calls, zero
errors in the 57-line agent.log window.
- Module-level import of run_agent + auxiliary_client + anthropic_adapter
no longer pulls 'anthropic' or 'firecrawl' at all.
* fix(gateway): restore top-level account_usage import for test-patch surface
CI caught two failures in tests/gateway/test_usage_command.py that I
missed locally:
AttributeError: 'module' object at gateway.run has no attribute 'fetch_account_usage'
The test uses monkeypatch.setattr('gateway.run.fetch_account_usage', ...)
to inject a fake account-fetch call. Moving the import inside the
handler deleted that module-level attribute, breaking the patch surface.
Restoring the top-level import in gateway/run.py gives up the ~230 ms
gateway-boot savings from that one lazy, but:
1. the gateway is a long-running daemon — boot cost is paid once per
install, not per turn;
2. the other four lazy-imports (firecrawl, openai, anthropic, cli's
account_usage) remain in place and still account for the bulk of
the savings reported in the PR body;
3. preserving the patch surface keeps the established
'gateway.run.fetch_account_usage' monkeypatch pattern working
without touching tests.
Verified: tests/gateway/test_usage_command.py — 8 passed, 0 failed.
Full targeted sweep (2336 tests across agent/gateway/hermes_cli/run_agent):
2332 passed, 4 failed — all 4 pre-existing on main.
---------
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
Previously, check_browser_requirements() only checked for the agent-browser
CLI, not the Chromium binary it drives. When the CLI was present but
Chromium wasn't (common in Docker images predating the playwright install
step), the browser tool was advertised to the agent, every call hung for
the full command timeout (~30s each, ~220s for a chained navigate), and
the agent eventually gave up with no useful error — users saw 'browser
not working' with empty errors.log.
Changes:
- tools/browser_tool.py: add _chromium_installed() checking
PLAYWRIGHT_BROWSERS_PATH + default Playwright cache paths for
chromium-* / chromium_headless_shell-* dirs; wire into
check_browser_requirements() for local mode (cloud providers
unaffected). _run_browser_command fails fast with an actionable
Docker vs. host message instead of hanging. _running_in_docker()
checks /.dockerenv and /proc/1/cgroup.
- hermes_cli/tools_config.py: post_setup for 'Local Browser' now runs
'agent-browser install --with-deps' after npm install to actually
download Chromium. In Docker, points user at the updated image pull
instead of trying to install into a read-only layer. Cloud-provider
post_setup (browserbase) skips Chromium install entirely.
- tests/tools/test_browser_chromium_check.py: new tests covering
search roots, install detection, requirements branches (local/cloud/
camofox), and the fast-fail guard in docker/non-docker contexts.
- tests/tools/test_browser_homebrew_paths.py: 5 existing subprocess-path
tests now mock _chromium_installed=True since they exercise the
post-guard subprocess path.
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
Mechanical cleanup across 43 files — removes 46 unused imports
(F401) and 14 unused local variables (F841) detected by
`ruff check --select F401,F841`. Net: -49 lines.
Also fixes a latent NameError in rl_cli.py where `get_hermes_home()`
was called at module line 32 before its import at line 65 — the
module never imported successfully on main. The ruff audit surfaced
this because it correctly saw the symbol as imported-but-unused
(the call happened before the import ran); the fix moves the import
to the top of the file alongside other stdlib imports.
One `# noqa: F401` kept in hermes_cli/status.py for `subprocess`:
tests monkeypatch `hermes_cli.status.subprocess` as a regression
guard that systemctl isn't called on Termux, so the name must
exist at module scope even though the module body doesn't reference
it. Docstring explains the reason.
Also fixes an invalid `# noqa:` directive in
gateway/platforms/discord.py:308 that lacked a rule code.
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
delegate_task runs inside the parent turn and is cancelled when the parent is interrupted (new user message, /stop, /new). The child status payload (status=interrupted, exit_reason=interrupted) is already honest, but the tool schema and user-facing docs did not set the expectation, so users reasonably assumed delegated subagents would keep running in the background after interrupting the parent.
Updates:
- tools/delegate_tool.py DELEGATE_TASK_SCHEMA description adds a WHEN NOT TO USE bullet pointing at cronjob / terminal(background=True, notify_on_complete=True) for durable long-running work.
- website/docs/user-guide/features/delegation.md gains a Lifetime and Durability callout above Key Properties.
- website/docs/guides/delegation-patterns.md expands the Use something else list and the Constraints section with the same guidance.
Reported by LizLiz (@lizliz404) via Teknium.
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
Extract the islink/realpath guard from the 16743 fix into a single
atomic_replace() helper in utils.py, then migrate every os.replace()
call site in the codebase to use it.
The original PR #16777 correctly identified and fixed the bug, but
only patched 9 of ~24 call sites. The same bug class (managed
deployments that symlink state files silently losing the link on
every write) still existed at auth.json, sessions file, gateway
config, env_loader, webhook subscriptions, debug store, model
catalog, pairing, google OAuth, nous rate guard, and more.
Rather than add another 10+ copies of the same three-line guard,
consolidate into atomic_replace(tmp, target) which:
- resolves symlinks via os.path.realpath before os.replace
- returns the resolved real path so callers can re-apply permissions
- is a drop-in replacement for os.replace at the use sites
Changes:
- utils.py: new atomic_replace() helper + atomic_json_write /
atomic_yaml_write now call it instead of inlining the guard
- 16 files: all os.replace() call sites migrated to atomic_replace()
- agent/{google_oauth, nous_rate_guard, shell_hooks}.py
- cron/jobs.py
- gateway/{pairing, session, platforms/telegram}.py
- hermes_cli/{auth, config, debug, env_loader, model_catalog, webhook}.py
- tools/{memory_tool, skill_manager_tool, skills_sync}.py
Tests: tests/test_atomic_replace_symlinks.py pins the invariant for
atomic_replace + atomic_json_write + atomic_yaml_write, covers plain
files, first-time creates, broken symlinks, and permission preservation.
Refs #16743
Builds on #16777 by @vominh1919.
os.replace(tmp, path) replaces the symlink itself with a regular file,
breaking users who symlink config.yaml, SOUL.md, or .env from ~/.hermes/
to a dotfiles repo or managed profile package.
Fix: resolve symlinks via os.path.realpath() before os.replace(), so the
real file is overwritten in-place while the symlink survives.
Fixed in 7 files covering all os.replace call sites:
- utils.py (atomic_json_write, atomic_yaml_write — fixes save_config)
- hermes_cli/config.py (env sanitizer, save_env_value, remove_env_value)
- tools/skill_manager_tool.py (_atomic_write_text — SOUL.md writes)
- tools/memory_tool.py (memory file writes)
- tools/skills_sync.py (manifest writes)
- cron/jobs.py (job state + output file writes)
- agent/shell_hooks.py (hook file writes)
FixesNousResearch/hermes-agent#16743
Adds tools.schema_sanitizer.strip_nullable_unions as the single
implementation for collapsing anyOf/oneOf nullable unions. Both the
MCP input-schema normalizer and the Anthropic tool-schema guard now
delegate to it instead of re-implementing the same walk three times.
The global sanitizer also gains a final pass so any tool that slips
past the two earlier hooks (plugin tools, non-MCP custom tools with
Pydantic-shaped schemas) still gets safe input_schemas on Anthropic.
- tools/schema_sanitizer.py:
* New public strip_nullable_unions(schema, keep_nullable_hint=True).
* _sanitize_single_tool() calls it as a final pass (hint preserved
so coerce_tool_args can still map string "null" to None).
- tools/mcp_tool.py: _normalize_mcp_input_schema delegates.
- agent/anthropic_adapter.py: _normalize_tool_input_schema delegates
with keep_nullable_hint=False (Anthropic does not recognize nullable).
No behavioral change for the fix itself; tests (73/73 targeted +
E2E across MCP→sanitizer→Anthropic paths) pass.
Extends the cua-driver computer-use backend to drive backgrounded macOS
windows without stealing keyboard or mouse focus from the foreground app.
All changes target the cua-driver MCP backend and the shared dispatcher.
## cua_backend.py
**Window-aware capture**: capture() now calls list_windows + get_window_state
instead of the removed capture tool. Prefers structuredContent.windows
(MCP 2024-11-05+ cua-driver) for zero-parse window enumeration; falls back
to regex-parsed text for older builds. Stores the selected (pid, window_id)
as sticky context so subsequent action calls do not need a redundant round-trip.
**Action routing**: click/scroll/type_text/key all carry the sticky pid
(and window_id for element-indexed clicks). type_text routes through
type_text_chars (individual key events) rather than AX attribute write --
WebKit AXTextFields reject attribute writes from backgrounded processes.
**Key parsing**: _parse_key_combo splits cmd+s-style strings into
(key, [modifiers]) and routes to hotkey (modifier present) or
press_key (bare key) -- cua-driver actual tool names.
**set_value method**: new set_value(value, element) calls the cua-driver
set_value MCP tool. For AXPopUpButton / HTML select in a backgrounded Safari,
AXPress opens the native macOS popup which closes immediately when the app is
non-frontmost; set_value AX-presses the matching child option directly
(no menu required, no focus steal).
**focus_app**: reimplemented as a pure window-selector (enumerates
list_windows, sets sticky pid/window_id) without ever raising the window
or stealing focus.
**list_apps**: fixed tool name from listApps to list_apps; handles plain-text
response via regex when structured data is absent.
**Structured-content extraction**: _extract_tool_result now surfaces
structuredContent from MCP results, enabling the list_windows window array
without text parsing.
**Helpers**: _parse_windows_from_text, _parse_elements_from_tree,
_split_tree_text, _parse_key_combo extracted as module-level functions.
## schema.py
Added set_value to the action enum with a description explaining when to
prefer it over click (select/popup elements, sliders, no focus steal).
Added value field for set_value payloads.
## tool.py
Routed set_value action through _dispatch to backend.set_value.
Added set_value to _DESTRUCTIVE_ACTIONS (approval-gated).
Fixed MIME-type detection in _capture_response: cua-driver may return
JPEG; detect from base64 magic bytes (/9j/ -> image/jpeg, else image/png)
rather than hardcoding image/png.
## agent/display.py + run_agent.py
Guard _detect_tool_failure and result-preview logic against non-string
function_result values: multimodal tool results (dicts with _multimodal=True)
are not string-sliceable; treat them as successes and fall back to str()
for length/preview.
Background macOS desktop control via cua-driver MCP — does NOT steal the
user's cursor or keyboard focus, works with any tool-capable model.
Replaces the Anthropic-native `computer_20251124` approach from the
abandoned #4562 with a generic OpenAI function-calling schema plus SOM
(set-of-mark) captures so Claude, GPT, Gemini, and open models can all
drive the desktop via numbered element indices.
- `tools/computer_use/` package — swappable ComputerUseBackend ABC +
CuaDriverBackend (stdio MCP client to trycua/cua's cua-driver binary).
- Universal `computer_use` tool with one schema for all providers.
Actions: capture (som/vision/ax), click, double_click, right_click,
middle_click, drag, scroll, type, key, wait, list_apps, focus_app.
- Multimodal tool-result envelope (`_multimodal=True`, OpenAI-style
`content: [text, image_url]` parts) that flows through
handle_function_call into the tool message. Anthropic adapter converts
into native `tool_result` image blocks; OpenAI-compatible providers
get the parts list directly.
- Image eviction in convert_messages_to_anthropic: only the 3 most
recent screenshots carry real image data; older ones become text
placeholders to cap per-turn token cost.
- Context compressor image pruning: old multimodal tool results have
their image parts stripped instead of being skipped.
- Image-aware token estimation: each image counts as a flat 1500 tokens
instead of its base64 char length (~1MB would have registered as
~250K tokens before).
- COMPUTER_USE_GUIDANCE system-prompt block — injected when the toolset
is active.
- Session DB persistence strips base64 from multimodal tool messages.
- Trajectory saver normalises multimodal messages to text-only.
- `hermes tools` post-setup installs cua-driver via the upstream script
and prints permission-grant instructions.
- CLI approval callback wired so destructive computer_use actions go
through the same prompt_toolkit approval dialog as terminal commands.
- Hard safety guards at the tool level: blocked type patterns
(curl|bash, sudo rm -rf, fork bomb), blocked key combos (empty trash,
force delete, lock screen, log out).
- Skill `apple/macos-computer-use/SKILL.md` — universal (model-agnostic)
workflow guide.
- Docs: `user-guide/features/computer-use.md` plus reference catalog
entries.
44 new tests in tests/tools/test_computer_use.py covering schema
shape (universal, not Anthropic-native), dispatch routing, safety
guards, multimodal envelope, Anthropic adapter conversion, screenshot
eviction, context compressor pruning, image-aware token estimation,
run_agent helpers, and universality guarantees.
469/469 pass across tests/tools/test_computer_use.py + the affected
agent/ test suites.
- `model_tools.py` provider-gating: the tool is available to every
provider. Providers without multi-part tool message support will see
text-only tool results (graceful degradation via `text_summary`).
- Anthropic server-side `clear_tool_uses_20250919` — deferred;
client-side eviction + compressor pruning cover the same cost ceiling
without a beta header.
- macOS only. cua-driver uses private SkyLight SPIs
(SLEventPostToPid, SLPSPostEventRecordTo,
_AXObserverAddNotificationAndCheckRemote) that can break on any macOS
update. Pin with HERMES_CUA_DRIVER_VERSION.
- Requires Accessibility + Screen Recording permissions — the post-setup
prints the Settings path.
Supersedes PR #4562 (pyautogui/Quartz foreground backend, Anthropic-
native schema). Credit @0xbyt4 for the original #3816 groundwork whose
context/eviction/token design is preserved here in generic form.
When delegation.provider is configured (e.g. minimax-cn), subagents
inherited the parent's acp_command unconditionally. This caused
run_agent.py to initialize CopilotACPClient, which bypassed the
override credentials entirely and used its own default model
(provider=copilot-acp model=qwen3.5-397b-a17b) instead of the
configured delegation.provider and delegation.model.
Fix: when override_provider is set but override_acp_command is not,
clear effective_acp_command and effective_acp_args so the child agent
uses direct API calls with the configured provider credentials.
The existing override_acp_command path is unchanged — explicit ACP
transport overrides still force provider=copilot-acp as before.
Fixes#16816
* Port from Kilo-Org/kilocode#9448: roll up subagent costs into parent session total
Child subagents built by delegate_task() each track their own
session_estimated_cost_usd, but the parent agent's total never folded
those numbers in. On runs where the parent mostly delegates and the
children do the expensive work, the footer/UI was reporting a fraction
of the actual spend — sometimes $0.00 when the parent itself made no
billed calls.
Fix:
- Capture each child's session_estimated_cost_usd into _child_cost_usd
on the result entry (before child.close() drops the counter).
- After the existing subagent_stop hook loop, sum the children's costs
and add the total to parent.session_estimated_cost_usd.
- Promote session_cost_source from 'none' -> 'subagent' when the parent
had no direct spend but children did, so the UI doesn't label the
total as having unknown provenance. Real sources (openrouter,
anthropic, etc.) are preserved.
Nested orchestrator -> worker trees roll up naturally: each layer's own
delegate_task() folds its direct children in, and when the orchestrator
itself returns, its parent folds the orchestrator's now-inflated total
on top.
Internal fields (_child_cost_usd, _child_role) are stripped from the
results dict before it's serialised back to the model — same contract
as _child_role already followed.
Tests: TestSubagentCostRollup (5 cases) covers single-child, batch,
zero-cost-children, preserved-source, and legacy-fixture paths.
Source: https://github.com/Kilo-Org/kilocode/pull/9448
* fix(web): scope dashboard config Reset button to the current tab
Reported by @ykmfb001 via X: clicking 'Restore Defaults' (恢复默认值) on
the Auxiliary page wiped the entire config.yaml to defaults, not just
the auxiliary section. The button sits next to the category tabs and
users reasonably assumed 'reset this tab', not 'reset everything'.
Changes:
- handleReset now scopes to the fields in the current view:
active category's fields (form mode) or search-matched fields
(search mode). Only those keys are copied from defaults; the rest
of the config is left alone.
- Added a window.confirm() with the scope name before applying.
- Button is hidden in YAML mode (scoping doesn't apply there).
- Tooltip/aria-label now name the scope, e.g. 'Reset Auxiliary to
defaults'.
- i18n: new resetScopeTooltip / confirmResetScope / resetScopeToast
strings in en + zh; resetDefaults key preserved for compat.
Plugins can now observe dangerous-command approval events in real time,
on both the CLI-interactive path and the async gateway path. This is the
missing hook surface external tools need to build approval notifiers
(macOS menu-bar allow/deny, Slack alerts, audit logs, etc.) without
forking Hermes or running a parallel gateway adapter.
Changes:
- hermes_cli/plugins.py: add two entries to VALID_HOOKS
- tools/approval.py: fire both hooks from check_all_command_guards --
around prompt_dangerous_approval (CLI surface) and around the
notify_cb + blocking event.wait loop (gateway surface)
- website/docs/user-guide/features/hooks.md: document both hooks with
a macOS-notification example
- tests/tools/test_approval_plugin_hooks.py: 5 tests covering CLI once,
CLI deny, plugin-crash resilience, gateway approve, gateway timeout
Hooks are observer-only: return values are ignored, so plugins cannot
veto or pre-answer an approval (use pre_tool_call for that). A crashing
plugin cannot break the approval flow -- invoke_hook swallows per-
callback errors, and the wrapper logs and swallows dispatch-layer
errors too.
Surface kwarg distinguishes "cli" from "gateway"; post hook reports
choice as one of once/session/always/deny/timeout.
PR #13734 fixed the concurrent-tool-executor vector (ThreadPoolExecutor
workers didn't inherit the CLI's TLS approval callback). Two vectors
remained that could still land in the deadlocking input() fallback:
1. _spawn_background_review spawns a raw threading.Thread with no
approval callback installed, so any dangerous-command guard the
review agent trips falls back to input() -> deadlock against the
parent's prompt_toolkit TUI (same class as delegate_task subagents,
fixed in 023b1bff1 / #15491). Install a _bg_review_auto_deny
callback at thread start, clear on finally.
2. prompt_dangerous_approval's fallback unconditionally spawned a
daemon thread calling input() when approval_callback was None.
That fallback can never succeed under prompt_toolkit because the
user's Enter goes to pt's raw-mode stdin capture. Detect an active
pt Application via get_app_or_none() and fail closed (deny + log)
instead, so future threads that forget to install a callback
degrade gracefully instead of hanging 60s invisibly.
Regression guards:
- tests/run_agent/test_background_review.py verifies the review
worker thread sees a callable auto-deny callback mid-run and that
the slot is cleared in the finally block.
- tests/tools/test_approval.py TestFailClosedUnderPromptToolkit
verifies prompt_dangerous_approval returns 'deny' fast under a
mocked pt Application, and that a real callback still wins over
the guard.
* feat(image-input): native multimodal routing based on model vision capability
Attach user-sent images as OpenAI-style content parts on the user turn when
the active model supports native vision, so vision-capable models see real
pixels instead of a lossy text description from vision_analyze.
Routing decision (agent/image_routing.py::decide_image_input_mode):
agent.image_input_mode = auto | native | text (default: auto)
In auto mode:
- If auxiliary.vision.provider/model is explicitly configured, keep the
text pipeline (user paid for a dedicated vision backend).
- Else if models.dev reports supports_vision=True for the active
provider/model, attach natively.
- Else fall back to text (current behaviour).
Call sites updated: gateway/run.py (all messaging platforms), tui_gateway
(dashboard/Ink), cli.py (interactive /attach + drag-drop).
run_agent.py changes:
- _prepare_anthropic_messages_for_api now passes image parts through
unchanged when the model supports vision — the Anthropic adapter
translates them to native image blocks. Previous behaviour
(vision_analyze → text) only runs for non-vision Anthropic models.
- New _prepare_messages_for_non_vision_model mirrors the same contract
for chat.completions and codex_responses paths, so non-vision models
on any provider get text-fallback instead of failing at the provider.
- New _model_supports_vision() helper reads models.dev caps.
vision_analyze description rewritten: positions it as a tool for images
NOT already visible in the conversation (URLs, tool output, deeper
inspection). Prevents the model from redundantly calling it on images
already attached natively.
Config default: agent.image_input_mode = auto.
Tests: 35 new (test_image_routing.py + test_vision_aware_preprocessing.py),
all existing tests that reference _prepare_anthropic_messages_for_api
still pass (198 targeted + new tests green).
* feat(image-input): size-cap + resize oversized images, charge image tokens in compressor
Two follow-ups that make the native image routing safer for long / heavy
sessions:
1) Oversize handling in build_native_content_parts:
- 20 MB ceiling per image (matches vision_tools._MAX_BASE64_BYTES,
the most restrictive provider — Gemini inline data).
- Delegates to vision_tools._resize_image_for_vision (Pillow-based,
already battle-tested) to downscale to 5 MB first-try.
- If Pillow is missing or resize still overshoots, the image is
dropped and reported back in skipped[]; caller falls back to text
enrichment for that image.
2) Image-token accounting in context_compressor:
- New _IMAGE_TOKEN_ESTIMATE = 1600 (matches Claude Code's constant;
within the realistic range for Anthropic/GPT-4o/Gemini billing).
- _content_length_for_budget() helper: sums text-part lengths and
charges _IMAGE_CHAR_EQUIVALENT (1600 * 4 chars) per image/image_url/
input_image part. Base64 payload inside image_url is NOT counted
as chars — dimensions don't matter, only image-presence.
- Both tail-cut sites (_prune_old_tool_results L527 and
_find_tail_cut_by_tokens L1126) now call the helper so multi-image
conversations don't slip past compression budget.
Tests: 9 new in test_image_routing.py (oversize triggers resize,
resize-fails-returns-None, oversize-skipped-reported), 11 new in
test_compressor_image_tokens.py (flat charge per image, multiple images,
Responses-API / Anthropic-native / OpenAI-chat shapes, no-inflation on
raw base64, bounds-check on the constant, integration test that an
image-heavy tail actually gets trimmed).
* fix(image-input): replace blanket 20MB ceiling with empirically-verified per-provider limits
The previous commit imposed a hardcoded 20 MB base64 ceiling on all
providers, triggering auto-resize on anything larger. This was wrong in
both directions:
* Too loose for Anthropic — actual limit is 5 MB (returns HTTP 400
'image exceeds 5 MB maximum' above that).
* Too strict for OpenAI / Codex / OpenRouter — accept 49 MB+ without
complaint (empirically verified April 2026 with progressive PNG
sizes).
New behaviour:
* _PROVIDER_BASE64_CEILING table: only anthropic and bedrock have a
ceiling (5 MB, since bedrock-on-Claude shares Anthropic's decoder).
* Providers NOT in the table get no ceiling — images attach at native
size and we trust the provider to return its own error if it
disagrees. A provider-specific 400 message is clearer than us
guessing wrong and silently degrading image quality.
* build_native_content_parts() gains a keyword-only provider arg;
gateway/CLI/TUI pass the active provider so Anthropic users get
auto-resize protection while OpenAI users don't pay it.
* Resize target dropped from 5 MB to 4 MB to slide safely under
Anthropic's boundary with header overhead.
Empirical measurements (direct API, no Hermes in the loop):
image b64 anthropic openrouter/gpt5.5 codex-oauth/gpt5.5
0.19 MB ✓ ✓ ✓
12.37 MB ✗ 400 5MB ✓ ✓
23.85 MB ✗ 400 5MB ✓ ✓
49.46 MB ✗ 413 ✓ ✓
Tests: rewrote TestOversizeHandling (5 tests): no-ceiling pass-through,
Anthropic resize fires, Anthropic skip on resize-fail, build_native_parts
routes ceiling by provider, unknown provider gets no ceiling. All 52
targeted tests pass.
* refactor(image-input): attempt native, shrink-and-retry on provider reject
Replace proactive per-provider size ceilings with a reactive shrink path
on the provider's actual rejection. All providers now attempt native
full-size attachment first; if the provider returns an image-too-large
error, the agent silently shrinks and retries once.
Why the previous design was wrong: hardcoding provider ceilings
(anthropic=5MB, others=unlimited) meant OpenAI users on a 10MB image
paid no tax, but Anthropic users lost quality on anything >5MB even
though the empirical behaviour at provider-reject time is the same
(shrink + retry). Baking the table into the routing layer also
requires updating Hermes every time a provider's limit changes.
Reactive design:
- image_routing.py: _file_to_data_url encodes native size, no ceiling.
build_native_content_parts drops its provider kwarg.
- error_classifier.py: new FailoverReason.image_too_large + pattern
match ("image exceeds", "image too large", etc.) checked BEFORE
context_overflow so Anthropic's 5MB rejection lands in the right
bucket.
- run_agent.py: new _try_shrink_image_parts_in_messages walks api
messages in-place, re-encodes oversized data: URL image parts
through vision_tools._resize_image_for_vision to fit under 4MB,
handles both chat.completions (dict image_url) and Responses
(string image_url) shapes, ignores http URLs (provider-fetched).
New image_shrink_retry_attempted flag in the retry loop fires the
shrink exactly once per turn after credential-pool recovery but
before auth retries.
E2E verified live against Anthropic claude-sonnet-4-6:
- 17.9MB PNG (23.9MB b64) attached at native size
- Anthropic returns 400 "image exceeds 5 MB maximum"
- Agent logs '📐 Image(s) exceeded provider size limit — shrank and
retrying...'
- Retry succeeds, correct response delivered in 6.8s total.
Tests: 12 new (8 shrink-helper shapes + 4 classifier signals),
replaces 5 proactive-ceiling tests with 3 simpler 'native attach works'
tests. 181 targeted tests pass. test_enum_members_exist in
test_error_classifier.py updated for the new enum value.
On macOS (bash 3.2 and some Homebrew bash builds) `source`ing a file that
contains `declare -x` statements prints each declaration to stdout. The
persistent-shell wrapper in tools/environments/base.py was only redirecting
stderr when sourcing the session snapshot, so ~60 lines of env vars leaked
into every terminal tool response — blowing out context and triggering
HTTP 400s on context-limited providers.
Fix: redirect both stdout and stderr when sourcing the snapshot. Linux
bash is silent here, so the redirect is harmless there; macOS no longer
leaks.
Closes#15459
Co-authored-by: Sanjays2402 <51058514+Sanjays2402@users.noreply.github.com>
read_file's dedup path returned a lightweight stub on re-reads of an
unchanged file, then returned early — so the consecutive-read loop
guard (hard block at count>=4) at the bottom of read_file_tool never
ran for stub-looped calls. Weaker tool-following models (local Qwen3.6
variants in the reported case) ignore the passive 'refer to earlier
result' hint and hammer the same read_file call until iteration budget
runs out.
Track per-key stub returns in task_data['dedup_hits'] and, on the
second stub for the same (path, offset, limit), return a hard BLOCKED
error mirroring the wording the real-read path already uses. A real
read, an intervening non-read tool call (notify_other_tool_call), or
reset_file_dedup (on context compression) all clear the counter so
the guard never stays engaged longer than the actual loop.
Closes#15759
Four small tool-description / skill-content tweaks addressing recurring
model mistakes seen in @versun's docx feedback (Kimi 2.6, but the patterns
apply to every model):
1. browser_navigate description: call out .md/.txt/.json/.yaml/.csv/.xml,
raw.githubusercontent.com, and API endpoints as specifically preferring
curl or web_extract. The generic "prefer web_search or web_extract" was
too weak; models kept firing up the browser for plain-text URLs.
2. delegate_task description: two additions.
(a) Pass user language / output-style preferences in 'context' when they
differ from English — otherwise subagents default to English and their
summaries contaminate the final reply (caused the bilingual digest bug).
(b) Subagent summaries are self-reports, not verified facts. For
operations with external side-effects (HTTP uploads, remote writes,
file creation at shared paths), require a verifiable handle (URL, ID,
path) and verify it yourself before claiming success.
3. agent/prompt_builder.py Skills-mandatory block: new explicit line
"Whenever the user asks to configure / set up / modify / install /
enable / disable / troubleshoot Hermes Agent itself, load the
`hermes-agent` skill first." The generic "load what's relevant" didn't
route Hermes-meta questions (like "how do I turn off redaction?") to
the one skill that has the answer.
4. skills/autonomous-ai-agents/hermes-agent/SKILL.md: new "Security &
Privacy Toggles" section covering security.redact_secrets (with the
import-time-snapshot restart-required caveat), privacy.redact_pii,
approvals.mode (manual/smart/off) + --yolo + HERMES_YOLO_MODE, shell
hooks allowlist, and how to disable network/media tools entirely.
Every command verified against the actual config keys — no invented
knobs.
Co-authored-by: teknium1 <teknium@noreply.github.com>
* feat(skills): install skills from a direct HTTP(S) URL
Adds UrlSource adapter so `hermes skills install <url-to-SKILL.md>` and
`/skills install <url>` work as first-class operations — no more
improvising with curl + patch + cp.
- Claims identifiers that start with http(s):// and end in .md
- Skips /.well-known/skills/ URLs (WellKnownSkillSource handles those)
- Skill name from YAML frontmatter, URL-slug fallback
- Single-file SKILL.md only (v1 scope — multi-file skills need a manifest)
- Trust level 'community'; full security scan still runs
- Lock file stores the URL as identifier so `hermes skills update`
re-fetches from the same URL cleanly
Scope matches real user need from @versun's docx feedback where
`https://sharethis.chat/SKILL.md` had no first-class install path.
* feat(skills): interactive name/category for URL installs + --name override
Follow-up to the UrlSource adapter. The previous commit fell back to weak
heuristics when frontmatter had no ``name:`` and could produce garbage names
like ``SKILL`` or ``unnamed-skill``. Now:
tools/skills_hub.py
- ``UrlSource._is_valid_skill_name()`` — strict identifier check
(``^[a-z][a-z0-9_-]*$``), rejects sentinel values (``SKILL``, ``README``,
``INDEX``, ``unnamed-skill``, empty, non-strings).
- ``_resolve_skill_name()`` returns ``Optional[str]`` — ``None`` when
nothing valid is resolvable. Also ignores unsafe frontmatter names
(``../evil``) and falls through to URL slug instead of returning None
immediately, so a URL with a bad frontmatter but a good path still
works.
- ``fetch()``/``inspect()`` carry an ``awaiting_name=True`` marker in
metadata/extra when resolution fails, letting ``do_install`` decide
whether to prompt, apply an override, or error out.
hermes_cli/skills_hub.py
- ``do_install`` gains a ``name_override`` parameter.
- On URL-sourced bundles with ``awaiting_name=True``:
1. If ``name_override`` is valid → use it.
2. If ``name_override`` is invalid → refuse with a clear error.
3. Else if ``skip_confirm=True`` (non-interactive: slash / TUI /
gateway / scripts) → refuse with an actionable retry hint pointing
at ``--name <your-name>`` on both CLI and slash forms.
4. Else (interactive TTY) → prompt for the name.
- Interactive TTY also prompts for a category when none is given for a
URL-sourced install, hinting existing category buckets so users can
reuse ``productivity``, ``devops``, etc. Empty input → flat install.
- ``_existing_categories()`` scans ``~/.hermes/skills/`` for subdirs that
look like category buckets (contain nested SKILL.md files); skips
top-level skills and hidden dirs.
- ``_prompt_for_skill_name()`` / ``_prompt_for_category()`` helpers
(EOF/Ctrl-C-safe, match the existing ``Confirm [y/N]`` prompt style).
hermes_cli/main.py
- ``hermes skills install`` argparse gains ``--name <name>``.
hermes_cli/skills_hub.py (slash)
- ``/skills install <url> --name <x>`` parsing added.
Tests
- tests/tools/test_skills_hub.py: updated ``UrlSource`` tests to assert
the new ``awaiting_name`` metadata; added 4 new tests for
``_is_valid_skill_name`` rejection sets and the awaiting-name marker.
- tests/hermes_cli/test_skills_hub.py: 8 new tests covering --name
override accept/reject, non-interactive error, interactive name prompt,
interactive category prompt, cancel-aborts-install, and
``_existing_categories`` scan behavior (buckets vs flat skills).
- E2E verified all four paths (no-name/no-override → error;
--name override → install; frontmatter name → install;
invalid --name → rejection).
---------
Co-authored-by: teknium1 <teknium@noreply.github.com>
_search_members() and _fetch_messages() call min(limit, 100) assuming
limit is int. Models can pass limit as a string (e.g. "10"), causing
TypeError: '<' not supported between instances of 'str' and 'int'.
Add try/except int() coercion with safe defaults at the top of both
functions, matching the pattern used in session_search fix (#10522).
Every working dir hermes ever touches gets its own shadow git repo under
~/.hermes/checkpoints/{sha256(abs_dir)[:16]}/. The per-repo _prune is a
no-op (comment in CheckpointManager._prune says so), so abandoned repos
from deleted/moved projects or one-off tmp dirs pile up forever. Field
reports put the typical offender at 1000+ repos / ~12 GB on active
contributor machines.
Adds an opt-in startup sweep that mirrors the sessions.auto_prune
pattern from #13861 / #16286:
- tools/checkpoint_manager.py: new prune_checkpoints() and
maybe_auto_prune_checkpoints() helpers. Deletes shadow repos that
are orphan (HERMES_WORKDIR marker points to a path that no longer
exists) or stale (newest in-repo mtime older than retention_days).
Idempotent via a CHECKPOINT_BASE/.last_prune marker file so it only
runs once per min_interval_hours regardless of how many hermes
processes start up.
- hermes_cli/config.py: new checkpoints.auto_prune /
retention_days / delete_orphans / min_interval_hours knobs.
Default auto_prune: false so users who rely on /rollback against
long-ago sessions never lose data silently.
- cli.py / gateway/run.py: startup hooks gated on checkpoints.auto_prune,
called right next to the existing state.db maintenance block.
- Docs updated with the new config knobs.
- 11 regression tests: orphan/stale deletion, precedence, byte-freed
tracking, non-shadow dir skip, interval gating, corrupt marker
recovery.
Refs #3015 (session-file disk growth was fixed in #16286; this covers
the checkpoint side noted out-of-scope there).
The write_file guard added in #16223 used strict equality against the
internal dedup status message. In practice, the model sometimes
prepends a short note or appends a trailing comment before calling
write_file, which slipped past the strict check.
Broaden the heuristic: reject writes whose stripped content equals
the status message OR contains it and is <=2x its length. Short,
status-dominated writes are always corruption; legitimate docs that
quote the message verbatim are always much longer.
Adds two tests: one for the small-wrapper corruption shape, one
confirming large legitimate files that quote the status still write.
write_file_tool and patch_tool both call _update_read_timestamp to
refresh the staleness tracker after writing, but they never invalidate
the dedup cache entries for the written path. The dedup cache keys are
(resolved_path, offset, limit) → mtime tuples populated by read_file_tool.
On filesystems where a read and write land in the same mtime second (or
when mtime granularity is 1s), the cached and current mtime are equal,
so the dedup check incorrectly returns a 'File unchanged since last
read' stub — even though the file was just overwritten.
The agent then sees stale content (or a stale 'File not found' error)
and enters expensive error-recovery loops, burning API calls.
Fix: add _invalidate_dedup_for_path(filepath, task_id) that removes all
dedup entries whose resolved path matches the written file. Called from
_update_read_timestamp so both write_file_tool and patch_tool benefit
automatically. Scoped to the writing task_id — other tasks' caches are
not affected.
6 regression tests added covering:
- read→write→read within same mtime second (core #13144 scenario)
- invalidation across all offset/limit combinations
- isolation: writing file A does not invalidate file B's cache
- isolation: writing in task A does not invalidate task B's cache
- _invalidate_dedup_for_path safety on missing task / empty dedup
All 25 tests pass (19 existing + 6 new).
Fixes#13144
MCP stdio servers are spawned via the SDK's stdio_client, which on
Linux uses start_new_session=True (setsid). When a cron job is
cancelled mid-way (timeout, agent finish, exception), the subprocess
often escapes the SDK's teardown and survives as a session leader.
Because setsid() detaches the child from the gateway's process group
/ cgroup tree, systemd does not reap it on service restart either —
so every cron tick that touches an MCP tool leaks a dangling server
process.
Fix:
* tools/mcp_tool.py — _run_stdio now wraps the whole stdio+session
context in try/finally. On any exit path (clean, exception,
cancellation), PIDs still alive are moved from the active
_stdio_pids set into a new _orphan_stdio_pids set. Orphan
detection is done via os.kill(pid, 0) — a cheap liveness probe
that never signals the target.
* tools/mcp_tool.py — _kill_orphaned_mcp_children gains an
include_active=False flag. Default behaviour now only reaps the
orphan set so concurrent sessions (other parallel cron jobs or
live user chats) are never disrupted. The existing shutdown path
passes include_active=True to keep the previous "kill everything"
semantics after the MCP loop is stopped.
* cron/scheduler.py — the cleanup hook is moved from run_job()'s
finally (which would race with parallel siblings after #13021)
into tick() after the ThreadPoolExecutor has joined every future.
At that point there are no in-flight sessions from this tick, so
sweeping the orphan set is always safe.
Net effect: zero regression for healthy sessions, and orphan MCP
servers no longer accumulate between gateway restarts.
Made-with: Cursor
Slack's chat.postMessage API rejects user IDs (U...) and workspace
IDs (W...) — they are not valid conversation IDs. Posting to them
fails because the API requires a channel ID (C/G/D). To DM a user,
the sender must first call conversations.open to obtain a D... ID.
Tighten _SLACK_TARGET_RE from [CGDUW] to [CGD] so the send path rejects
U/W values as explicit targets and instead falls through to channel-
name resolution (where they'll fail with a clear 'could not resolve'
error rather than silently getting stuck in a retry loop on the API).
Flip the corresponding regression test to assert U/W values are not
explicit. Matches the narrower regex briandevans proposed in #15939.
Co-authored-by: briandevans <brian@bde.io>
send_message(target='slack:<channel_id>') failed with "Could not
resolve" because _parse_target_ref had no Slack branch — Slack's
uppercase alphanumeric IDs fell through to channel-name resolution,
which only matched by name. As a fallback, the agent would retry with
bare target='slack' and post to the home channel instead.
Three fixes:
- _parse_target_ref recognizes Slack IDs (C/G/D/U/W prefix) as
explicit targets so the name-resolver is bypassed entirely.
- resolve_channel_name tries a case-sensitive raw-ID match before
the existing name match, so any platform's IDs resolve cleanly.
- _build_slack now actually calls users.conversations against each
workspace's AsyncWebClient (paginated), instead of only returning
session-history entries. This populates the directory with public
and private channels the bot has joined, so action='list' shows
them and they can also be addressed by name. Errors from one
workspace don't block others.
build_channel_directory becomes async (Slack web calls require it).
The two async-context callers in gateway/run.py are awaited; the
cron ticker thread call bridges via asyncio.run_coroutine_threadsafe.
Slack bot needs channels:read and groups:read scopes for full
enumeration; missing scopes degrade gracefully per-workspace.
addressing #15927
Before: delegate_task children each allocated their own terminal
sandbox keyed by child task_id. Starting extra containers (or Modal
sandboxes / Daytona workspaces) is expensive, and the subagent's work
is invisible to the parent — files written by the child in its
container don't exist in the parent's when the subagent returns.
After: a single `_resolve_container_task_id` helper maps any
tool-call task_id to "default" UNLESS an env override is registered
for it. The parent agent and all delegate_task children therefore
share one long-lived sandbox — installed packages, cwd, /workspace
files, and /tmp scratch carry over freely between them.
RL and benchmark environments (TerminalBench2, HermesSweEnv, ...)
opt in to isolation via `register_task_env_overrides(task_id, {...})`;
those task_ids survive the collapse and get their own sandbox,
preserving the per-task Docker image behavior these benchmarks rely on.
file_state / active-subagents registry / TUI events still key off the
original child task_id, so the 'subagent wrote a file the parent read'
warning and UI per-subagent panels keep working.
Tradeoff: parallel delegate_task children (tasks=[...]) now share one
bash/container. Concurrent cd, env-var mutations, and writes to the
same path will collide. If that bites a specific workflow, the
subagent can opt back into isolation via register_task_env_overrides.
Applied at four lookup sites:
- tools/terminal_tool.py terminal_tool() and get_active_env()
- tools/file_tools.py _get_file_ops() and _get_live_tracking_cwd()
- tools/code_execution_tool.py _get_or_create_environment()
Docs: website/docs/user-guide/configuration.md updated to reflect the
shared-container reality and document the RL/benchmark carve-out.
Tests: tests/tools/test_shared_container_task_id.py (9 cases).
When a cloud browser provider (Browserbase / Browser-Use / Firecrawl) is
configured, browser_navigate now transparently spawns a local Chromium
sidecar for URLs whose host resolves to a private/loopback/LAN address
(localhost, 127.0.0.1, 192.168.x.x, 10.x.x.x, *.local, *.lan, *.internal,
::1, 169.254.x.x). Public URLs continue to use the cloud provider in the
same conversation.
Previously, setting BROWSERBASE_API_KEY / cloud_provider: browserbase
pinned the whole tool to cloud for the process — localhost URLs were
either SSRF-blocked (default) or sent to Browserbase (where they 404'd
because the cloud can't reach your LAN). Users who wanted 'cloud for
public, local for localhost' had no way to express it short of toggling
providers mid-session.
Implementation uses a composite session key scheme: the bare task_id
serves the cloud session, and a '{task_id}::local' sidecar serves the
local Chromium. _last_active_session_key[task_id] tracks which of the
two served the most recent nav so snapshot/click/fill/etc. hit the
correct one. cleanup_browser(bare_task_id) reaps both.
Feature is on by default. Opt out via:
browser:
auto_local_for_private_urls: false
The cloud provider never sees private URLs. Post-redirect SSRF guard
is preserved: redirects from public onto private addresses still block.
Follow-up to #6616 covering the remaining user-injected prompt markers that
the original PR did not touch (reporter's second comment on #6576 explicitly
flagged these). Azure OpenAI Default/DefaultV2 content filters treat any
bracketed [SYSTEM: ...] as prompt-injection and reject with HTTP 400.
Remaining call sites renamed:
- cli.py: background-process notifications (watch_disabled, watch_match,
completion), MCP reload notice (4 live + 1 docstring)
- gateway/run.py: same notification paths + auto-loaded skill banner +
MCP reload notice (5 live + 1 docstring)
- tools/process_registry.py: comment reference
Not renamed:
- environments/hermes_base_env.py '[SYSTEM]\n{content}' — RL training
trajectory rendering only, never sent to Azure, part of a symmetric
[USER]/[ASSISTANT]/[TOOL] scheme.
AUTHOR_MAP: buraysandro9@gmail.com -> ygd58.
Stop pre-stripping the path from the configured MCP server URL before
constructing OAuthClientProvider. The MCP SDK strips the path itself via
OAuthContext.get_authorization_base_url() for authorization-server
discovery, but uses the full server_url through
resource_url_from_server_url() + check_resource_allowed() to validate
against the server's RFC 9728 Protected Resource Metadata.
For servers whose PRM advertises a path-scoped resource (e.g. Notion's
https://mcp.notion.com/mcp), our _parse_base_url() collapsed the URL to
the origin, so check_resource_allowed() saw requested='/' vs
configured='/mcp/' and refused the token. Fixes OAuth against Notion MCP
(and any other path-scoped resource).
Closes#16015.
Adds a floor below --yolo: a tiny set of commands so catastrophic they
should never run via the agent, regardless of --yolo, gateway /yolo,
approvals.mode=off, or cron approve mode. Opting into yolo is trusting
the agent with your files and services — not trusting it to wipe the
disk or power the box off.
The list is deliberately small (12 patterns), covering only
unrecoverable ops:
- rm -rf targeting /, /home, /etc, /usr, /var, /boot, /bin, /sbin,
/lib, ~, $HOME
- mkfs (any variant)
- dd + redirection to raw block devices (/dev/sd*, /dev/nvme*, etc.)
- fork bomb
- kill -1 / kill -9 -1
- shutdown, reboot, halt, poweroff, init 0/6, telinit 0/6,
systemctl poweroff/reboot/halt/kexec
Recoverable-but-costly commands (git reset --hard, rm -rf /tmp/x,
chmod -R 777, curl | sh) stay in DANGEROUS_PATTERNS where yolo can
still pass them through — that's what yolo is for.
Container backends (docker/singularity/modal/daytona) continue to
bypass both hardline and dangerous checks, since nothing they do can
touch the host.
Inspired by Mercury Agent's permission-hardened blocklist.
* fix(terminal): three-layer defense against watch_patterns notification spam
Background processes that stack notify_on_complete=True with watch_patterns
can flood the user with duplicate, delayed notifications — matches deliver
asynchronously via the completion queue and continue arriving minutes after
the process has exited. The docstring warning against this (PR #12113) has
proven insufficient; agents still misuse the combination.
Three layered defenses, each sufficient on its own:
1. Mutual exclusion (terminal_tool.py): When both flags are set on a
background process, drop watch_patterns with a warning. notify_on_complete
wins because 'let me know when it's done' is the more useful signal and
fires exactly once. Extracted as _resolve_notification_flag_conflict() so
the rule is testable in isolation.
2. Suppress-after-exit (process_registry.py): _check_watch_patterns() now
bails the moment session.exited is True. Post-exit chunks (buffered reads
draining after the process is gone) no longer produce notifications. This
is the fix flagged as future work in session 20260418_020302_79881c.
3. Global circuit breaker (process_registry.py): Per-session rate limits don't
catch the sibling-flood case — N concurrent processes can each stay under
8/10s and still collectively spam. New WATCH_GLOBAL_MAX_PER_WINDOW=15 cap
trips a 30-second cooldown across ALL sessions, emits a single
watch_overflow_tripped event, silently counts dropped events, and emits a
watch_overflow_released summary when the cooldown ends.
Also updates the tool schema + docstring to document the new behavior.
Tests: 8 new tests covering all three fixes (suppress-after-exit x2,
mutual-exclusion resolver x4, global breaker trip/cooldown/release x2).
All 60 tests across test_watch_patterns.py, test_notify_on_complete.py,
test_terminal_tool.py pass.
Real-world trigger: self-inflicted in session 20260425_051924 — three
concurrent hermes-sweeper review subprocesses each set watch_patterns=
['failed validation', 'errored'] AND notify_on_complete=True, then iterated
over multiple items, producing enough matches per process to defeat the
per-session cap while staying under the global cap that didn't yet exist.
* fix(terminal): aggressive 1-per-15s watch_patterns rate limit + strike-3 promotion
Per Teknium's direction, the watch_patterns rate limit is now much more
aggressive and self-healing.
## New rule — per session
- HARD cap: 1 watch-match notification per 15 seconds per process.
- Any match arriving inside the cooldown window is dropped and counts as
ONE strike for that window (many drops in the same window still = 1 strike).
- After 3 consecutive strike windows, watch_patterns is permanently disabled
for the session and the session is auto-promoted to notify_on_complete
semantics — exactly one notification when the process actually exits.
- A cooldown window that expires with zero drops resets the consecutive
strike counter — healthy cadence is forgiven.
## Schema + docstring rewritten
The tool schema description now gives the model explicit guidance:
- notify_on_complete is 'the right choice for almost every long-running task'
- watch_patterns is for RARE one-shot signals on LONG-LIVED processes
- Do NOT use watch_patterns with loops/batch jobs — error patterns fire every
iteration and will hit the strike limit fast
- Mutual exclusion is stated on both parameter descriptions
- 1/15s cooldown and 3-strike promotion are stated in the watch_patterns
description so the model sees the contract every turn
## Removed
- WATCH_MAX_PER_WINDOW (8/10s) and WATCH_OVERLOAD_KILL_SECONDS (45) — the
new 1/15s limit subsumes both; keeping them would double-count.
- _watch_window_hits / _watch_window_start / _watch_overload_since fields
on ProcessSession. Replaced by _watch_last_emit_at / _watch_cooldown_until
/ _watch_strike_candidate / _watch_consecutive_strikes.
## Kept
- Global circuit breaker across all sessions (15/10s → 30s cooldown) as a
secondary safety net for concurrent siblings. Still valuable when 20
short-lived processes each fire once — none individually violates the
per-session limit.
- Suppress-after-exit guard.
- Mutual exclusion resolver at the tool entry point.
## Tests
- 6 new tests in TestPerSessionRateLimit covering: first match delivers,
second in cooldown suppressed, multi-drop = single strike, 3 strikes
disables + promotes, clean window resets counter, suppressed count
carried to next emit.
- Global circuit breaker tests rewritten to use fresh sessions instead of
hacking removed per-window fields.
- 50/50 watch_patterns + notify_on_complete tests pass.
- 60/60 including test_terminal_tool.py pass.
Split the monolithic discord_server tool (14 actions) into two:
- discord: core actions (fetch_messages, search_members, create_thread)
that are useful for the agent's normal operation. Auto-enabled on
the discord platform via the pipeline fix.
- discord_admin: server management actions (list channels/roles, pins,
role assignment) that require explicit opt-in via hermes tools.
Added to CONFIGURABLE_TOOLSETS and _DEFAULT_OFF_TOOLSETS.
The tool schema promised 'On update, pass an empty array to clear' but the
update branch ignored the context_from kwarg entirely — users could set
the field at create time and never modify or clear it afterward.
- tools/cronjob_tools.py: handle context_from in the update branch the
same way script/enabled_toolsets/workdir are handled: normalize str/list
to refs, validate each referenced job exists (same check the create
branch does), store as list-or-None to match create_job()'s shape.
Empty string or empty list clears the field.
- tests/cron/test_cron_context_from.py: 6 new tests covering add/change/
clear (both shapes)/bad-ref/preserve-across-unrelated-update.
Subagents run inside a ThreadPoolExecutor. The CLI's interactive approval
callback lives in tools/terminal_tool.py's threading.local(), which worker
threads do not inherit. When a subagent hits a dangerous-command guard,
prompt_dangerous_approval() falls back to input() from the worker thread,
deadlocking against the parent's prompt_toolkit TUI that owns stdin.
Fix: install a non-interactive callback into every subagent worker thread
via ThreadPoolExecutor(initializer=set_approval_callback, initargs=(cb,)).
The callback is config-gated by delegation.subagent_auto_approve:
false (default) -> _subagent_auto_deny (safe; matches leaf tool blocklist)
true -> _subagent_auto_approve (opt-in YOLO for cron/batch)
Both emit a logger.warning audit line. Gateway sessions are unaffected
because they resolve approvals via tools/approval.py's per-session queue,
not through these TLS callbacks. Diagnosis credit: @MorAlekss (#14685).
- hermes_cli/config.py: DEFAULT_CONFIG.delegation.subagent_auto_approve: False
- cli-config.yaml.example: documented, commented (default)
- tools/delegate_tool.py: _subagent_auto_deny, _subagent_auto_approve,
_get_subagent_approval_callback, wired into the child timeout executor
- tests/tools/test_delegate.py: 7 tests covering defaults, truthy coercion,
and TLS scoping in the worker thread
skill_view response went to the model verbatim; duplicating the SKILL.md
body as raw_content on every tool call added token cost with no agent-facing
benefit. Remove the field and update tests to assert on content only.
The slash/preload caller (agent/skill_commands.py) already falls back to
content when raw_content is absent, and it calls skill_view(preprocess=False)
anyway, so content is already unrendered on that path.
A child running a legitimately long-running tool (terminal command,
browser fetch, big file read) holds current_tool set and keeps
api_call_count frozen while the tool runs. The previous stale check
treated that as idle after 5 heartbeat cycles (~150s), stopped
touching the parent, and let the gateway kill the session.
Split the threshold in two:
- _HEARTBEAT_STALE_CYCLES_IDLE=5 (~150s) — applied only when
current_tool is None (child wedged between turns)
- _HEARTBEAT_STALE_CYCLES_IN_TOOL=20 (~600s) — applied when the child
is inside a tool call
Stale counter also resets when current_tool changes (new tool =
progress). The hard child_timeout_seconds (default 600s) is still
the final cap, so genuinely stuck tools don't get to block forever.
Moves the Spotify integration from tools/ into plugins/spotify/,
matching the existing pattern established by plugins/image_gen/ for
third-party service integrations.
Why:
- tools/ should be reserved for foundational capabilities (terminal,
read_file, web_search, etc.). tools/providers/ was a one-off
directory created solely for spotify_client.py.
- plugins/ is already the home for image_gen backends, memory
providers, context engines, and standalone hook-based plugins.
Spotify is a third-party service integration and belongs alongside
those, not in tools/.
- Future service integrations (eventually: Deezer, Apple Music, etc.)
now have a pattern to copy.
Changes:
- tools/spotify_tool.py → plugins/spotify/tools.py (handlers + schemas)
- tools/providers/spotify_client.py → plugins/spotify/client.py
- tools/providers/ removed (was only used for Spotify)
- New plugins/spotify/__init__.py with register(ctx) calling
ctx.register_tool() × 7. The handler/check_fn wiring is unchanged.
- New plugins/spotify/plugin.yaml (kind: backend, bundled, auto-load).
- tests/tools/test_spotify_client.py: import paths updated.
tools_config fix — _DEFAULT_OFF_TOOLSETS now wins over plugin auto-enable:
- _get_platform_tools() previously auto-enabled unknown plugin
toolsets for new platforms. That was fine for image_gen (which has
no toolset of its own) but bad for Spotify, which explicitly
requires opt-in (don't ship 7 tool schemas to users who don't use
it). Added a check: if a plugin toolset is in _DEFAULT_OFF_TOOLSETS,
it stays off until the user picks it in 'hermes tools'.
Pre-existing test bug fix:
- tests/hermes_cli/test_plugins.py::test_list_returns_sorted
asserted names were sorted, but list_plugins() sorts by key
(path-derived, e.g. image_gen/openai). With only image_gen plugins
bundled, name and key order happened to agree. Adding plugins/spotify
broke that coincidence (spotify sorts between openai-codex and xai
by name but after xai by key). Updated test to assert key order,
which is what the code actually documents.
Validation:
- scripts/run_tests.sh tests/hermes_cli/test_plugins.py \
tests/hermes_cli/test_tools_config.py \
tests/hermes_cli/test_spotify_auth.py \
tests/tools/test_spotify_client.py \
tests/tools/test_registry.py
→ 143 passed
- E2E plugin load: 'spotify' appears in loaded plugins, all 7 tools
register into the spotify toolset, check_fn gating intact.
Three quality improvements on top of #15121 / #15130 / #15135:
1. Tool consolidation (9 → 7)
- spotify_saved_tracks + spotify_saved_albums → spotify_library with
kind='tracks'|'albums'. Handler code was ~90 percent identical
across the two old tools; the merge is a behavioral no-op.
- spotify_activity dropped. Its 'now_playing' action was a duplicate
of spotify_playback.get_currently_playing (both return identical
204/empty payloads). Its 'recently_played' action moves onto
spotify_playback as a new action — history belongs adjacent to
live state.
- Net: each API call ships 2 fewer tool schemas when the Spotify
toolset is enabled, and the action surface is more discoverable
(everything playback-related is on one tool).
2. Spotify skill (skills/media/spotify/SKILL.md)
Teaches the agent canonical usage patterns so common requests don't
balloon into 4+ tool calls:
- 'play X' = one search, then play by URI (not search + scan +
describe + play)
- 'what's playing' = single get_currently_playing (no preflight
get_state chain)
- Don't retry on '403 Premium required' or '403 No active device' —
both require user action
- URI/URL/bare-ID format normalization
- Full failure-mode reference for 204/401/403/429
3. Surfaced in 'hermes setup' tool status
Adds 'Spotify (PKCE OAuth)' to the tool status list when
auth.json has a Spotify access/refresh token. Matches the
homeassistant pattern but reads from auth.json (OAuth-based) rather
than env vars.
Docs updated to reflect the new 7-tool surface, and mention the
companion skill in the 'Using it' section.
Tests: 54 passing (client 22, auth 15, tools_config 35 — 18 = 54 after
renaming/replacing the spotify_activity tests with library +
recently_played coverage). Docusaurus build clean.
Streamable HTTP MCP servers may garbage-collect their server-side
session state while the OAuth token remains valid — idle TTL, server
restart, pod rotation, etc. Before this fix, the tool-call handler
treated the resulting "Invalid or expired session" error as a plain
tool failure with no recovery path, so **every subsequent call on
the affected server failed until the gateway was manually
restarted**. Reporter: #13383.
The OAuth-based recovery path (``_handle_auth_error_and_retry``)
already exists for 401s, but it only fires on auth errors. Session
expiry slipped through because the access token is still valid —
nothing 401'd, so the existing recovery branch was skipped.
Fix
---
Add a sibling function ``_handle_session_expired_and_retry`` that
detects MCP session-expiry via ``_is_session_expired_error`` (a
narrow allow-list of known-stable substrings: ``"invalid or expired
session"``, ``"session expired"``, ``"session not found"``,
``"unknown session"``, etc.) and then uses the existing transport
reconnect mechanism:
* Sets ``MCPServerTask._reconnect_event`` — the server task's
lifecycle loop already interprets this as "tear down the current
``streamablehttp_client`` + ``ClientSession`` and rebuild them,
reusing the existing OAuth provider instance".
* Waits up to 15 s for the new session to come back ready.
* Retries the original call once. If the retry succeeds, returns
its result and resets the circuit-breaker error count. If the
retry raises, or if the reconnect doesn't ready in time, falls
through to the caller's generic error path.
Unlike the 401 path, this does **not** call ``handle_401`` — the
access token is already valid and running an OAuth refresh would be
a pointless round-trip.
All 5 MCP handlers (``call_tool``, ``list_resources``, ``read_resource``,
``list_prompts``, ``get_prompt``) now consult both recovery paths
before falling through:
recovered = _handle_auth_error_and_retry(...) # 401 path
if recovered is not None: return recovered
recovered = _handle_session_expired_and_retry(...) # new
if recovered is not None: return recovered
# generic error response
Narrow scope — explicitly not changed
-------------------------------------
* **Detection is string-based on a 5-entry allow-list.** The MCP
SDK wraps JSON-RPC errors in ``McpError`` whose exception type +
attributes vary across SDK versions, so matching on message
substrings is the durable path. Kept narrow to avoid false
positives — a regular ``RuntimeError("Tool failed")`` will NOT
trigger spurious reconnects (pinned by
``test_is_session_expired_rejects_unrelated_errors``).
* **No change to the existing 401 recovery flow.** The new path is
consulted only after the auth path declines (returns ``None``).
* **Retry count stays at 1.** If the reconnect-then-retry also
fails, we don't loop — the error surfaces normally so the model
sees a failed tool call rather than a hang.
* **``InterruptedError`` is explicitly excluded** from session-expired
detection so user-cancel signals always short-circuit the same
way they did before (pinned by
``test_is_session_expired_rejects_interrupted_error``).
Regression coverage
-------------------
``tests/tools/test_mcp_tool_session_expired.py`` (new, 16 cases):
Unit tests for ``_is_session_expired_error``:
* ``test_is_session_expired_detects_invalid_or_expired_session`` —
reporter's exact wpcom-mcp text.
* ``test_is_session_expired_detects_expired_session_variant`` —
"Session expired" / "expired session" variants.
* ``test_is_session_expired_detects_session_not_found`` — server GC
variant ("session not found", "unknown session").
* ``test_is_session_expired_is_case_insensitive``.
* ``test_is_session_expired_rejects_unrelated_errors`` — narrow-scope
canary: random RuntimeError / ValueError / 401 don't trigger.
* ``test_is_session_expired_rejects_interrupted_error`` — user cancel
must never route through reconnect.
* ``test_is_session_expired_rejects_empty_message``.
Handler integration tests:
* ``test_call_tool_handler_reconnects_on_session_expired`` — reporter's
full repro: first call raises "Invalid or expired session", handler
signals ``_reconnect_event``, retries once, returns the retry's
success result with no ``error`` key.
* ``test_call_tool_handler_non_session_expired_error_falls_through``
— preserved-behaviour canary: random tool failures do NOT trigger
reconnect.
* ``test_session_expired_handler_returns_none_without_loop`` —
defensive: cold-start / shutdown race.
* ``test_session_expired_handler_returns_none_without_server_record``
— torn-down server falls through cleanly.
* ``test_session_expired_handler_returns_none_when_retry_also_fails``
— no retry loop on repeated failure.
Parametrised across all 4 non-``tools/call`` handlers:
* ``test_non_tool_handlers_also_reconnect_on_session_expired``
[list_resources / read_resource / list_prompts / get_prompt].
**15 of 16 fail on clean ``origin/main`` (``6fb69229``)** with
``ImportError: cannot import name '_is_session_expired_error'``
— the fix's surface symbols don't exist there yet. The 1 passing
test is an ordering artefact of pytest-xdist worker collection.
Validation
----------
``source venv/bin/activate && python -m pytest
tests/tools/test_mcp_tool_session_expired.py -q`` → **16 passed**.
Broader MCP suite (5 files:
``test_mcp_tool.py``, ``test_mcp_tool_401_handling.py``,
``test_mcp_tool_session_expired.py``, ``test_mcp_reconnect_signal.py``,
``test_mcp_oauth.py``) → **230 passed, 0 regressions**.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add event hook to httpx.AsyncClient in MCP HTTP transport that strips
Authorization headers when a redirect targets a different origin,
preventing credential leakage to third-party servers.
``tools/mcp_oauth.py`` relied on ``assert _oauth_port is not None`` to
guard the module-level port set by ``build_oauth_auth``. Python's
``-O`` / ``-OO`` optimization flags strip ``assert`` statements
entirely, so a deployment that runs ``python -O -m hermes ...``
silently loses the check: ``_oauth_port`` stays ``None`` and the
failure surfaces much later as an obscure ``int()`` or
``http.server.HTTPServer((host, None))`` TypeError rather than the
intended "OAuth callback port not set" signal.
Replace with an explicit ``if … raise RuntimeError(...)`` so the
invariant is preserved regardless of the interpreter's optimization
level. Docstring updated to document the new exception.
Found during a proactive audit of ``assert`` statements in
non-test code paths.
OAuth client information and token responses from the MCP SDK contain
Pydantic AnyUrl fields (client_uri, redirect_uris, etc.). The previous
model_dump() call returned a dict with these AnyUrl objects still as
their native Python type, which then crashed json.dumps with:
TypeError: Object of type AnyUrl is not JSON serializable
This caused any OAuth-based MCP server (e.g. alphaxiv) to fail
registration with an "OAuth flow error" traceback during startup.
Adding mode="json" tells Pydantic to serialize all fields to
JSON-compatible primitives (AnyUrl -> str, datetime -> ISO string, etc.)
before returning the dict, so the standard json.dumps can handle it.
Three call sites fixed:
- HermesTokenStorage.set_tokens
- HermesTokenStorage.set_client_info
- build_oauth_auth pre-registration write
Cron jobs can now specify a per-job working directory. When set, the job
runs as if launched from that directory: AGENTS.md / CLAUDE.md /
.cursorrules from that dir are injected into the system prompt, and the
terminal / file / code-exec tools use it as their cwd (via TERMINAL_CWD).
When unset, old behaviour is preserved (no project context files, tools
use the scheduler's cwd).
Requested by @bluthcy.
## Mechanism
- cron/jobs.py: create_job / update_job accept 'workdir'; validated to
be an absolute existing directory at create/update time.
- cron/scheduler.py run_job: if job.workdir is set, point TERMINAL_CWD
at it and flip skip_context_files to False before building the agent.
Restored in finally on every exit path.
- cron/scheduler.py tick: workdir jobs run sequentially (outside the
thread pool) because TERMINAL_CWD is process-global. Workdir-less jobs
still run in the parallel pool unchanged.
- tools/cronjob_tools.py + hermes_cli/cron.py + hermes_cli/main.py:
expose 'workdir' via the cronjob tool and 'hermes cron create/edit
--workdir ...'. Empty string on edit clears the field.
## Validation
- tests/cron/test_cron_workdir.py (21 tests): normalize, create, update,
JSON round-trip via cronjob tool, tick partition (workdir jobs run on
the main thread, not the pool), run_job env toggle + restore in finally.
- Full targeted suite (tests/cron/, test_cronjob_tools.py, test_cron.py,
test_config_cwd_bridge.py, test_worktree.py): 314/314 passed.
- Live smoke: hermes cron create --workdir $(pwd) works; relative path
rejected; list shows 'Workdir:'; edit --workdir '' clears.
When a subagent in delegate_task times out before making its first LLM
request, write a structured diagnostic file under
~/.hermes/logs/subagent-timeout-<sid>-<ts>.log capturing enough state
for the user (and us) to debug the hang. The old error message —
'Subagent timed out after Ns with no response. The child may be stuck
on a slow API call or unresponsive network request.' — gave no
observability for the 0-API-call case, which is the hardest to reason
about remotely.
The diagnostic captures:
- timeout config vs actual duration
- goal (truncated to 1000 chars)
- child config: model, provider, api_mode, base_url, max_iterations,
quiet_mode, platform, _delegate_role, _delegate_depth
- enabled_toolsets + loaded tool names
- system prompt byte/char count (catches oversized prompts that
providers silently choke on)
- tool schema count + byte size
- child's get_activity_summary() snapshot
- Python stack of the worker thread at the moment of timeout
(reveals whether the hang is in credential resolution, transport,
prompt construction, etc.)
Wiring:
- _run_single_child captures the worker thread via a small wrapper
around child.run_conversation so we can look up its stack at
timeout.
- After a FuturesTimeoutError, we pull child.get_activity_summary()
to read api_call_count. If 0 AND it was a timeout (not a raise),
_dump_subagent_timeout_diagnostic() is invoked.
- The returned path is surfaced in the error string so the parent
agent (and therefore the user / gateway) sees exactly where to look.
- api_calls > 0 timeouts keep the old 'stuck on slow API call'
phrasing since that's the correct diagnosis for those.
This does NOT change any behavior for successful subagent runs,
non-timeout errors, or subagents that made at least one API call
before hanging.
Tests: 7 cases (tests/tools/test_delegate_subagent_timeout_diagnostic.py)
- output format + required sections + field values
- long-goal truncation with [truncated] marker
- missing / already-exited worker thread branches
- unwritable HERMES_HOME/logs/ returns None without raising
- _run_single_child wiring: 0 API calls → dump + diagnostic_path in error
- _run_single_child wiring: N>0 API calls → no dump, old message
Refs: #14726
Make the main-branch test suite pass again. Most failures were tests
still asserting old shapes after recent refactors; two were real source
bugs.
Source fixes:
- tools/mcp_tool.py: _kill_orphaned_mcp_children() slept 2s on every
shutdown even when no tracked PIDs existed, making test_shutdown_is_parallel
measure ~3s for 3 parallel 1s shutdowns. Early-return when pids is empty.
- hermes_cli/tips.py: tip 105 was 157 chars; corpus max is 150.
Test fixes (mostly stale mock targets / missing fixture fields):
- test_zombie_process_cleanup, test_agent_cache: patch run_agent.cleanup_vm
(the local name bound at import), not tools.terminal_tool.cleanup_vm.
- test_browser_camofox: patch tools.browser_camofox.load_config, not
hermes_cli.config.load_config (the source module, not the resolved one).
- test_flush_memories_codex._chat_response_with_memory_call: add
finish_reason, tool_call.id, tool_call.type so the chat_completions
transport normalizer doesn't AttributeError.
- test_concurrent_interrupt: polling_tool signature now accepts
messages= kwarg that _invoke_tool() passes through.
- test_minimax_provider: add _fallback_chain=[] to the __new__'d agent
so switch_model() doesn't AttributeError.
- test_skills_config: SKILLS_DIR MagicMock + .rglob stopped working
after the scanner switched to agent.skill_utils.iter_skill_index_files
(os.walk-based). Point SKILLS_DIR at a real tmp_path and patch
agent.skill_utils.get_external_skills_dirs.
- test_browser_cdp_tool: browser_cdp toolset was intentionally split into
'browser-cdp' (commit 96b0f3700) so its stricter check_fn doesn't gate
the whole browser toolset; test now expects 'browser-cdp'.
- test_registry: add tools.browser_dialog_tool to the expected
builtin-discovery set (PR #14540 added it).
- test_file_tools TestPatchHints: patch_tool surfaces hints as a '_hint'
key on the JSON payload, not inline '[Hint: ...' text.
- test_write_deny test_hermes_env: resolve .env via get_hermes_home() so
the path matches the profile-aware denylist under hermetic HERMES_HOME.
- test_checkpoint_manager test_falls_back_to_parent: guard the walk-up
so a stray /tmp/pyproject.toml on the host doesn't pick up /tmp as the
project root.
- test_quick_commands: set cli.session_id in the __new__'d CLI so the
alias-args path doesn't trip AttributeError when fuzzy-matching leaks
a skill command across xdist test distribution.
faster-whisper's device="auto" picks CUDA when ctranslate2's wheel
ships CUDA shared libs, even on hosts without the NVIDIA runtime
(libcublas.so.12 / libcudnn*). On those hosts the model often loads
fine but transcribe() fails at first dlopen, and the broken model
stays cached in the module-global — every subsequent voice message
in the gateway process fails identically until restart.
- Add _load_local_whisper_model() wrapper: try auto, catch missing-lib
errors, retry on device=cpu compute_type=int8.
- Wrap transcribe() with the same fallback: evict cached model, reload
on CPU, retry once. Required because the dlopen failure only surfaces
at first kernel launch, not at model construction.
- Narrow marker list (libcublas, libcudnn, libcudart, 'cannot be loaded',
'no kernel image is available', 'no CUDA-capable device', driver
mismatch). Deliberately excludes 'CUDA out of memory' and similar —
those are real runtime failures that should surface, not be silently
retried on CPU.
- Tests for load-time fallback, runtime fallback (with cached-model
eviction verified), and the OOM non-fallback path.
Reported via Telegram voice-message dumps on WSL2 hosts where libcublas
isn't installed by default.
Local llama.cpp servers (e.g. ggml-org/llama.cpp:full-cuda) fail the entire
request with HTTP 400 'Unable to generate parser for this template. ...
Unrecognized schema: "object"' when any tool schema contains shapes its
json-schema-to-grammar converter can't handle:
* 'type': 'object' without 'properties'
* bare string schema values ('additionalProperties: "object"')
* 'type': ['X', 'null'] arrays (nullable form)
Cloud providers accept these silently, so they ship from external MCP
servers (Atlassian, GCloud, Datadog) and from a couple of our own tools.
Changes
- tools/schema_sanitizer.py: walks the finalized tool list right before it
leaves get_tool_definitions() and repairs the hostile shapes in a deep
copy. No-op on well-formed schemas. Recurses into properties, items,
additionalProperties, anyOf/oneOf/allOf, and $defs.
- model_tools.get_tool_definitions(): invoke the sanitizer as the last
step so all paths (built-in, MCP, plugin, dynamically-rebuilt) get
covered uniformly.
- tools/browser_cdp_tool.py, tools/mcp_tool.py: fix our own bare-object
schemas so sanitization isn't load-bearing for in-repo tools.
- tui_gateway/server.py: _load_enabled_toolsets() was passing
include_default_mcp_servers=False at runtime. That's the config-editing
variant (see PR #3252) — it silently drops every default MCP server
from the TUI's enabled_toolsets, which is why the TUI didn't hit the
llama.cpp crash (no MCP tools sent at all). Switch to True so TUI
matches CLI behavior.
Tests
tests/tools/test_schema_sanitizer.py (17 tests) covers the individual
failure modes, well-formed pass-through, deep-copy isolation, and
required-field pruning.
E2E: loaded the default 'hermes-cli' toolset with MCP discovery and
confirmed all 27 resolved tool schemas pass a llama.cpp-compatibility
walk (no 'object' node missing 'properties', no bare-string schema
values).
* docs: browser CDP supervisor design (for upcoming PR)
Design doc ahead of implementation — dialog + iframe detection/interaction
via a persistent CDP supervisor. Covers backend capability matrix (verified
live 2026-04-23), architecture, lifecycle, policy, agent surface, PR split,
non-goals, and test plan.
Supersedes #12550.
No code changes in this commit.
* feat(browser): add persistent CDP supervisor for dialog + frame detection
Single persistent CDP WebSocket per Hermes task_id that subscribes to
Page/Runtime/Target events and maintains thread-safe state for pending
dialogs, frame tree, and console errors.
Supervisor lives in its own daemon thread running an asyncio loop;
external callers use sync API (snapshot(), respond_to_dialog()) that
bridges onto the loop.
Auto-attaches to OOPIF child targets via Target.setAutoAttach{flatten:true}
and enables Page+Runtime on each so iframe-origin dialogs surface through
the same supervisor.
Dialog policies: must_respond (default, 300s safety timeout),
auto_dismiss, auto_accept.
Frame tree capped at 30 entries + OOPIF depth 2 to keep snapshot
payloads bounded on ad-heavy pages.
E2E verified against real Chrome via smoke test — detects + responds
to main-frame alerts, iframe-contentWindow alerts, preserves frame
tree, graceful no-dialog error path, clean shutdown.
No agent-facing tool wiring in this commit (comes next).
* feat(browser): add browser_dialog tool wired to CDP supervisor
Agent-facing response-only tool. Schema:
action: 'accept' | 'dismiss' (required)
prompt_text: response for prompt() dialogs (optional)
dialog_id: disambiguate when multiple dialogs queued (optional)
Handler:
SUPERVISOR_REGISTRY.get(task_id).respond_to_dialog(...)
check_fn shares _browser_cdp_check with browser_cdp so both surface and
hide together. When no supervisor is attached (Camofox, default
Playwright, or no browser session started yet), tool is hidden; if
somehow invoked it returns a clear error pointing the agent to
browser_navigate / /browser connect.
Registered in _HERMES_CORE_TOOLS and the browser / hermes-acp /
hermes-api-server toolsets alongside browser_cdp.
* feat(browser): wire CDP supervisor into session lifecycle + browser_snapshot
Supervisor lifecycle:
* _get_session_info lazy-starts the supervisor after a session row is
materialized — covers every backend code path (Browserbase, cdp_url
override, /browser connect, future providers) with one hook.
* cleanup_browser(task_id) stops the supervisor for that task first
(before the backend tears down CDP).
* cleanup_all_browsers() calls SUPERVISOR_REGISTRY.stop_all().
* /browser connect eagerly starts the supervisor for task 'default'
so the first snapshot already shows pending_dialogs.
* /browser disconnect stops the supervisor.
CDP URL resolution for the supervisor:
1. BROWSER_CDP_URL / browser.cdp_url override.
2. Fallback: session_info['cdp_url'] from cloud providers (Browserbase).
browser_snapshot merges supervisor state (pending_dialogs + frame_tree)
into its JSON output when a supervisor is active — the agent reads
pending_dialogs from the snapshot it already requests, then calls
browser_dialog to respond. No extra tool surface.
Config defaults:
* browser.dialog_policy: 'must_respond' (new)
* browser.dialog_timeout_s: 300 (new)
No version bump — new keys deep-merge into existing browser section.
Deadlock fix in supervisor event dispatch:
* _on_dialog_opening and _on_target_attached used to await CDP calls
while the reader was still processing an event — but only the reader
can set the response Future, so the call timed out.
* Both now fire asyncio.create_task(...) so the reader stays pumping.
* auto_dismiss/auto_accept now actually close the dialog immediately.
Tests (tests/tools/test_browser_supervisor.py, 11 tests, real Chrome):
* supervisor start/snapshot
* main-frame alert detection + dismiss
* iframe.contentWindow alert
* prompt() with prompt_text reply
* respond with no pending dialog -> clean error
* auto_dismiss clears on event
* registry idempotency
* registry stop -> snapshot reports inactive
* browser_dialog tool no-supervisor error
* browser_dialog invalid action
* browser_dialog end-to-end via tool handler
xdist-safe: chrome_cdp fixture uses a per-worker port.
Skipped when google-chrome/chromium isn't installed.
* docs(browser): document browser_dialog tool + CDP supervisor
- user-guide/features/browser.md: new browser_dialog section with
workflow, availability gate, and dialog_policy table
- reference/tools-reference.md: row for browser_dialog, tool count
bumped 53 -> 54, browser tools count 11 -> 12
- reference/toolsets-reference.md: browser_dialog added to browser
toolset row with note on pending_dialogs / frame_tree snapshot fields
Full design doc lives at
developer-guide/browser-supervisor.md (committed earlier).
* fix(browser): reconnect loop + recent_dialogs for Browserbase visibility
Found via Browserbase E2E test that revealed two production-critical issues:
1. **Supervisor WebSocket drops when other clients disconnect.** Browserbase's
CDP proxy tears down our long-lived WebSocket whenever a short-lived
client (e.g. agent-browser CLI's per-command CDP connection) disconnects.
Fixed with a reconnecting _run loop that re-attaches with exponential
backoff on drops. _page_session_id and _child_sessions are reset on each
reconnect; pending_dialogs and frames are preserved across reconnects.
2. **Browserbase auto-dismisses dialogs server-side within ~10ms.** Their
Playwright-based CDP proxy dismisses alert/confirm/prompt before our
Page.handleJavaScriptDialog call can respond. So pending_dialogs is
empty by the time the agent reads a snapshot on Browserbase.
Added a recent_dialogs ring buffer (capacity 20) that retains a
DialogRecord for every dialog that opened, with a closed_by tag:
* 'agent' — agent called browser_dialog
* 'auto_policy' — local auto_dismiss/auto_accept fired
* 'watchdog' — must_respond timeout auto-dismissed (300s default)
* 'remote' — browser/backend closed it on us (Browserbase)
Agents on Browserbase now see the dialog history with closed_by='remote'
so they at least know a dialog fired, even though they couldn't respond.
3. **Page.javascriptDialogClosed matching bug.** The event doesn't include a
'message' field (CDP spec has only 'result' and 'userInput') but our
_on_dialog_closed was matching on message. Fixed to match by session_id
+ oldest-first, with a safety assumption that only one dialog is in
flight per session (the JS thread is blocked while a dialog is up).
Docs + tests updated:
* browser.md: new availability matrix showing the three backends and
which mode (pending / recent / response) each supports
* developer-guide/browser-supervisor.md: three-field snapshot schema
with closed_by semantics
* test_browser_supervisor.py: +test_recent_dialogs_ring_buffer (12/12
passing against real Chrome)
E2E verified both backends:
* Local Chrome via /browser connect: detect + respond full workflow
(smoke_supervisor.py all 7 scenarios pass)
* Browserbase: detect via recent_dialogs with closed_by='remote'
(smoke_supervisor_browserbase_v2.py passes)
Camofox remains out of scope (REST-only, no CDP) — tracked for
upstream PR 3.
* feat(browser): XHR bridge for dialog response on Browserbase (FIXED)
Browserbase's CDP proxy auto-dismisses native JS dialogs within ~10ms, so
Page.handleJavaScriptDialog calls lose the race. Solution: bypass native
dialogs entirely.
The supervisor now injects Page.addScriptToEvaluateOnNewDocument with a
JavaScript override for window.alert/confirm/prompt. Those overrides
perform a synchronous XMLHttpRequest to a magic host
('hermes-dialog-bridge.invalid'). We intercept those XHRs via Fetch.enable
with a requestStage=Request pattern.
Flow when a page calls alert('hi'):
1. window.alert override intercepts, builds XHR GET to
http://hermes-dialog-bridge.invalid/?kind=alert&message=hi
2. Sync XHR blocks the page's JS thread (mirrors real dialog semantics)
3. Fetch.requestPaused fires on our WebSocket; supervisor surfaces
it as a pending dialog with bridge_request_id set
4. Agent reads pending_dialogs from browser_snapshot, calls browser_dialog
5. Supervisor calls Fetch.fulfillRequest with JSON body:
{accept: true|false, prompt_text: '...', dialog_id: 'd-N'}
6. The injected script parses the body, returns the appropriate value
from the override (undefined for alert, bool for confirm, string|null
for prompt)
This works identically on Browserbase AND local Chrome — no native dialog
ever fires, so Browserbase's auto-dismiss has nothing to race. Dialog
policies (must_respond / auto_dismiss / auto_accept) all still work.
Bridge is installed on every attached session (main page + OOPIF child
sessions) so iframe dialogs are captured too.
Native-dialog path kept as a fallback for backends that don't auto-dismiss
(so a page that somehow bypasses our override — e.g. iframes that load
after Fetch.enable but before the init-script runs — still gets observed
via Page.javascriptDialogOpening).
E2E VERIFIED:
* Local Chrome: 13/13 pytest tests green (12 original + new
test_bridge_captures_prompt_and_returns_reply_text that asserts
window.__ret === 'AGENT-SUPPLIED-REPLY' after agent responds)
* Browserbase: smoke_bb_bridge_v2.py runs 4/4 PASS:
- alert('BB-ALERT-MSG') dismiss → page.alert_ret = undefined ✓
- prompt('BB-PROMPT-MSG', 'default-xyz') accept with 'AGENT-REPLY'
→ page.prompt_ret === 'AGENT-REPLY' ✓
- confirm('BB-CONFIRM-MSG') accept → page.confirm_ret === true ✓
- confirm('BB-CONFIRM-MSG') dismiss → page.confirm_ret === false ✓
Docs updated in browser.md and developer-guide/browser-supervisor.md —
availability matrix now shows Browserbase at full parity with local
Chrome for both detection and response.
* feat(browser): cross-origin iframe interaction via browser_cdp(frame_id=...)
Adds iframe interaction to the CDP supervisor PR (was queued as PR 2).
Design: browser_cdp gets an optional frame_id parameter. When set, the
tool looks up the frame in the supervisor's frame_tree, grabs its child
cdp_session_id (OOPIF session), and dispatches the CDP call through the
supervisor's already-connected WebSocket via run_coroutine_threadsafe.
Why not stateless: on Browserbase, each fresh browser_cdp WebSocket
must re-negotiate against a signed connectUrl. The session info carries
a specific URL that can expire while the supervisor's long-lived
connection stays valid. Routing via the supervisor sidesteps this.
Agent workflow:
1. browser_snapshot → frame_tree.children[] shows OOPIFs with is_oopif=true
2. browser_cdp(method='Runtime.evaluate', frame_id=<OOPIF frame_id>,
params={'expression': 'document.title', 'returnByValue': True})
3. Supervisor dispatches the call on the OOPIF's child session
Supervisor state fixes needed along the way:
* _on_frame_detached now skips reason='swap' (frame migrating processes)
* _on_frame_detached also skips when the frame is an OOPIF with a live
child session — Browserbase fires spurious remove events when a
same-origin iframe gets promoted to OOPIF
* _on_target_detached clears cdp_session_id but KEEPS the frame record
so the agent still sees the OOPIF in frame_tree during transient
session flaps
E2E VERIFIED on Browserbase (smoke_bb_iframe_agent_path.py):
browser_cdp(method='Runtime.evaluate',
params={'expression': 'document.title', 'returnByValue': True},
frame_id=<OOPIF>)
→ {'success': True, 'result': {'value': 'Example Domain'}}
The iframe is <iframe src='https://example.com/'> inside a top-level
data: URL page on a real Browserbase session. The agent Runtime.evaluates
INSIDE the cross-origin iframe and gets example.com's title back.
Tests (tests/tools/test_browser_supervisor.py — 16 pass total):
* test_browser_cdp_frame_id_routes_via_supervisor — injects fake OOPIF,
verifies routing via supervisor, Runtime.evaluate returns 1+1=2
* test_browser_cdp_frame_id_missing_supervisor — clean error when no
supervisor attached
* test_browser_cdp_frame_id_not_in_frame_tree — clean error on bad
frame_id
Docs (browser.md and developer-guide/browser-supervisor.md) updated with
the iframe workflow, availability matrix now shows OOPIF eval as shipped
for local Chrome + Browserbase.
* test(browser): real-OOPIF E2E verified manually + chrome_cdp uses --site-per-process
When asked 'did you test the iframe stuff' I had only done a mocked
pytest (fake injected OOPIF) plus a Browserbase E2E. Closed the
local-Chrome real-OOPIF gap by writing /tmp/dialog-iframe-test/
smoke_local_oopif.py:
* 2 http servers on different hostnames (localhost:18905 + 127.0.0.1:18906)
* Chrome with --site-per-process so the cross-origin iframe becomes a
real OOPIF in its own process
* Navigate, find OOPIF in supervisor.frame_tree, call
browser_cdp(method='Runtime.evaluate', frame_id=<OOPIF>) which routes
through the supervisor's child session
* Asserts iframe document.title === 'INNER-FRAME-XYZ' (from the
inner page, retrieved via OOPIF eval)
PASSED on 2026-04-23.
Tried to embed this as a pytest but hit an asyncio version quirk between
venv (3.11) and the system python (3.13) — Page.navigate hangs in the
pytest harness but works in standalone. Left a self-documenting skip
test that points to the smoke script + describes the verification.
chrome_cdp fixture now passes --site-per-process so future iframe tests
can rely on OOPIF behavior.
Result: 16 pass + 1 documented-skip = 17 tests in
tests/tools/test_browser_supervisor.py.
* docs(browser): add dialog_policy + dialog_timeout_s to configuration.md, fix tool count
Pre-merge docs audit revealed two gaps:
1. user-guide/configuration.md browser config example was missing the
two new dialog_* knobs. Added with a short table explaining
must_respond / auto_dismiss / auto_accept semantics and a link to
the feature page for the full workflow.
2. reference/tools-reference.md header said '54 built-in tools' — real
count on main is 54, this branch adds browser_dialog so it's 55.
Fixed the header. (browser count was already correctly bumped
11 -> 12 in the earlier docs commit.)
No code changes.
* feat(config): make tool output truncation limits configurable
Port from anomalyco/opencode#23770: expose a new `tool_output` config
section so users can tune the hardcoded truncation caps that apply to
terminal output and read_file pagination.
Three knobs under `tool_output`:
- max_bytes (default 50_000) — terminal stdout/stderr cap
- max_lines (default 2000) — read_file pagination cap
- max_line_length (default 2000) — per-line cap in line-numbered view
All three keep their existing hardcoded values as defaults, so behaviour
is unchanged when the section is absent. Power users on big-context
models can raise them; small-context local models can lower them.
Implementation:
- New `tools/tool_output_limits.py` reads the section with defensive
fallback (missing/invalid values → defaults, never raises).
- `tools/terminal_tool.py` MAX_OUTPUT_CHARS now comes from
get_max_bytes().
- `tools/file_operations.py` normalize_read_pagination() and
_add_line_numbers() now pull the limits at call time.
- `hermes_cli/config.py` DEFAULT_CONFIG gains the `tool_output` section
so `hermes setup` writes defaults into fresh configs.
- Docs page `user-guide/configuration.md` gains a "Tool Output
Truncation Limits" section with large-context and small-context
example configs.
Tests (18 new in tests/tools/test_tool_output_limits.py):
- Default resolution with missing / malformed / non-dict config.
- Full and partial user overrides.
- Coercion of bad values (None, negative, wrong type, str int).
- Shortcut accessors delegate correctly.
- DEFAULT_CONFIG exposes the section with the right defaults.
- Integration: normalize_read_pagination clamps to the configured
max_lines.
* feat(skills): add design-md skill for Google's DESIGN.md spec
Built-in skill under skills/creative/ that teaches the agent to author,
lint, diff, and export DESIGN.md files — Google's open-source
(Apache-2.0) format for describing a visual identity to coding agents.
Covers:
- YAML front matter + markdown body anatomy
- Full token schema (colors, typography, rounded, spacing, components)
- Canonical section order + duplicate-heading rejection
- Component property whitelist + variants-as-siblings pattern
- CLI workflow via 'npx @google/design.md' (lint/diff/export/spec)
- Lint rule reference including WCAG contrast checks
- Common YAML pitfalls (quoted hex, negative dimensions, dotted refs)
- Starter template at templates/starter.md
Package verified live on npm (@google/design.md@0.1.1).
MCP stdio servers' stderr was being dumped directly onto the user's
terminal during hermes launch. Servers like FastMCP-based ones print a
large ASCII banner at startup; slack-mcp-server emits JSON logs; etc.
With prompt_toolkit / Rich rendering the TUI concurrently, these
unsolicited writes corrupt the terminal state — hanging the session
~80% of the time for one user with Google Ads Tools + slack-mcp
configured, forcing Ctrl+C and restart loops.
Root cause: `stdio_client(server_params)` in tools/mcp_tool.py was
called without `errlog=`, and the SDK's default is `sys.stderr` —
i.e. the real parent-process stderr, which is the TTY.
Fix: open a shared, append-mode log at $HERMES_HOME/logs/mcp-stderr.log
(created once per process, line-buffered, real fd required by asyncio's
subprocess machinery) and pass it as `errlog` to every stdio_client.
Each server's spawn writes a timestamped header so the shared log stays
readable when multiple servers are running. Falls back to /dev/null if
the log file cannot be opened.
Verified by E2E spawning a subprocess with the log fd as its stderr:
banner lines land in the log file, nothing reaches the calling TTY.
The 300s default was too tight for high-reasoning models on non-trivial
delegated tasks — e.g. gpt-5.5 xhigh reviewing 12 files would burn >5min
on reasoning tokens before issuing its first tool call, tripping the
hard wall-clock timeout with 0 api_calls logged.
- tools/delegate_tool.py: DEFAULT_CHILD_TIMEOUT 300 -> 600
- hermes_cli/config.py: surface delegation.child_timeout_seconds in
DEFAULT_CONFIG so it's discoverable (previously the key was read by
_get_child_timeout() but absent from the default config schema)
Users can still override via config.yaml delegation.child_timeout_seconds
or DELEGATION_CHILD_TIMEOUT_SECONDS env var (floor 30s, no ceiling).
Fixes a broader class of 'tools.function.parameters is not a valid
moonshot flavored json schema' errors on Nous / OpenRouter aggregators
routing to moonshotai/kimi-k2.6 with MCP tools loaded.
## Moonshot sanitizer (agent/moonshot_schema.py, new)
Model-name-routed (not base-URL-routed) so Nous / OpenRouter users are
covered alongside api.moonshot.ai. Applied in
ChatCompletionsTransport.build_kwargs when is_moonshot_model(model).
Two repairs:
1. Fill missing 'type' on every property / items / anyOf-child schema
node (structural walk — only schema-position dicts are touched, not
container maps like properties/$defs).
2. Strip 'type' at anyOf parents; Moonshot rejects it.
## MCP normalizer hardened (tools/mcp_tool.py)
Draft-07 $ref rewrite from PR #14802 now also does:
- coerce missing / null 'type' on object-shaped nodes (salvages #4897)
- prune 'required' arrays to names that exist in 'properties'
(salvages #4651; Gemini 400s on dangling required)
- apply recursively, not just top-level
These repairs are provider-agnostic so the same MCP schema is valid on
OpenAI, Anthropic, Gemini, and Moonshot in one pass.
## Crash fix: safe getattr for Tool.inputSchema
_convert_mcp_schema now uses getattr(t, 'inputSchema', None) so MCP
servers whose Tool objects omit the attribute entirely no longer abort
registration (salvages #3882).
## Validation
- tests/agent/test_moonshot_schema.py: 27 new tests (model detection,
missing-type fill, anyOf-parent strip, non-mutation, real-world MCP
shape)
- tests/tools/test_mcp_tool.py: 7 new tests (missing / null type,
required pruning, nested repair, safe getattr)
- tests/agent/transports/test_chat_completions.py: 2 new integration
tests (Moonshot route sanitizes, non-Moonshot route doesn't)
- Targeted suite: 49 passed
- E2E via execute_code with a realistic MCP tool carrying all three
Moonshot rejection modes + dangling required + draft-07 refs:
sanitizer produces a schema valid on Moonshot and Gemini
- _stdio_pids: set → Dict[int,str] tracks pid→server_name
- SIGTERM-first with 2s grace before SIGKILL escalation
- hasattr guard for SIGKILL on platforms without it
- Updated tests for dict-based tracking and 3-phase kill sequence
The original regex only matched relative paths (./foo/.env or bare
.env), so the exact command from the bug report —
`cp /opt/data/.env.local /opt/data/.env` — did not trigger approval.
Broaden the leading-path prefix to accept an absolute leading slash
alongside ./ and ../, and add regressions for the bug-report command
and its redirection variant.
Previously delegate_task exposed 'max_iterations' in its JSON schema and used
`max_iterations or default_max_iter` — so a model guessing conservatively (or
copy-pasting a docstring hint like 'Only set lower for simple tasks') could
silently shrink a subagent's budget below the user's configured
delegation.max_iterations. One such call this session capped a deep forensic
audit at 40 iterations while the user's config was set to 250.
Changes:
- Drop 'max_iterations' from DELEGATE_TASK_SCHEMA['parameters']['properties'].
Models can no longer emit it.
- In delegate_task(): ignore any caller-supplied max_iterations, always use
delegation.max_iterations from config. Log at debug if a stale schema or
internal caller still passes one through.
- Keep the Python kwarg on the function signature for internal callers
(_build_child_agent tests pass it through the plumbing layer).
- Update test_schema_valid to assert the param is now absent (intentional
contract change, not a change-detector).
Replaces the blanket 'always allow' change from the previous commit with
an opt-in config flag so users who want belt-and-suspenders security can
still get the keyword scan on skill_manage output.
## Default behavior (flag off)
skill_manage(action='create'|'edit'|'patch') no longer runs the keyword
scanner. The agent can write skills that mention risky keywords in prose
(documenting what reviewers should watch for, describing cache-bust
semantics in a PR-review skill, referencing AGENTS.md, etc.) without
getting blocked.
Rationale: the agent can already execute the same code paths via
terminal() with no gate, so the scan adds friction without meaningful
security against a compromised or malicious agent.
## Opt-in behavior (flag on)
Set skills.guard_agent_created: true in config.yaml to get the original
behavior back. Scanner runs on every skill_manage write; dangerous
verdicts surface as a tool error the agent can react to (retry without
the flagged content).
## External hub installs unaffected
trusted/community sources (hermes skills install) always get scanned
regardless of this flag. The gate is specifically for skill_manage,
which only agents call.
## Changes
- hermes_cli/config.py: add skills.guard_agent_created: False to DEFAULT_CONFIG
- tools/skill_manager_tool.py: _guard_agent_created_enabled() reads the flag;
_security_scan_skill() short-circuits to None when the flag is off
- tools/skills_guard.py: restore INSTALL_POLICY['agent-created'] =
('allow', 'allow', 'ask') so the scan remains strict when it does run
- tests/tools/test_skills_guard.py: restore original ask/force tests
- tests/tools/test_skill_manager_tool.py: new TestSecurityScanGate class
covering both flag states + config error handling
## Validation
- tests/tools/test_skills_guard.py + test_skill_manager_tool.py: 115/115 pass
- E2E: flagged-keyword skill creates with default config, blocks with flag on
The security scanner is meant to protect against hostile external skills
pulled from GitHub via hermes skills install — trusted/community policies
block or ask on dangerous verdicts accordingly. But agent-created skills
(from skill_manage) run in the same process as the agent that wrote them.
The agent can already execute the same code paths via terminal() with no
gate, so the ask-on-dangerous policy adds friction without meaningful
security.
Concrete trigger: an agent writing a PR-review skill that describes
cache-busting or persistence semantics in prose gets blocked because
those words appear in the patterns list. The skill isn't actually doing
anything dangerous — it's just documenting what reviewers should watch
for in other PRs.
Change: agent-created dangerous verdict maps to 'allow' instead of 'ask'.
External hub installs (trusted/community) keep their stricter policies
intact. Tests updated: renamed test_dangerous_agent_created_asks →
test_dangerous_agent_created_allowed; renamed force-override test and
updated assertion since force is now a no-op for agent-created (the allow
branch returns first).
The environment-snapshot login shell was auto-sourcing only ~/.bashrc when
building the PATH snapshot. On Debian/Ubuntu the default ~/.bashrc starts
with a non-interactive short-circuit:
case $- in *i*) ;; *) return;; esac
Sourcing it from a non-interactive shell returns before any PATH export
below that guard runs. Node version managers like n and nvm append their
PATH line under that guard, so Hermes was capturing a PATH without
~/n/bin — and the terminal tool saw 'node: command not found' even when
node was on the user's interactive shell PATH.
Expand the auto-source list (when auto_source_bashrc is on) to:
~/.profile → ~/.bash_profile → ~/.bashrc
~/.profile and ~/.bash_profile have no interactivity guard — installers
that write their PATH there (n's n-install, nvm's curl installer on most
setups) take effect. ~/.bashrc still runs last to preserve behaviour for
users who put PATH logic there without the guard.
Added two tests covering the new behaviour plus an E2E test that spins up
a real LocalEnvironment with a guard-prefixed ~/.bashrc and a ~/.profile
PATH export, and verifies the captured snapshot PATH contains the profile
entry.
When a newly-bundled skill's name collides with a pre-existing user
skill, sync silently kept the user's copy. Users never learned that
a bundled version shipped by that name.
Now (on non-quiet sync only) print:
⚠ <name>: bundled version shipped but you already have a local
skill by this name — yours was kept. Run `hermes skills reset
<name>` to replace it with the bundled version.
No behavior change to manifest writes or to the kept user copy —
purely additive warning on the existing collision-skip path.
When a new bundled skill's name collided with a pre-existing user skill
(from hub, custom, or leftover), sync_skills() recorded the bundled hash
in the manifest even though the on-disk copy was unrelated to bundled.
On the next sync, user_hash != origin_hash (bundled_hash) marked the
skill as "user-modified" permanently, blocking all bundled updates for
that skill until the user ran `hermes skills reset`.
Fix: only baseline the manifest entry when the user's on-disk copy is
byte-identical to bundled (safe to track — this is the reset re-sync or
coincidentally-identical install case). Otherwise skip the manifest
write entirely: the on-disk skill is unrelated to bundled and shouldn't
be tracked as if it were.
This preserves reset_bundled_skill()'s re-baseline flow (its post-delete
sync still writes to the manifest when user copy matches bundled) while
fixing the poisoning scenario for genuinely unrelated collisions.
Adds two tests following the existing test_failed_copy_does_not_poison_manifest
pattern: one verifying the manifest stays clean after a collision with
differing content, one verifying no false user_modified flag on resync.
When _resolve_tirith_path() returns None (e.g. install failed on
unsupported platform or all resolution paths exhausted), the function
passed None directly to subprocess.run(), causing a TypeError instead
of respecting the fail_open config.
Add a None check before the subprocess call that allows or blocks
according to the configured fail_open policy, matching the existing
error handling behavior for OSError and TimeoutExpired.
When override_acp_command was passed to _build_child_agent, it failed to
override effective_provider to 'copilot-acp' and effective_api_mode to
'chat_completions'. This caused the child AIAgent to inherit the parent's
native API configuration (e.g. Anthropic) and attempt real HTTP requests
using the parent's API key, leading to HTTP 401 errors and completely
bypassing the ACP subprocess.
Ensure that if an ACP command override is provided, the child agent
correctly routes through CopilotACPClient.
Refs #2653
Adds MiniMax-AI/cli to the default taps list so the mmx-cli skill
is discoverable and installable out of the box via /skills browse
and /skills install. The skill definition lives upstream at
github.com/MiniMax-AI/cli/skill/SKILL.md, keeping updates decoupled.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace hardcoded 'fr' default with DEFAULT_LOCAL_STT_LANGUAGE ('en')
— removes locale leak, matches other providers
- Drop redundant default=True on is_truthy_value (dict .get already defaults)
- Update auto-detect comment to include 'xai' in the chain
- Fix docstring: 21 languages (match PR body + actual xAI API)
- Update test_sends_language_and_format to set HERMES_LOCAL_STT_LANGUAGE=fr
explicitly, since default is no longer 'fr'
All 18 xAI STT tests pass locally.
In _detect_target(), platform.system() returns "Android" on Termux,
not "Linux". Without this change tirith's auto-installer skips
Android even though the Linux GNU binaries are ABI-compatible.
When an MCP server config has ssl_verify: false (e.g. local dev with
a self-signed cert), the setting was read from config.yaml but never
passed to the httpx client, causing CERTIFICATE_VERIFY_FAILED errors
and silent connection failures.
Fix: read ssl_verify from config and pass it as the 'verify' kwarg to
both code paths:
- New API (mcp >= 1.24.0): httpx.AsyncClient(verify=ssl_verify)
- Legacy API (mcp < 1.24.0): streamablehttp_client(..., verify=ssl_verify)
Fixes local dev setups using ServBay, LocalWP, MAMP, or any stack with
a self-signed TLS certificate.
On Windows, Path.open() defaults to the system ANSI code page (cp1252).
If the .env file contains UTF-8 characters, decoding fails with
'gbk codec can't decode byte 0x94'. Specify encoding='utf-8'
explicitly to ensure consistent behavior across platforms.
The Docker terminal backend runs containers with `--cap-drop ALL`
and re-adds only DAC_OVERRIDE, CHOWN, FOWNER. Since commit fee0e0d3
("run as non-root user, use virtualenv") the image entrypoint drops
from root to the `hermes` user via `gosu`, which requires CAP_SETUID
and CAP_SETGID. Without them every sandbox container exits
immediately with:
Dropping root privileges
error: failed switching to 'hermes': operation not permitted
Breaking every terminal/file tool invocation in `terminal.backend: docker`
mode.
Fix: add SETUID and SETGID to the cap-add list. The `no-new-privileges`
security-opt is kept, so gosu still cannot escalate back to root after
the one-way drop — the hardening posture is preserved.
Reproduction
------------
With any image whose ENTRYPOINT calls `gosu <user>`, the container
exits immediately under the pre-fix cap set. Post-fix, the drop
succeeds and the container proceeds normally.
docker run --rm \
--cap-drop ALL \
--cap-add DAC_OVERRIDE --cap-add CHOWN --cap-add FOWNER \
--security-opt no-new-privileges \
--entrypoint /usr/local/bin/gosu \
hermes-claude:latest hermes id
# -> error: failed switching to 'hermes': operation not permitted
# Same command with SETUID+SETGID added:
# -> uid=10000(hermes) gid=10000(hermes) groups=10000(hermes)
Tests
-----
Added `test_security_args_include_setuid_setgid_for_gosu_drop` that
asserts both caps are present and the overall hardening posture
(cap-drop ALL + no-new-privileges) is preserved.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
LocalEnvironment._run_bash() spawned subprocess.Popen without a cwd
argument, so init_session()'s pwd -P ran in the gateway process's
startup directory and overwrote self.cwd. Pass cwd=self.cwd so the
initial snapshot captures the user-configured working directory.
Tested:
- pytest tests/ -q (255 env-related tests passed)
- Full suite: 13,537 passed; 70 pre-existing failures unrelated to local env
The container_config builder in terminal_tool.py was missing
docker_forward_env and docker_env keys, causing config.yaml's
docker_forward_env setting to be silently ignored. Environment
variables listed in docker_forward_env were never injected into
Docker containers.
This fix adds both keys to the container_config dict so they are
properly passed to _create_environment().
Follow-up on helix4u's PR #14211:
- Flip default to true: narrowing toolsets=['web','browser'] expresses
'I want these extras', not 'silently strip MCP'. Parent MCP tools
(registered at runtime) should survive narrowing by default.
- Drop _config_version bump (22->23); additive nested key under
delegation.* is handled by _deep_merge, no migration needed.
- Update tests to reflect new default behavior.
browser_cdp_tool.py registers before browser_tool.py (alphabetical
import order), so its stricter check_fn (requires CDP endpoint) becomes
the toolset-level check for all 11 browser tools. This causes
'hermes doctor' to report the entire browser toolset as unavailable
even when agent-browser is correctly installed.
Move browser_cdp to toolset='browser-cdp' so it is evaluated
independently. browser_navigate et al. only need agent-browser;
browser_cdp additionally requires a reachable CDP endpoint.
Version managers like frum (Ruby), rvm, nvm, and others commonly alias
cd to a wrapper function that runs additional logic after directory
changes. When Hermes captures the shell environment into a session
snapshot, these aliases are preserved. If the wrapper function fails
in the subprocess context (e.g. frum not on PATH), every cd fails,
causing all terminal commands to exit with code 126.
Using builtin cd bypasses any aliases or functions, ensuring the
directory change always uses the real bash builtin regardless of
what version managers are installed.
Make Tavily client respect a TAVILY_BASE_URL environment variable,
defaulting to https://api.tavily.com for backward compatibility.
Consistent with FIRECRAWL_API_URL pattern already used in this module.
The code execution sandbox creates a Unix domain socket in /tmp with
default permissions, allowing any local user to connect and execute
tool calls. Restrict to 0o600 after bind.
Closes#6230
Upgrades agent-browser from 0.13.0 to 0.26.0, picking up 13 releases of
daemon reliability fixes:
- Daemon hang on Linux from waitpid(-1) race in SIGCHLD handler (#1098)
- Chrome killed after ~10s idle due to PR_SET_PDEATHSIG thread tracking (#1157)
- Orphaned Chrome processes via process-group kill on shutdown (#1137)
- Stale daemon after upgrade via .version sidecar and auto-restart (#1134)
- Idle timeout not firing (sleep future recreated each loop) (#1110)
- Navigation hanging on lifecycle events that never fire (#1059, #1092)
- CDP attach hang on Chrome 144+ (#1133)
- Windows daemon TCP bind with Hyper-V port conflicts (#1041)
- Shadow DOM traversal in accessibility tree snapshots
- doctor command for user self-diagnosis
Also wires AGENT_BROWSER_IDLE_TIMEOUT_MS into the browser subprocess
environment so the daemon self-terminates after our configured inactivity
timeout (default 300s). This is the daemon-side counterpart to the
Python-side inactivity reaper — the daemon kills itself and its Chrome
children when no commands arrive, preventing orphan accumulation even
when the Python process dies without running atexit handlers.
Addresses #7343 (daemon socket hangs, shadow DOM) and #13793 (orphan
accumulation from force-killed sessions).
Adds security.allow_private_urls / HERMES_ALLOW_PRIVATE_URLS toggle so
users on OpenWrt routers, TUN-mode proxies (Clash/Mihomo/Sing-box),
corporate split-tunnel VPNs, and Tailscale networks — where DNS resolves
public domains to 198.18.0.0/15 or 100.64.0.0/10 — can use web_extract,
browser, vision URL fetching, and gateway media downloads.
Single toggle in tools/url_safety.py; all 23 is_safe_url() call sites
inherit automatically. Cached for process lifetime.
Cloud metadata endpoints stay ALWAYS blocked regardless of the toggle:
169.254.169.254 (AWS/GCP/Azure/DO/Oracle), 169.254.170.2 (AWS ECS task
IAM creds), 169.254.169.253 (Azure IMDS wire server), 100.100.100.200
(Alibaba), fd00:ec2::254 (AWS IPv6), the entire 169.254.0.0/16
link-local range, and the metadata.google.internal / metadata.goog
hostnames (checked pre-DNS so they can't be bypassed on networks where
those names resolve to local IPs).
Supersedes #3779 (narrower HERMES_ALLOW_RFC2544 for the same class of
users).
Co-authored-by: kshitijk4poor <82637225+kshitijk4poor@users.noreply.github.com>
- delegate_task: use shared tool_error() for the paused-spawn early return
so the error envelope matches the rest of the tool.
- Disk snapshot label: treat orphaned nodes (parentId missing from the
snapshot) as top-level, matching buildSubagentTree / summarizeLabel.
Four real issues Copilot flagged:
1. delegate_tool: `_build_child_agent` never passed `toolsets` to the
progress callback, so the event payload's `toolsets` field (wired
through every layer) was always empty and the overlay's toolsets
row never populated. Thread `child_toolsets` through.
2. event handler: the race-protection on subagent.spawn_requested /
subagent.start only preserved `completed`, so a late-arriving queued
event could clobber `failed` / `interrupted` too. Preserve any
terminal status (`completed | failed | interrupted`).
3. SpawnHud: comment claimed concurrency was approximated by "widest
level in the tree" but code used `totals.activeCount` (total across
all parents). `max_concurrent_children` is a per-parent cap, so
activeCount over-warns for multi-orchestrator runs. Switch to
`max(widthByDepth(tree))`; the label now reads `⚡W/cap+extra` where
W is the widest level (drives the ratio) and `+extra` is the rest.
4. spawn_tree.list: comment said "peek header without parsing full list"
but the code json.loads()'d every snapshot. Adds a per-session
`_index.jsonl` sidecar written on save; list() reads only the index
(with a full-scan fallback for pre-index sessions). O(1) per
snapshot now vs O(file-size).
Adds a live + post-hoc audit surface for recursive delegate_task fan-out.
None of cc/oc/oclaw tackle nested subagent trees inside an Ink overlay;
this ships a view-switched dashboard that handles arbitrary depth + width.
Python
- delegate_tool: every subagent event now carries subagent_id, parent_id,
depth, model, tool_count; subagent.complete also ships input/output/
reasoning tokens, cost, api_calls, files_read/files_written, and a
tail of tool-call outputs
- delegate_tool: new subagent.spawn_requested event + _active_subagents
registry so the overlay can kill a branch by id and pause new spawns
- tui_gateway: new RPCs delegation.status, delegation.pause,
subagent.interrupt, spawn_tree.save/list/load (disk under
\$HERMES_HOME/spawn-trees/<session>/<ts>.json)
TUI
- /agents overlay: full-width list mode (gantt strip + row picker) and
Enter-to-drill full-width scrollable detail mode; inverse+amber
selection, heat-coloured branch markers, wall-clock gantt with tick
ruler, per-branch rollups
- Detail pane: collapsible accordions (Budget, Files, Tool calls, Output,
Progress, Summary); open-state persists across agents + mode switches
via a shared atom
- /replay [N|last|list|load <path>] for in-memory + disk history;
/replay-diff <a> <b> for side-by-side tree comparison
- Status-bar SpawnHud warns as depth/concurrency approaches caps;
overlay auto-follows the just-finished turn onto history[1]
- Theme: bump DARK dim #B8860B → #CC9B1F for readable secondary text
globally; keep LIGHT untouched
Tests: +29 new subagentTree unit tests; 215/215 passing.
* feat(plugins): pluggable image_gen backends + OpenAI provider
Adds a ImageGenProvider ABC so image generation backends register as
bundled plugins under `plugins/image_gen/<name>/`. The plugin scanner
gains three primitives to make this work generically:
- `kind:` manifest field (`standalone` | `backend` | `exclusive`).
Bundled `kind: backend` plugins auto-load — no `plugins.enabled`
incantation. User-installed backends stay opt-in.
- Path-derived keys: `plugins/image_gen/openai/` gets key
`image_gen/openai`, so a future `tts/openai` cannot collide.
- Depth-2 recursion into category namespaces (parent dirs without a
`plugin.yaml` of their own).
Includes `OpenAIImageGenProvider` as the first consumer (gpt-image-1.5
default, plus gpt-image-1, gpt-image-1-mini, DALL-E 3/2). Base64
responses save to `$HERMES_HOME/cache/images/`; URL responses pass
through.
FAL stays in-tree for this PR — a follow-up ports it into
`plugins/image_gen/fal/` so the in-tree `image_generation_tool.py`
slims down. The dispatch shim in `_handle_image_generate` only fires
when `image_gen.provider` is explicitly set to a non-FAL value, so
existing FAL setups are untouched.
- 41 unit tests (scanner recursion, kind parsing, gate logic,
registry, OpenAI payload shapes)
- E2E smoke verified: bundled plugin autoloads, registers, and
`_handle_image_generate` routes to OpenAI when configured
* fix(image_gen/openai): don't send response_format to gpt-image-*
The live API rejects it: 'Unknown parameter: response_format'
(verified 2026-04-21 with gpt-image-1.5). gpt-image-* models return
b64_json unconditionally, so the parameter was both unnecessary and
actively broken.
* feat(image_gen/openai): gpt-image-2 only, drop legacy catalog
gpt-image-2 is the latest/best OpenAI image model (released 2026-04-21)
and there's no reason to expose the older gpt-image-1.5 / gpt-image-1 /
dall-e-3 / dall-e-2 alongside it — slower, lower quality, or awkward
(dall-e-2 squares only). Trim the catalog down to a single model.
Live-verified end-to-end: landscape 1536x1024 render of a Moog-style
synth matches prompt exactly, 2.4MB PNG saved to cache.
* feat(image_gen/openai): expose gpt-image-2 as three quality tiers
Users pick speed/fidelity via the normal model picker instead of a
hidden quality knob. All three tier IDs resolve to the single underlying
gpt-image-2 API model with a different quality parameter:
gpt-image-2-low ~15s fast iteration
gpt-image-2-medium ~40s default
gpt-image-2-high ~2min highest fidelity
Live-measured on OpenAI's API today: 15.4s / 40.8s / 116.9s for the
same 1024x1024 prompt.
Config:
image_gen.openai.model: gpt-image-2-high
# or
image_gen.model: gpt-image-2-low
# or env var for scripts/tests
OPENAI_IMAGE_MODEL=gpt-image-2-medium
Live-verified end-to-end with the low tier: 18.8s landscape render of a
golden retriever in wildflowers, vision-confirmed exact match.
* feat(tools_config): plugin image_gen providers inject themselves into picker
'hermes tools' → Image Generation now shows plugin-registered backends
alongside Nous Subscription and FAL.ai without tools_config.py needing
to know about them. OpenAI appears as a third option today; future
backends appear automatically as they're added.
Mechanism:
- ImageGenProvider gains an optional get_setup_schema() hook
(name, badge, tag, env_vars). Default derived from display_name.
- tools_config._plugin_image_gen_providers() pulls the schemas from
every registered non-FAL plugin provider.
- _visible_providers() appends those rows when rendering the Image
Generation category.
- _configure_provider() handles the new image_gen_plugin_name marker:
writes image_gen.provider and routes to the plugin's list_models()
catalog for the model picker.
- _toolset_needs_configuration_prompt('image_gen') stops demanding a
FAL key when any plugin provider reports is_available().
FAL is skipped in the plugin path because it already has hardcoded
TOOL_CATEGORIES rows — when it gets ported to a plugin in a follow-up
PR the hardcoded rows go away and it surfaces through the same path
as OpenAI.
Verified live: picker shows Nous Subscription / FAL.ai / OpenAI.
Picking OpenAI prompts for OPENAI_API_KEY, then shows the
gpt-image-2-low/medium/high model picker sourced from the plugin.
397 tests pass across plugins/, tools_config, registry, and picker.
* fix(image_gen): close final gaps for plugin-backend parity with FAL
Two small places that still hardcoded FAL:
- hermes_cli/setup.py status line: an OpenAI-only setup showed
'Image Generation: missing FAL_KEY'. Now probes plugin providers
and reports '(OpenAI)' when one is_available() — or falls back to
'missing FAL_KEY or OPENAI_API_KEY' if nothing is configured.
- image_generate tool schema description: said 'using FAL.ai, default
FLUX 2 Klein 9B'. Rewrote provider-neutral — 'backend and model are
user-configured' — and notes the 'image' field can be a URL or an
absolute path, which the gateway delivers either way via
extract_local_files().
- Wrap child.run_conversation() in a ThreadPoolExecutor with configurable
timeout (delegation.child_timeout_seconds, default 300s) to prevent
indefinite blocking when a subagent's API call or tool HTTP request hangs.
- Add heartbeat stale detection: if a child's api_call_count doesn't
advance for 5 consecutive heartbeat cycles (~2.5 min), stop touching
the parent's activity timestamp so the gateway inactivity timeout
can fire as a last resort.
- Add 'timeout' as a new exit_reason/status alongside the existing
completed/max_iterations/interrupted states.
- Use shutdown(wait=False) on the timeout executor to avoid the
ThreadPoolExecutor.__exit__ deadlock when a child is stuck on
blocking I/O.
Closes#13768
A single global MAX_TEXT_LENGTH = 4000 truncated every TTS provider at
4000 chars, causing long inputs to be silently chopped even though the
underlying APIs allow much more:
- OpenAI: 4096
- xAI: 15000
- MiniMax: 10000
- ElevenLabs: 5000 / 10000 / 30000 / 40000 (model-aware)
- Gemini: ~5000
- Edge: ~5000
The schema description also told the model 'Keep under 4000 characters',
which encouraged the agent to self-chunk long briefs into multiple TTS
calls (producing 3 separate audio files instead of one).
New behavior:
- PROVIDER_MAX_TEXT_LENGTH table + ELEVENLABS_MODEL_MAX_TEXT_LENGTH
encode the documented per-provider limits.
- _resolve_max_text_length(provider, cfg) resolves:
1. tts.<provider>.max_text_length user override
2. ElevenLabs model_id lookup
3. provider default
4. 4000 fallback
- text_to_speech_tool() and stream_tts_to_speaker() both call the
resolver; old MAX_TEXT_LENGTH alias kept for back-compat.
- Schema description no longer hardcodes 4000.
Tests: 27 new unit + E2E tests; all 53 existing TTS tests and 253
voice-command/voice-cli tests still pass.
* feat(models): hide OpenRouter models that don't advertise tool support
Port from Kilo-Org/kilocode#9068.
hermes-agent is tool-calling-first — every provider path assumes the
model can invoke tools. Models whose OpenRouter supported_parameters
doesn't include 'tools' (e.g. image-only or completion-only models)
cannot be driven by the agent loop and fail at the first tool call.
Filter them out of fetch_openrouter_models() so they never appear in
the model picker (`hermes model`, setup wizard, /model slash command).
Permissive when the field is missing — OpenRouter-compatible gateways
(Nous Portal, private mirrors, older snapshots) don't always populate
supported_parameters. Treat missing as 'unknown → allow' rather than
silently emptying the picker on those gateways. Only hide models
whose supported_parameters is an explicit list that omits tools.
Tests cover: tools present → kept, tools absent → dropped, field
missing → kept, malformed non-list → kept, non-dict item → kept,
empty list → dropped.
* feat(delegate): cross-agent file state coordination for concurrent subagents
Prevents mangled edits when concurrent subagents touch the same file
(same process, same filesystem — the mangle scenario from #11215).
Three layers, all opt-out via HERMES_DISABLE_FILE_STATE_GUARD=1:
1. FileStateRegistry (tools/file_state.py) — process-wide singleton
tracking per-agent read stamps and the last writer globally.
check_stale() names the sibling subagent in the warning when a
non-owning agent wrote after this agent's last read.
2. Per-path threading.Lock wrapped around the read-modify-write
region in write_file_tool and patch_tool. Concurrent siblings on
the same path serialize; different paths stay fully parallel.
V4A multi-file patches lock in sorted path order (deadlock-free).
3. Delegate-completion reminder in tools/delegate_tool.py: after a
subagent returns, writes_since(parent, child_start, parent_reads)
appends '[NOTE: subagent modified files the parent previously
read — re-read before editing: ...]' to entry.summary when the
child touched anything the parent had already seen.
Complements (does not replace) the existing path-overlap check in
run_agent._should_parallelize_tool_batch — batch check prevents
same-file parallel dispatch within one agent's turn (cheap prevention,
zero API cost), registry catches cross-subagent and cross-turn
staleness at write time (detection).
Behavior is warning-only, not hard-failing — matches existing project
style. Errors surface naturally: sibling writes often invalidate the
old_string in patch operations, which already errors cleanly.
Tests: tests/tools/test_file_state_registry.py — 16 tests covering
registry state transitions, per-path locking, per-path-not-global
locking, writes_since filtering, kill switch, and end-to-end
integration through the real read_file/write_file/patch handlers.
Adds role='leaf'|'orchestrator' to delegate_task. With max_spawn_depth>=2,
an orchestrator child retains the 'delegation' toolset and can spawn its
own workers; leaf children cannot delegate further (identical to today).
Default posture is flat — max_spawn_depth=1 means a depth-0 parent's
children land at the depth-1 floor and orchestrator role silently
degrades to leaf. Users opt into nested delegation by raising
max_spawn_depth to 2 or 3 in config.yaml.
Also threads acp_command/acp_args through the main agent loop's delegate
dispatch (previously silently dropped in the schema) via a new
_dispatch_delegate_task helper, and adds a DelegateEvent enum with
legacy-string back-compat for gateway/ACP/CLI progress consumers.
Config (hermes_cli/config.py defaults):
delegation.max_concurrent_children: 3 # floor-only, no upper cap
delegation.max_spawn_depth: 1 # 1=flat (default), 2-3 unlock nested
delegation.orchestrator_enabled: true # global kill switch
Salvaged from @pefontana's PR #11215. Overrides vs. the original PR:
concurrency stays at 3 (PR bumped to 5 + cap 8 — we keep the floor only,
no hard ceiling); max_spawn_depth defaults to 1 (PR defaulted to 2 which
silently enabled one level of orchestration for every user).
Co-authored-by: pefontana <fontana.pedro93@gmail.com>
Adds OpenAI's new GPT Image 2 model via FAL.ai, selectable through
`hermes tools` → Image Generation. SOTA text rendering (including CJK)
and world-aware photorealism.
- FAL_MODELS entry with image_size_preset style
- 4:3 presets on all aspect ratios — 16:9 (1024x576) falls below
GPT-Image-2's 655,360 min-pixel floor and would be rejected
- quality pinned to medium (same rule as gpt-image-1.5) for
predictable Nous Portal billing
- BYOK (openai_api_key) deliberately omitted from supports so all
users stay on shared FAL billing
- 6 new tests covering preset mapping, quality pinning, and
supports-whitelist integrity
- Docs table + aspect-ratio map updated
Live-tested end-to-end: 39.9s cold request, clean 1024x768 PNG
Two related ACP approval issues:
GHSA-96vc-wcxf-jjff — ACP's _run_agent never set HERMES_INTERACTIVE
(or any other flag recognized by tools.approval), so check_all_command_guards
took the non-interactive auto-approve path and never consulted the
ACP-supplied approval callback (conn.request_permission). Dangerous
commands executed in ACP sessions without operator approval despite
the callback being installed. Fix: set HERMES_INTERACTIVE=1 around
the agent run so check_all_command_guards routes through
prompt_dangerous_approval(approval_callback=...) — the correct shape
for ACP's per-session request_permission call. HERMES_EXEC_ASK would
have routed through the gateway-queue path instead, which requires a
notify_cb registered in _gateway_notify_cbs (not applicable to ACP).
GHSA-qg5c-hvr5-hjgr — _approval_callback and _sudo_password_callback
were module-level globals in terminal_tool. Concurrent ACP sessions
running in ThreadPoolExecutor threads each installed their own callback
into the same slot, racing. Fix: store both callbacks in threading.local()
so each thread has its own slot. CLI mode (single thread) is unaffected;
gateway mode uses a separate queue-based approval path and was never
touched.
set_approval_callback is now called INSIDE _run_agent (the executor
thread) rather than before dispatching — so the TLS write lands on the
correct thread.
Tests: 5 new in tests/acp/test_approval_isolation.py covering
thread-local isolation of both callbacks and the HERMES_INTERACTIVE
callback routing. Existing tests/acp/ (159 tests) and tests/tools/
approval-related tests continue to pass.
Fixes GHSA-96vc-wcxf-jjff
Fixes GHSA-qg5c-hvr5-hjgr
A skill declaring `required_environment_variables: [ANTHROPIC_TOKEN]` in
its SKILL.md frontmatter silently bypassed the `execute_code` sandbox's
credential-scrubbing guarantee. `register_env_passthrough` had no
blocklist, so any name a skill chose flipped `is_env_passthrough(name) =>
True`, which shortcircuits the sandbox's secret filter.
Fix: reject registration when the name appears in
`_HERMES_PROVIDER_ENV_BLOCKLIST` (the canonical list of Hermes-managed
credentials — provider keys, gateway tokens, etc.). Log a warning naming
GHSA-rhgp-j443-p4rf so operators see the rejection in logs.
Non-Hermes third-party API keys (TENOR_API_KEY for gif-search,
NOTION_TOKEN for notion skills, etc.) remain legitimately registerable —
they were never in the sandbox scrub list in the first place.
Tests: 16 -> 17 passing. Two old tests that documented the bypass
(`test_passthrough_allows_blocklisted_var`, `test_make_run_env_passthrough`)
are rewritten to assert the new fail-closed behavior. New
`test_non_hermes_api_key_still_registerable` locks in that legitimate
third-party keys are unaffected.
Reported in GHSA-rhgp-j443-p4rf by @q1uf3ng. Hardening; not CVE-worthy
on its own per the decision matrix (attacker must already have operator
consent to install a malicious skill).
Fixes#13027
Previously, `_is_skill_disabled()` only checked the explicit `platform`
argument and `os.getenv('HERMES_PLATFORM')`, missing the gateway session
context (`HERMES_SESSION_PLATFORM`). This caused `skill_view()` to expose
skills that were platform-disabled for the active gateway session.
Add `_get_session_platform()` helper that resolves the platform from
`gateway.session_context.get_session_env`, mirroring the logic in
`agent.skill_utils.get_disabled_skill_names()`.
Now the platform resolution follows the same precedence as skill_utils:
1. Explicit `platform` argument
2. `HERMES_PLATFORM` environment variable
3. `HERMES_SESSION_PLATFORM` from gateway session context
Previously the breaker was only cleared when the post-reconnect retry
call itself succeeded (via _reset_server_error at the end of the try
block). If OAuth recovery succeeded but the retry call happened to
fail for a different reason, control fell through to the
needs_reauth path which called _bump_server_error — adding to an
already-tripped count instead of the fresh count the reconnect
justified. With fix#1 in place this would still self-heal on the
next cooldown, but we should not pay a 60s stall when we already
have positive evidence the server is viable.
Move _reset_server_error(server_name) up to immediately after the
reconnect-and-ready-wait block, before the retry_call. The
subsequent retry still goes through _bump_server_error on failure,
so a genuinely broken server re-trips the breaker as normal — but
the retry starts from a clean count (1 after a failure), not a
stale one.
The MCP circuit breaker previously had no path back to the closed
state: once _server_error_counts[srv] reached _CIRCUIT_BREAKER_THRESHOLD
the gate short-circuited every subsequent call, so the only reset
path (on successful call) was unreachable. A single transient
3-failure blip (bad network, server restart, expired token) permanently
disabled every tool on that MCP server for the rest of the agent
session.
Introduce a classic closed/open/half-open state machine:
- Track a per-server breaker-open timestamp in _server_breaker_opened_at
alongside the existing failure count.
- Add _CIRCUIT_BREAKER_COOLDOWN_SEC (60s). Once the count reaches
threshold, calls short-circuit for the cooldown window.
- After the cooldown elapses, the *next* call falls through as a
half-open probe that actually hits the session. Success resets the
breaker via _reset_server_error; failure re-bumps the count via
_bump_server_error, which re-stamps the open timestamp and re-arms
the cooldown.
The error message now includes the live failure count and an
"Auto-retry available in ~Ns" hint so the model knows the breaker
will self-heal rather than giving up on the tool for the whole
session.
Covers tests 1 (half-opens after cooldown) and 2 (reopens on probe
failure); test 3 (cleared on reconnect) still fails pending fix#2.
Follow-up to PR #2504. The original fix covered the two direct FAL_KEY
checks in image_generation_tool but left four other call sites intact,
including the managed-gateway gate where a whitespace-only FAL_KEY
falsely claimed 'user has direct FAL' and *skipped* the Nous managed
gateway fallback entirely.
Introduce fal_key_is_configured() in tools/tool_backend_helpers.py as a
single source of truth (consults os.environ, falls back to .env for
CLI-setup paths) and route every FAL_KEY presence check through it:
- tools/image_generation_tool.py : _resolve_managed_fal_gateway,
image_generate_tool's upfront check, check_fal_api_key
- hermes_cli/nous_subscription.py : direct_fal detection, selected
toolset gating, tools_ready map
- hermes_cli/tools_config.py : image_gen needs-setup check
Verified by extending tests/tools/test_image_generation_env.py and by
E2E exercising whitespace + managed-gateway composition directly.
Treat whitespace-only FAL_KEY the same as unset so users who export
FAL_KEY=" " (or CI that leaves a blank token) get the expected
'not set' error path instead of a confusing downstream fal_client
failure.
Applied to the two direct FAL_KEY checks in image_generation_tool.py:
image_generate_tool's upfront credential check and check_fal_api_key().
Both keep the existing managed-gateway fallback intact.
Adapted the original whitespace/valid tests to pin the managed gateway
to None so the whitespace assertion exercises the direct-key path
rather than silently relying on gateway absence.
Follow-ups on top of @teyrebaz33's cherry-picked commit:
1. New shared helper format_no_match_hint() in fuzzy_match.py with a
startswith('Could not find') gate so the snippet only appends to
genuine no-match errors — not to 'Found N matches' (ambiguous),
'Escape-drift detected', or 'identical strings' errors, which would
all mislead the model.
2. file_tools.patch_tool suppresses the legacy generic '[Hint: old_string
not found...]' string when the rich 'Did you mean?' snippet is
already attached — no more double-hint.
3. Wire the same helper into patch_parser.py (V4A patch mode, both
_validate_operations and _apply_update) and skill_manager_tool.py so
all three fuzzy callers surface the hint consistently.
Tests: 7 new gating tests in TestFormatNoMatchHint cover every error
class (ambiguous, drift, identical, non-zero match count, None error,
no similar content, happy path). 34/34 test_fuzzy_match, 96/96
test_file_tools + test_patch_parser + test_skill_manager_tool pass.
E2E verified across all four scenarios: no-match-with-similar,
no-match-no-similar, ambiguous, success. V4A mode confirmed
end-to-end with a non-matching hunk.
When patch_replace() cannot find old_string in a file, the error message
now includes the closest matching lines from the file with line numbers
and context. This helps the LLM self-correct without a separate read_file
call.
Implements Phase 1 of #536: enhanced patch error feedback with no
architectural changes.
- tools/fuzzy_match.py: new find_closest_lines() using SequenceMatcher
- tools/file_operations.py: attach closest-lines hint to patch errors
- tests/tools/test_fuzzy_match.py: 5 new tests for find_closest_lines
Builds on @AxDSan's PR #2109 to finish the KittenTTS wiring so the
provider behaves like every other TTS backend end to end.
- tools/tts_tool.py: `_check_kittentts_available()` helper and wire
into `check_tts_requirements()`; extend Opus-conversion list to
include kittentts (WAV → Opus for Telegram voice bubbles); point the
missing-package error at `hermes setup tts`.
- hermes_cli/tools_config.py: add KittenTTS entry to the "Text-to-Speech"
toolset picker, with a `kittentts` post_setup hook that auto-installs
the wheel + soundfile via pip.
- hermes_cli/setup.py: `_install_kittentts_deps()`, new choice + install
flow in `_setup_tts_provider()`, provider_labels entry, and status row
in the `hermes setup` summary.
- website/docs/user-guide/features/tts.md: add KittenTTS to the provider
table, config example, ffmpeg note, and the zero-config voice-bubble tip.
- tests/tools/test_tts_kittentts.py: 10 unit tests covering generation,
model caching, config passthrough, ffmpeg conversion, availability
detection, and the missing-package dispatcher branch.
E2E verified against the real `kittentts` wheel:
- WAV direct output (pcm_s16le, 24kHz mono)
- MP3 conversion via ffmpeg (from WAV)
- Telegram flow (provider in Opus-conversion list) produces
`codec_name=opus`, 48kHz mono, `voice_compatible=True`, and the
`[[audio_as_voice]]` marker
- check_tts_requirements() returns True when kittentts is installed
Add support for KittenTTS - a lightweight, local TTS engine with models
ranging from 25-80MB that runs on CPU without requiring a GPU or API key.
Features:
- Support for 8 built-in voices (Jasper, Bella, Luna, etc.)
- Configurable model size (nano 25MB, micro 41MB, mini 80MB)
- Adjustable speech speed
- Model caching for performance
- Automatic WAV to Opus conversion for Telegram voice messages
Configuration example (config.yaml):
tts:
provider: kittentts
kittentts:
model: KittenML/kitten-tts-nano-0.8-int8
voice: Jasper
speed: 1.0
clean_text: true
Installation:
pip install https://github.com/KittenML/KittenTTS/releases/download/0.8.1/kittentts-0.8.1-py3-none-any.whl
Full AST-based scan of all .py files to find every case where a module
or name is imported locally inside a function body but is already
available at module level. This is the second pass — the first commit
handled the known cases from the lint report; this one catches
everything else.
Files changed (19):
cli.py — 16 removals: time as _time/_t/_tmod (×10),
re / re as _re (×2), os as _os, sys,
partial os from combo import,
from model_tools import get_tool_definitions
gateway/run.py — 8 removals: MessageEvent as _ME /
MessageType as _MT (×3), os as _os2,
MessageEvent+MessageType (×2), Platform,
BasePlatformAdapter as _BaseAdapter
run_agent.py — 6 removals: get_hermes_home as _ghh,
partial (contextlib, os as _os),
cleanup_vm, cleanup_browser,
set_interrupt as _sif (×2),
partial get_toolset_for_tool
hermes_cli/main.py — 4 removals: get_hermes_home, time as _time,
logging as _log, shutil
hermes_cli/config.py — 1 removal: get_hermes_home as _ghome
hermes_cli/runtime_provider.py
— 1 removal: load_config as _load_bedrock_config
hermes_cli/setup.py — 2 removals: importlib.util (×2)
hermes_cli/nous_subscription.py
— 1 removal: from hermes_cli.config import load_config
hermes_cli/tools_config.py
— 1 removal: from hermes_cli.config import load_config, save_config
cron/scheduler.py — 3 removals: concurrent.futures, json as _json,
from hermes_cli.config import load_config
batch_runner.py — 1 removal: list_distributions as get_all_dists
(kept print_distribution_info, not at top level)
tools/send_message_tool.py
— 2 removals: import os (×2)
tools/skills_tool.py — 1 removal: logging as _logging
tools/browser_camofox.py
— 1 removal: from hermes_cli.config import load_config
tools/image_generation_tool.py
— 1 removal: import fal_client
environments/tool_context.py
— 1 removal: concurrent.futures
gateway/platforms/bluebubbles.py
— 1 removal: httpx as _httpx
gateway/platforms/whatsapp.py
— 1 removal: import asyncio
tui_gateway/server.py — 2 removals: from datetime import datetime,
import time
All alias references (_time, _t, _tmod, _re, _os, _os2, _json, _ghh,
_ghome, _sif, _ME, _MT, _BaseAdapter, _load_bedrock_config, _httpx,
_logging, _log, get_all_dists) updated to use the top-level names.
Sweep ~74 redundant local imports across 21 files where the same module
was already imported at the top level. Also includes type fixes and lint
cleanups on the same branch.
* feat(skills): inject absolute skill dir and expand ${HERMES_SKILL_DIR} templates
When a skill loads, the activation message now exposes the absolute
skill directory and substitutes ${HERMES_SKILL_DIR} /
${HERMES_SESSION_ID} tokens in the SKILL.md body, so skills with
bundled scripts can instruct the agent to run them by absolute path
without an extra skill_view round-trip.
Also adds opt-in inline-shell expansion: !`cmd` snippets in SKILL.md
are pre-executed (with the skill directory as CWD) and their stdout is
inlined into the message before the agent reads it. Off by default —
enable via skills.inline_shell in config.yaml — because any snippet
runs on the host without approval.
Changes:
- agent/skill_commands.py: template substitution, inline-shell
expansion, absolute skill-dir header, supporting-files list now
shows both relative and absolute forms.
- hermes_cli/config.py: new skills.template_vars,
skills.inline_shell, skills.inline_shell_timeout knobs.
- tests/agent/test_skill_commands.py: coverage for header, both
template tokens (present and missing session id), template_vars
disable, inline-shell default-off, enabled, CWD, and timeout.
- website/docs/developer-guide/creating-skills.md: documents the
template tokens, the absolute-path header, and the opt-in inline
shell with its security caveat.
Validation: tests/agent/ 1591 passed (includes 9 new tests).
E2E: loaded a real skill in an isolated HERMES_HOME; confirmed
${HERMES_SKILL_DIR} resolves to the absolute path, ${HERMES_SESSION_ID}
resolves to the passed task_id, !`date` runs when opt-in is set, and
stays literal when it isn't.
* feat(terminal): source ~/.bashrc (and user-listed init files) into session snapshot
bash login shells don't source ~/.bashrc, so tools that install themselves
there — nvm, asdf, pyenv, cargo, custom PATH exports — stay invisible to
the environment snapshot Hermes builds once per session. Under systemd
or any context with a minimal parent env, that surfaces as
'node: command not found' in the terminal tool even though the binary
is reachable from every interactive shell on the machine.
Changes:
- tools/environments/local.py: before the login-shell snapshot bootstrap
runs, prepend guarded 'source <file>' lines for each resolved init
file. Missing files are skipped, each source is wrapped with a
'[ -r ... ] && . ... || true' guard so a broken rc can't abort the
bootstrap.
- hermes_cli/config.py: new terminal.shell_init_files (explicit list,
supports ~ and ${VAR}) and terminal.auto_source_bashrc (default on)
knobs. When shell_init_files is set it takes precedence; when it's
empty and auto_source_bashrc is on, ~/.bashrc gets auto-sourced.
- tests/tools/test_local_shell_init.py: 10 tests covering the resolver
(auto-bashrc, missing file, explicit override, ~/${VAR} expansion,
opt-out) and the prelude builder (quoting, guarded sourcing), plus
a real-LocalEnvironment snapshot test that confirms exports in the
init file land in subsequent commands' environment.
- website/docs/reference/faq.md: documents the fix in Troubleshooting,
including the zsh-user pattern of sourcing ~/.zshrc or nvm.sh
directly via shell_init_files.
Validation: 10/10 new tests pass; tests/tools/test_local_*.py 40/40
pass; tests/agent/ 1591/1591 pass; tests/hermes_cli/test_config.py
50/50 pass. E2E in an isolated HERMES_HOME: confirmed that a fake
~/.bashrc setting a marker var and PATH addition shows up in a real
LocalEnvironment().execute() call, that auto_source_bashrc=false
suppresses it, that an explicit shell_init_files entry wins over the
auto default, and that a missing bashrc is silently skipped.
Aslaaen's fix in the original PR covered _detect_api_mode_for_url and the
two openai/xai sites in run_agent.py. This finishes the sweep: the same
substring-match false-positive class (e.g. https://api.openai.com.evil/v1,
https://proxy/api.openai.com/v1, https://api.anthropic.com.example/v1)
existed in eight more call sites, and the hostname helper was duplicated
in two modules.
- utils: add shared base_url_hostname() (single source of truth).
- hermes_cli/runtime_provider, run_agent: drop local duplicates, import
from utils. Reuse the cached AIAgent._base_url_hostname attribute
everywhere it's already populated.
- agent/auxiliary_client: switch codex-wrap auto-detect, max_completion_tokens
gate (auxiliary_max_tokens_param), and custom-endpoint max_tokens kwarg
selection to hostname equality.
- run_agent: native-anthropic check in the Claude-style model branch
and in the AIAgent init provider-auto-detect branch.
- agent/model_metadata: Anthropic /v1/models context-length lookup.
- hermes_cli/providers.determine_api_mode: anthropic / openai URL
heuristics for custom/unknown providers (the /anthropic path-suffix
convention for third-party gateways is preserved).
- tools/delegate_tool: anthropic detection for delegated subagent
runtimes.
- hermes_cli/setup, hermes_cli/tools_config: setup-wizard vision-endpoint
native-OpenAI detection (paired with deduping the repeated check into
a single is_native_openai boolean per branch).
Tests:
- tests/test_base_url_hostname.py covers the helper directly
(path-containing-host, host-suffix, trailing dot, port, case).
- tests/hermes_cli/test_determine_api_mode_hostname.py adds the same
regression class for determine_api_mode, plus a test that the
/anthropic third-party gateway convention still wins.
Also: add asslaenn5@gmail.com → Aslaaen to scripts/release.py AUTHOR_MAP.
Users can declare shell scripts in config.yaml under a hooks: block that
fire on plugin-hook events (pre_tool_call, post_tool_call, pre_llm_call,
subagent_stop, etc). Scripts receive JSON on stdin, can return JSON on
stdout to block tool calls or inject context pre-LLM.
Key design:
- Registers closures on existing PluginManager._hooks dict — zero changes
to invoke_hook() call sites
- subprocess.run(shell=False) via shlex.split — no shell injection
- First-use consent per (event, command) pair, persisted to allowlist JSON
- Bypass via --accept-hooks, HERMES_ACCEPT_HOOKS=1, or hooks_auto_accept
- hermes hooks list/test/revoke/doctor CLI subcommands
- Adds subagent_stop hook event fired after delegate_task children exit
- Claude Code compatible response shapes accepted
Cherry-picked from PR #13143 by @pefontana.
Cherry-picked from PR #13159 by @cdanis.
Adds native media attachment delivery to Signal via signal-cli JSON-RPC
attachments param. Signal messages with media now follow the same
early-return pattern as Telegram/Discord/Matrix — attachments are sent
only with the last chunk to avoid duplicates.
Follow-up fixes on top of the original PR:
- Moved Signal into its own early-return block above the restriction
check (matches Telegram/Discord/Matrix pattern)
- Fixed media_files being sent on every chunk in the generic loop
- Restored restriction/warning guards to simple form (Signal exits early)
- Fixed non-hermetic test writing to /tmp instead of tmp_path
Adds a _resolve_path() helper that reads TERMINAL_CWD and uses it as
the base for relative path resolution. Applied to _check_sensitive_path,
read_file_tool, _update_read_timestamp, and _check_file_staleness.
Absolute paths and non-worktree sessions (no TERMINAL_CWD) are
unaffected — falls back to os.getcwd().
Fixes#12689.
Replaces the serial for-loop in tick() with ThreadPoolExecutor so all
jobs due in a single tick run concurrently. A slow job no longer blocks
others from executing, fixing silent job skipping (issue #9086).
Thread safety:
- Session/delivery env vars migrated from os.environ to ContextVars
(gateway/session_context.py) so parallel jobs can't clobber each
other's delivery targets. Each thread gets its own copied context.
- jobs.json read-modify-write cycles (advance_next_run, mark_job_run)
protected by threading.Lock to prevent concurrent save clobber.
- send_message_tool reads delivery vars via get_session_env() for
ContextVar-aware resolution with os.environ fallback.
Configuration:
- cron.max_parallel_jobs in config.yaml (null = unbounded, 1 = serial)
- HERMES_CRON_MAX_PARALLEL env var override
Based on PR #9169 by @VenomMoth1.
Fixes#9086
Cherry-picked from PR #2545 by @Mibayy.
The setup wizard could leave stt.model: "whisper-1" in config.yaml.
When using the local faster-whisper provider, this crashed with
"Invalid model size 'whisper-1'". Voice messages were silently ignored.
_normalize_local_model() now detects cloud-only names (whisper-1,
gpt-4o-transcribe, etc.) and maps them to the default local model
with a warning. Valid local sizes (tiny, base, small, medium, large-v3)
pass through unchanged.
- Renamed _normalize_local_command_model -> _normalize_local_model
(backward-compat wrapper preserved)
- 6 new tests including integration test
- Added lowercase AUTHOR_MAP alias for @Mibayy
Closes#2544
On macOS, Unix domain socket paths are capped at 104 bytes (sun_path).
SSH appends a 16-byte random suffix to the ControlPath when operating
in ControlMaster mode. With an IPv6 host embedded literally in the
filename and a deeply-nested macOS $TMPDIR like
/var/folders/XX/YYYYYYYYYYYY/T/, the full path reliably exceeds the
limit — every terminal/file-op tool call then fails immediately with
``unix_listener: path "…" too long for Unix domain socket``.
Swap the ``user@host:port.sock`` filename for a sha256-derived 16-char
hex digest. The digest is deterministic for a given (user, host, port)
triple, so ControlMaster reuse across reconnects is preserved, and the
full path fits comfortably under the limit even after SSH's random
suffix. Collision space is 2^64 — effectively unreachable for the
handful of concurrent connections any single Hermes process holds.
Regression tests cover: path length under realistic macOS $TMPDIR with
the IPv6 host from the issue report, determinism for reconnects, and
distinctness across different (user, host, port) triples.
Closes#11840
Follow-up to #12704. The SignalAdapter can resolve +E164 numbers to
UUIDs via listContacts, but _parse_target_ref() in the send_message
tool rejected '+' as non-digit and fell through to channel-name
resolution — which fails for contacts without a prior session entry.
Adds an E.164 branch in _parse_target_ref for phone-based platforms
(signal, sms, whatsapp) that preserves the leading '+' so downstream
adapters keep the format they expect. Non-phone platforms are
unaffected.
Reported by @qdrop17 on Discord after pulling #12704.
file_tools._get_file_ops() built a container_config dict for Docker/
Singularity/Modal/Daytona backends but omitted docker_mount_cwd_to_workspace
and docker_forward_env. Both are read by _create_environment() from
container_config, so file tools (read_file, write_file, patch, search)
silently ignored those config values when running in Docker.
Add the two missing keys to match the container_config already built by
terminal_tool.terminal_tool().
Fixes#2672.
The vision tool hardcoded temperature=0.1, ignoring the user's
config.yaml setting. This broke providers like Kimi/Moonshot that
require temperature=1 for vision models. Now reads temperature
from auxiliary.vision.temperature, falling back to 0.1.
bash parses `A && B &` with `&&` tighter than `&`, so it forks a subshell
for the compound and backgrounds the subshell. Inside the subshell, B
runs foreground, so the subshell waits for B. When B is a process that
doesn't naturally exit (`python3 -m http.server`, `yes > /dev/null`, a
long-running daemon), the subshell is stuck in `wait4` forever and leaks
as an orphan reparented to init.
Observed in production: agents running `cd X && python3 -m http.server
8000 &>/dev/null & sleep 1 && curl ...` as a "start a local server, then
verify it" one-liner. Outer bash exits cleanly; the subshell never does.
Across ~3 days of use, 8 unique stuck-terminal events and 7 leaked
bash+server pairs accumulated on the fleet, with some sessions appearing
hung from the user's perspective because the subshell's open stdout pipe
kept the terminal tool's drain thread blocked.
This is distinct from the `set +m` fix in 933fbd8f (which addressed
interactive-shell job-control waiting at exit). `set +m` doesn't help
here because `bash -c` is non-interactive and job control is already
off; the problem is the subshell's own internal wait for its foreground
B, not the outer shell's job-tracking.
The fix: walk the command shell-aware (respecting quotes, parens, brace
groups, `&>`/`>&` redirects), find `A && B &` / `A || B &` at depth 0
and rewrite the tail to `A && { B & }`. Brace groups don't fork a
subshell — they run in the current shell. `B &` inside the group is a
simple background (no subshell wait). The outer `&` is absorbed into
the group, so the compound no longer needs an explicit subshell.
`&&` error-propagation is preserved exactly: if A fails, `&&`
short-circuits and B never runs.
- Skips quoted strings, comment lines, and `(…)` subshells
- Handles `&>/dev/null`, `2>&1`, `>&2` without mistaking them for `&`
- Resets chain state at `;`, `|`, and newlines
- Tracks brace depth so already-rewritten output is idempotent
- Walks using the existing `_read_shell_token` tokenizer, matching the
pattern of `_rewrite_real_sudo_invocations`
Called once from `BaseEnvironment.execute` right after
`_prepare_command`, so it runs for every backend (local, ssh, docker,
modal, etc.) with no per-backend plumbing.
34 new tests covering rewrite cases, preservation cases, redirect
edge-cases, quoting/parens/backticks, idempotency, and empty/edge
inputs. End-to-end verified on a test VM: the exact vela-incident
command now returns in ~1.3s with no leaked bash, only the intentional
backgrounded server.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* [verified] fix(mcp-oauth): bridge httpx auth_flow bidirectional generator
HermesMCPOAuthProvider.async_auth_flow wrapped the SDK's auth_flow with
'async for item in super().async_auth_flow(request): yield item', which
discards httpx's .asend(response) values and resumes the inner generator
with None. This broke every OAuth MCP server on the first HTTP response
with 'NoneType' object has no attribute 'status_code' crashing at
mcp/client/auth/oauth2.py:505.
Replace with a manual bridge that forwards .asend() values into the
inner generator, preserving httpx's bidirectional auth_flow contract.
Add tests/tools/test_mcp_oauth_bidirectional.py with two regression
tests that drive the flow through real .asend() round-trips. These
catch the bug at the unit level; prior tests only exercised
_initialize() and disk-watching, never the full generator protocol.
Verified against BetterStack MCP:
Before: 'Connection failed (11564ms): NoneType...' after 3 retries
After: 'Connected (2416ms); Tools discovered: 83'
Regression from #11383.
* [verified] fix(mcp-oauth): seed token_expiry_time + pre-flight AS discovery on cold-load
PR #11383's consolidation fixed external-refresh reloading and 401 dedup
but left two latent bugs that surfaced on BetterStack and any other OAuth
MCP with a split-origin authorization server:
1. HermesTokenStorage persisted only a relative 'expires_in', which is
meaningless after a process restart. The MCP SDK's OAuthContext
does NOT seed token_expiry_time in _initialize, so is_token_valid()
returned True for any reloaded token regardless of age. Expired
tokens shipped to servers, and app-level auth failures (e.g.
BetterStack's 'No teams found. Please check your authentication.')
were invisible to the transport-layer 401 handler.
2. Even once preemptive refresh did fire, the SDK's _refresh_token
falls back to {server_url}/token when oauth_metadata isn't cached.
For providers whose AS is at a different origin (BetterStack:
mcp.betterstack.com for MCP, betterstack.com/oauth/token for the
token endpoint), that fallback 404s and drops into full browser
re-auth on every process restart.
Fix set:
- HermesTokenStorage.set_tokens persists an absolute wall-clock
expires_at alongside the SDK's OAuthToken JSON (time.time() + TTL
at write time).
- HermesTokenStorage.get_tokens reconstructs expires_in from
max(expires_at - now, 0), clamping expired tokens to zero TTL.
Legacy files without expires_at fall back to file-mtime as a
best-effort wall-clock proxy, self-healing on the next set_tokens.
- HermesMCPOAuthProvider._initialize calls super(), then
update_token_expiry on the reloaded tokens so token_expiry_time
reflects actual remaining TTL. If tokens are loaded but
oauth_metadata is missing, pre-flight PRM + ASM discovery runs
via httpx.AsyncClient using the MCP SDK's own URL builders and
response handlers (build_protected_resource_metadata_discovery_urls,
handle_auth_metadata_response, etc.) so the SDK sees the correct
token_endpoint before the first refresh attempt. Pre-flight is
skipped when there are no stored tokens to keep fresh-install
paths zero-cost.
Test coverage (tests/tools/test_mcp_oauth_cold_load_expiry.py):
- set_tokens persists absolute expires_at
- set_tokens skips expires_at when token has no expires_in
- get_tokens round-trips expires_at -> remaining expires_in
- expired tokens reload with expires_in=0
- legacy files without expires_at fall back to mtime proxy
- _initialize seeds token_expiry_time from stored tokens
- _initialize flags expired-on-disk tokens as is_token_valid=False
- _initialize pre-flights PRM + ASM discovery with mock transport
- _initialize skips pre-flight when no tokens are stored
Verified against BetterStack MCP:
hermes mcp test betterstack -> Connected (2508ms), 83 tools
mcp_betterstack_telemetry_list_teams_tool -> real team data, not
'No teams found. Please check your authentication.'
Reference: mcp-oauth-token-diagnosis skill, Fix A.
* chore: map hermes@noushq.ai to benbarclay in AUTHOR_MAP
Needed for CI attribution check on cherry-picked commits from PR #12025.
---------
Co-authored-by: Hermes Agent <hermes@noushq.ai>
Two hardening layers in the patch tool, triggered by a real silent failure
in the previous session:
(1) Post-write verification in patch_replace — after write_file succeeds,
re-read the file and confirm the bytes on disk match the intended write.
If not, return an error instead of the current success-with-diff. Catches
silent persistence failures from any cause (backend FS oddities, stdin
pipe truncation, concurrent task races, mount drift).
(2) Escape-drift guard in fuzzy_find_and_replace — when a non-exact
strategy matches and both old_string and new_string contain literal
\' or \" sequences but the matched file region does not, reject the
patch with a clear error pointing at the likely cause (tool-call
serialization adding a spurious backslash around apostrophes/quotes).
Exact matches bypass the guard, and legitimate edits that add or
preserve escape sequences in files that already have them still work.
Why: in a prior tool call, old_string was sent with \' where the file
has ' (tool-call transport drift). The fuzzy matcher's block_anchor
strategy matched anyway and produced a diff the tool reported as
successful — but the file was never modified on disk. The agent moved
on believing the edit landed when it hadn't.
Tests: added TestPatchReplacePostWriteVerification (3 cases) and
TestEscapeDriftGuard (6 cases). All pass, existing fuzzy match and
file_operations tests unaffected.
* feat: add Discord server introspection and management tool
Add a discord_server tool that gives the agent the ability to interact
with Discord servers when running on the Discord gateway. Uses Discord
REST API directly with the bot token — no dependency on the gateway
adapter's discord.py client.
The tool is only included in the hermes-discord toolset (zero cost for
users on other platforms) and gated on DISCORD_BOT_TOKEN via check_fn.
Actions (14):
- Introspection: list_guilds, server_info, list_channels, channel_info,
list_roles, member_info, search_members
- Messages: fetch_messages, list_pins, pin_message, unpin_message
- Management: create_thread, add_role, remove_role
This addresses a gap where users on Discord could not ask Hermes to
review server structure, channels, roles, or members — a task competing
agents (OpenClaw) handle out of the box.
Files changed:
- tools/discord_tool.py (new): Tool implementation + registration
- model_tools.py: Add to discovery list
- toolsets.py: Add to hermes-discord toolset only
- tests/tools/test_discord_tool.py (new): 43 tests covering all actions,
validation, error handling, registration, and toolset scoping
* feat(discord): intent-aware schema filtering + config allowlist + schema cleanup
- _detect_capabilities() hits GET /applications/@me once per process
to read GUILD_MEMBERS / MESSAGE_CONTENT privileged intent bits.
- Schema is rebuilt per-session in model_tools.get_tool_definitions:
hides search_members / member_info when GUILD_MEMBERS intent is off,
annotates fetch_messages description when MESSAGE_CONTENT is off.
- New config key discord.server_actions (comma-separated or YAML list)
lets users restrict which actions the agent can call, intersected
with intent availability. Unknown names are warned and dropped.
- Defense-in-depth: runtime handler re-checks the allowlist so a stale
cached schema cannot bypass a tightened config.
- Schema description rewritten as an action-first manifest (signature
per action) instead of per-parameter 'required for X, Y, Z' cross-refs.
~25% shorter; model can see each action's required params at a glance.
- Added bounds: limit gets minimum=1 maximum=100, auto_archive_duration
becomes an enum of the 4 valid Discord values.
- 403 enrichment: runtime 403 errors are mapped to actionable guidance
(which permission is missing and what to do about it) instead of the
raw Discord error body.
- 36 new tests: capability detection with caching and force refresh,
config allowlist parsing (string/list/invalid/unknown), intent+allowlist
intersection, dynamic schema build, runtime allowlist enforcement,
403 enrichment, and model_tools integration wiring.
The first draft of the fix called `chunk.decode("utf-8")` directly on
each 4096-byte `os.read()` result, which corrupts output whenever a
multi-byte UTF-8 character straddles a read boundary:
* `UnicodeDecodeError` fires on the valid-but-truncated byte sequence.
* The except handler clears ALL previously-decoded output and replaces
the whole buffer with `[binary output detected ...]`.
Empirically: 10000 '日' chars (30001 bytes) through the wrapper loses
all 10000 characters on the first draft; the baseline TextIOWrapper
drain (which uses `encoding='utf-8', errors='replace'` on Popen)
preserves them all. This regression affects any command emitting
non-ASCII output larger than one chunk — CJK/Arabic/emoji in
`npm install`, `pip install`, `docker logs`, `kubectl logs`, etc.
Fix: swap to `codecs.getincrementaldecoder('utf-8')(errors='replace')`,
which buffers partial multi-byte sequences across chunks and substitutes
U+FFFD for genuinely invalid bytes. Flush on drain exit via
`decoder.decode(b'', final=True)` to emit any trailing replacement
character for a dangling partial sequence.
Adds two regression tests:
* test_utf8_multibyte_across_read_boundary — 10000 U+65E5 chars,
verifies count round-trips and no fallback fires.
* test_invalid_utf8_uses_replacement_not_fallback — deliberate
\xff\xfe between valid ASCII, verifies surrounding text survives.
When a user's command backgrounds a child (`cmd &`, `setsid cmd & disown`,
etc.), the backgrounded grandchild inherits the write-end of our stdout
pipe via fork(). The old `for line in proc.stdout` drain never EOF'd
until the grandchild closed the pipe — so for a uvicorn server, the
terminal tool hung indefinitely (users reported the whole session
deadlocking when asking the agent to restart a backend).
Fix: switch _drain() to select()-based non-blocking reads and stop
draining shortly after bash exits even if the pipe hasn't EOF'd. Any
output the grandchild writes after that point goes to an orphaned pipe,
which is exactly what the user asked for when they said '&'.
Adds regression tests covering the issue's exact repro and 5 related
patterns (plain bg, setsid+disown, streaming output, high volume,
timeout, UTF-8).
Agents can now send arbitrary CDP commands to the browser. The tool is
gated on a reachable CDP endpoint at session start — it only appears in
the toolset when BROWSER_CDP_URL is set (from '/browser connect') or
'browser.cdp_url' is configured in config.yaml. Backends that don't
currently expose CDP to the Python side (Camofox, default local
agent-browser, cloud providers whose per-session cdp_url is not yet
surfaced) do not see the tool at all.
Tool schema description links to the CDP method reference at
https://chromedevtools.github.io/devtools-protocol/ so the agent can
web_extract specific method docs on demand.
Stateless per call. Browser-level methods (Target.*, Browser.*,
Storage.*) omit target_id. Page-level methods attach to the target
with flatten=true and dispatch the method on the returned sessionId.
Clean errors when the endpoint becomes unreachable mid-session or
the URL isn't a WebSocket.
Tests: 19 unit (mock CDP server + gate checks) + E2E against real
headless Chrome (Target.getTargets, Browser.getVersion,
Runtime.evaluate with target_id, Page.navigate + re-eval, bogus
method, bogus target_id, missing endpoint) + E2E of the check_fn
gate (tool hidden without CDP URL, visible with it, hidden again
after unset).
Add approvals.cron_mode config option that controls how cron jobs handle
dangerous commands. Previously, cron jobs silently auto-approved all
dangerous commands because there was no user present to approve them.
Now the behavior is configurable:
- deny (default): block dangerous commands and return a message telling
the agent to find an alternative approach. The agent loop continues —
it just can't use that specific command.
- approve: auto-approve all dangerous commands (previous behavior).
When a command is blocked, the agent receives the same response format as
a user denial in the CLI — exit_code=-1, status=blocked, with a message
explaining why and pointing to the config option. This keeps the agent
loop running and encourages it to adapt.
Implementation:
- config.py: add approvals.cron_mode to DEFAULT_CONFIG
- scheduler.py: set HERMES_CRON_SESSION=1 env var before agent runs
- approval.py: both check_command_approval() and check_all_command_guards()
now check for cron sessions and apply the configured mode
- 21 new tests covering config parsing, deny/approve behavior, and
interaction with other bypass mechanisms (yolo, containers)
Stacking both features on the same event produces duplicate, delayed
notifications — delivery is async and continues firing after the process
exits, so matches on end-of-run markers (SUMMARY, DONE, PASS) arrive
after the agent has already polled/waited and moved on.
Updates both the terminal tool JSON schema description and the
terminal_tool() function docstring to make the split explicit:
- watch_patterns: mid-process signals only (errors, readiness markers,
intermediate steps you want to react to before the process exits)
- notify_on_complete: end-of-run completion signal
No behavioural change.
Weaker models (Gemma-class) repeatedly rediscover and forget that
execute_code uses a different CWD and Python interpreter than terminal(),
causing them to flip-flop on whether user files exist and to hit import
errors on project dependencies like pandas.
Adds a new 'code_execution.mode' config key (default 'project') that
brings execute_code into line with terminal()'s filesystem/interpreter:
project (new default):
- cwd = session's TERMINAL_CWD (falls back to os.getcwd())
- python = active VIRTUAL_ENV/bin/python or CONDA_PREFIX/bin/python
with a Python 3.8+ version check; falls back cleanly to
sys.executable if no venv or the candidate fails
- result : 'import pandas' works, '.env' resolves, matches terminal()
strict (opt-in):
- cwd = staging tmpdir (today's behavior)
- python = sys.executable (today's behavior)
- result : maximum reproducibility and isolation; project deps
won't resolve
Security-critical invariants are identical across both modes and covered by
explicit regression tests:
- env scrubbing (strips *_API_KEY, *_TOKEN, *_SECRET, *_PASSWORD,
*_CREDENTIAL, *_PASSWD, *_AUTH substrings)
- SANDBOX_ALLOWED_TOOLS whitelist (no execute_code recursion, no
delegate_task, no MCP from inside scripts)
- resource caps (5-min timeout, 50KB stdout, 50 tool calls)
Deliberately avoids 'sandbox'/'isolated'/'cloud' language in tool
descriptions (regression from commit 39b83f34 where agents on local
backends falsely believed they were sandboxed and refused networking).
Override via env var: HERMES_EXECUTE_CODE_MODE=strict|project
Weaker models (Gemma-class) repeatedly rediscover and forget that execute_code's
working directory differs from terminal()/read_file()'s, leading to
os.path.exists('.env') returning False even though the file exists in the
session's CWD. They then bounce between 'the file exists' and 'the file is
missing' across tool calls.
Adds a 'Working directory' note to the execute_code schema description
pointing agents at absolute paths (os.path.expanduser) or terminal()/read_file()
for inspecting user files.
Carefully avoids the 'sandbox'/'isolated'/'cloud' language that commit
39b83f34 removed (it caused agents on local backends to refuse networking
tasks and save false sandbox beliefs to persistent memory). Purely factual
CWD guidance — no restriction implications.
Error messages that tell users to install optional extras now use
{sys.executable} -m pip install ... instead of a bare 'pip install
hermes-agent[extra]' string. Under the curl installer, bare 'pip'
resolves to system pip, which either fails with PEP 668
externally-managed-environment or installs into the wrong Python.
Affects: hermes dashboard, hermes web server startup, mcp_serve,
hermes doctor Bedrock check, CLI voice mode, voice_mode tool runtime
error, Discord voice-channel join failure message.
* fix(interrupt): propagate to concurrent-tool workers + opt-in debug trace
interrupt() previously only flagged the agent's _execution_thread_id.
Tools running inside _execute_tool_calls_concurrent execute on
ThreadPoolExecutor worker threads whose tids are distinct from the
agent's, so is_interrupted() inside those tools returned False no matter
how many times the gateway called .interrupt() — hung ssh / curl / long
make-builds ran to their own timeout.
Changes:
- run_agent.py: track concurrent-tool worker tids in a per-agent set,
fan interrupt()/clear_interrupt() out to them, and handle the
register-after-interrupt race at _run_tool entry. getattr fallback
for the tracker so test stubs built via object.__new__ keep working.
- tools/environments/base.py: opt-in _wait_for_process trace (ENTER,
per-30s HEARTBEAT with interrupt+activity-cb state, INTERRUPT
DETECTED, TIMEOUT, EXIT) behind HERMES_DEBUG_INTERRUPT=1.
- tools/interrupt.py: opt-in set_interrupt() trace (caller tid, target
tid, set snapshot) behind the same env flag.
- tests: new regression test runs a polling tool on a concurrent worker
and asserts is_interrupted() flips to True within ~1s of interrupt().
Second new test guards clear_interrupt() clearing tracked worker bits.
Validation: tests/run_agent/ all 762 pass; tests/tools/ interrupt+env
subset 216 pass.
* fix(interrupt-debug): bypass quiet_mode logger filter so trace reaches agent.log
AIAgent.__init__ sets logging.getLogger('tools').setLevel(ERROR) when
quiet_mode=True (the CLI default). This would silently swallow every
INFO-level trace line from the HERMES_DEBUG_INTERRUPT=1 instrumentation
added in the parent commit — confirmed by running hermes chat -q with
the flag and finding zero trace lines in agent.log even though
_wait_for_process was clearly executing (subprocess pid existed).
Fix: when HERMES_DEBUG_INTERRUPT=1, each traced module explicitly sets
its own logger level to INFO at import time, overriding the 'tools'
parent-level filter. Scoped to the opt-in case only, so production
(quiet_mode default) logs stay quiet as designed.
Validation: hermes chat -q with HERMES_DEBUG_INTERRUPT=1 now writes
'_wait_for_process ENTER/EXIT' lines to agent.log as expected.
* fix(cli): SIGTERM/SIGHUP no longer orphans tool subprocesses
Tool subprocesses spawned by the local environment backend use
os.setsid so they run in their own process group. Before this fix,
SIGTERM/SIGHUP to the hermes CLI killed the main thread via
KeyboardInterrupt but the worker thread running _wait_for_process
never got a chance to call _kill_process — Python exited, the child
was reparented to init (PPID=1), and the subprocess ran to its
natural end (confirmed live: sleep 300 survived 4+ min after SIGTERM
to the agent until manual cleanup).
Changes:
- cli.py _signal_handler (interactive) + _signal_handler_q (-q mode):
route SIGTERM/SIGHUP through agent.interrupt() so the worker's poll
loop sees the per-thread interrupt flag and calls _kill_process
(os.killpg) on the subprocess group. HERMES_SIGTERM_GRACE (default
1.5s) gives the worker time to complete its SIGTERM+SIGKILL
escalation before KeyboardInterrupt unwinds main.
- tools/environments/base.py _wait_for_process: wrap the poll loop in
try/except (KeyboardInterrupt, SystemExit) so the cleanup fires
even on paths the signal handlers don't cover (direct sys.exit,
unhandled KI from nested code, etc.). Emits EXCEPTION_EXIT trace
line when HERMES_DEBUG_INTERRUPT=1.
- New regression test: injects KeyboardInterrupt into a running
_wait_for_process via PyThreadState_SetAsyncExc, verifies the
subprocess process group is dead within 3s of the exception and
that KeyboardInterrupt re-raises cleanly afterward.
Validation:
| Before | After |
|---------------------------------------------------------|--------------------|
| sleep 300 survives 4+ min as PPID=1 orphan after SIGTERM | dies within 2 s |
| No INTERRUPT DETECTED in trace | INTERRUPT DETECTED fires + killing process group |
| tests/tools/test_local_interrupt_cleanup | 1/1 pass |
| tests/run_agent/test_concurrent_interrupt | 4/4 pass |
Extend forum support from PR #10145:
- REST path (_send_discord): forum thread creation now uploads media
files as multipart attachments on the starter message in a single
call. Previously media files were silently dropped on the forum
path.
- Websocket media paths (_send_file_attachment, send_voice, send_image,
send_animation — covers send_image_file, send_video, send_document
transitively): forum channels now go through a new _forum_post_file
helper that creates a thread with the file as starter content,
instead of failing via channel.send(file=...) which forums reject.
- _send_to_forum chunk follow-up failures are collected into
raw_response['warnings'] so partial-send outcomes surface.
- Process-local probe cache (_DISCORD_CHANNEL_TYPE_PROBE_CACHE) avoids
GET /channels/{id} on every uncached send after the first.
- Dedup of TestSendDiscordMedia that the PR merge-resolution left
behind.
- Docs: Forum Channels section under website/docs/user-guide/messaging/discord.md.
Tests: 117 passed (22 new for forum+media, probe cache, warnings).
ShellFileOperations captured the terminal env's cwd at __init__ time and
used that stale value for every subsequent _exec() call. When the user
ran `cd` via the terminal tool, `env.cwd` updated but `ops.cwd` did not.
Relative paths passed to patch_replace / read_file / write_file / search
then targeted the ORIGINAL directory instead of the current one.
Observed symptom in agent sessions:
terminal: cd .worktrees/my-branch
patch hermes_cli/main.py <old> <new>
→ returns {"success": true} with a plausible unified diff
→ but `git diff` in the worktree shows nothing
→ the patch landed in the main repo's checkout of main.py instead
The diff looked legitimate because patch_replace computes it from the
IN-MEMORY content vs new_content, not by re-reading the file. The
write itself DID succeed — it just wrote to the wrong directory's copy
of the same-named file.
Fix: _exec() now resolves cwd from live sources in this order:
1. Explicit `cwd` arg (if provided by the caller)
2. Live `self.env.cwd` (tracks `cd` commands run via terminal)
3. Init-time `self.cwd` (fallback when env has no cwd attribute)
Includes a 5-test regression suite covering:
- cd followed by relative read follows live cwd
- the exact reported bug: patch_replace with relative path after cd
- explicit cwd= arg still wins over env.cwd
- env without cwd attribute falls back to init-time cwd
- patch_replace success reflects real file state (safety rail)
Co-authored-by: teknium1 <teknium@nousresearch.com>
Follow-up polish on top of the cherry-picked #11023 commit.
- feishu_comment_rules.py: replace import-time "~/.hermes" expanduser fallback
with get_hermes_home() from hermes_constants (canonical, profile-safe).
- tools/feishu_doc_tool.py, tools/feishu_drive_tool.py: drop the
asyncio.get_event_loop().run_until_complete(asyncio.to_thread(...)) dance.
Tool handlers run synchronously in a worker thread with no running loop, so
the RuntimeError branch was always the one that executed. Calls client.request
directly now. Unused asyncio import removed.
- tests/gateway/test_feishu.py: add register_p2_customized_event to the mock
EventDispatcher builder so the existing adapter test matches the new handler
registration for drive.notice.comment_add_v1.
- scripts/release.py: map liujinkun@bytedance.com -> liujinkun2025 for
contributor attribution on release notes.
- Full comment handler: parse drive.notice.comment_add_v1 events, build
timeline, run agent, deliver reply with chunking support.
- 5 tools: feishu_doc_read, feishu_drive_list_comments,
feishu_drive_list_comment_replies, feishu_drive_reply_comment,
feishu_drive_add_comment.
- 3-tier access control rules (exact doc > wildcard "*" > top-level >
defaults) with per-field fallback. Config via
~/.hermes/feishu_comment_rules.json, mtime-cached hot-reload.
- Self-reply filter using generalized self_open_id (supports future
user-identity subscriptions). Receiver check: only process events
where the bot is the @mentioned target.
- Smart timeline selection, long text chunking, semantic text extraction,
session sharing per document, wiki link resolution.
Change-Id: I31e82fd6355173dbcc400b8934b6d9799e3137b9
Both fixes close process leaks observed in production (18+ orphaned
agent-browser node daemons, 15+ orphaned paste.rs sleep interpreters
accumulated over ~3 days, ~2.7 GB RSS).
## agent-browser daemon leak
Previously the orphan reaper (_reap_orphaned_browser_sessions) only ran
from _start_browser_cleanup_thread, which is only invoked on the first
browser tool call in a process. Hermes sessions that never used the
browser never swept orphans, and the cross-process orphan detection
relied on in-process _active_sessions, which doesn't see other hermes
PIDs' sessions (race risk).
- Write <session>.owner_pid alongside the socket dir recording the
hermes PID that owns the daemon (extracted into _write_owner_pid for
direct testability).
- Reaper prefers owner_pid liveness over in-process _active_sessions.
Cross-process safe: concurrent hermes instances won't reap each
other's daemons. Legacy tracked_names fallback kept for daemons
that predate owner_pid.
- atexit handler (_emergency_cleanup_all_sessions) now always runs
the reaper, not just when this process had active sessions —
every clean hermes exit sweeps accumulated orphans.
## paste.rs auto-delete leak
_schedule_auto_delete spawned a detached Python subprocess per call
that slept 6 hours then issued DELETE requests. No dedup, no tracking —
every 'hermes debug share' invocation added ~20 MB of resident Python
interpreters that stuck around until the sleep finished.
- Replaced the spawn with ~/.hermes/pastes/pending.json: records
{url, expire_at} entries.
- _sweep_expired_pastes() synchronously DELETEs past-due entries on
every 'hermes debug' invocation (run_debug() dispatcher).
- Network failures stay in pending.json for up to 24h, then give up
(paste.rs's own retention handles the 'user never runs hermes again'
edge case).
- Zero subprocesses; regression test asserts subprocess/Popen/time.sleep
never appear in the function source (skipping docstrings via AST).
## Validation
| | Before | After |
|------------------------------|---------------|--------------|
| Orphan agent-browser daemons | 18 accumulated| 2 (live) |
| paste.rs sleep interpreters | 15 accumulated| 0 |
| RSS reclaimed | - | ~2.7 GB |
| Targeted tests | - | 2253 pass |
E2E verified: alive-owner daemons NOT reaped; dead-owner daemons
SIGTERM'd and socket dirs cleaned; pending.json sweep deletes expired
entries without spawning subprocesses.
Two accretion-over-time leaks that compound over long CLI / gateway
lifetimes. Both were flagged in the memory-leak audit.
## file_tools._read_tracker
_read_tracker[task_id] holds three sub-containers that grew unbounded:
read_history set of (path, offset, limit) tuples — 1 per unique read
dedup dict of (path, offset, limit) → mtime — same growth pattern
read_timestamps dict of resolved_path → mtime — 1 per unique path
A CLI session uses one stable task_id for its lifetime, so these were
uncapped. A 10k-read session accumulated ~1.5MB of tracker state that
the tool no longer needed (only the most recent reads are relevant for
dedup, consecutive-loop detection, and write/patch external-edit
warnings).
Fix: _cap_read_tracker_data() enforces hard caps on each container
after every add. Defaults: read_history=500, dedup=1000,
read_timestamps=1000. Eviction is insertion-order (Python 3.7+ dict
guarantee) for the dicts; arbitrary for the set (which only feeds
diagnostic summaries).
## process_registry._completion_consumed
Module-level set that recorded every session_id ever polled / waited /
logged. No pruning. Each entry is ~20 bytes, so the absolute leak is
small, but on a gateway processing thousands of background commands
per day the set grows until process exit.
Fix: _prune_if_needed() now discards _completion_consumed entries
alongside the session dict evictions it already performs (both the
TTL-based prune and the LRU-over-cap prune). Adds a final
belt-and-suspenders pass that drops any dangling entries whose
session_id no longer appears in _running or _finished.
Tests: tests/tools/test_accretion_caps.py — 9 cases
* Each container bound respected, oldest evicted
* No-op when under cap (no unnecessary work)
* Handles missing sub-containers without crashing
* Live read_file_tool path enforces caps end-to-end
* _completion_consumed pruned on TTL expiry
* _completion_consumed pruned on LRU eviction
* Dangling entries (no backing session) cleared
Broader suite: 3486 tests/tools + tests/cli pass. The single flake
(test_alias_command_passes_args) reproduces on unchanged main — known
cross-test pollution under suite-order load.
- gateway/platforms/weixin.py:
- Split aiohttp.ClientSession into _poll_session and _send_session
- Add _LIVE_ADAPTERS registry so send_weixin_direct() reuses the connected gateway adapter instead of creating a competing session
- Fixes silent message loss when gateway is running (iLink token contention)
- cron/scheduler.py:
- Support comma-separated deliver values (e.g. 'feishu,weixin') for multi-target delivery
- Delay pconfig/enabled check until standalone fallback so live adapters work even when platform is not in gateway config
- tools/send_message_tool.py:
- Synthesize PlatformConfig from WEIXIN_* env vars when gateway config lacks a weixin entry
- Fall back to WEIXIN_HOME_CHANNEL env var for home channel resolution
- tests/gateway/test_weixin.py:
- Update mocks to include _send_session
The send_message tool's direct-REST QQBot path used "QQBotAccessToken {token}"
which QQ's API rejects with 401. The correct format is "QQBot {token}" — the
gateway adapter at gateway/platforms/qqbot.py uses this format in all 5 header
sites (lines 341, 551, 579, 1068, 1467); this was the one outlier.
Credit to @Quon for surfacing this in #10257 (that PR had unrelated issues in
its media-upload logic and was closed; this salvages the genuine 1-line fix).
When a user edits a bundled skill, sync flags it as user_modified and
skips it forever. The problem: if the user later tries to undo the edit
by copying the current bundled version back into ~/.hermes/skills/, the
manifest still holds the old origin hash from the last successful
sync, so the fresh bundled hash still doesn't match and the skill stays
stuck as user_modified.
Adds an escape hatch for this case.
hermes skills reset <name>
Drops the skill's entry from ~/.hermes/skills/.bundled_manifest and
re-baselines against the user's current copy. Future 'hermes update'
runs accept upstream changes again. Non-destructive.
hermes skills reset <name> --restore
Also deletes the user's copy and re-copies the bundled version.
Use when you want the pristine upstream skill back.
Also available as /skills reset in chat.
- tools/skills_sync.py: new reset_bundled_skill(name, restore=False)
- hermes_cli/skills_hub.py: do_reset() + wired into skills_command and
handle_skills_slash; added to the slash /skills help panel
- hermes_cli/main.py: argparse entry for 'hermes skills reset'
- tests/tools/test_skills_sync.py: 5 new tests covering the stuck-flag
repro, --restore, unknown-skill error, upstream-removed-skill, and
no-op on already-clean state
- website/docs/user-guide/features/skills.md: new 'Bundled skill updates'
section explaining the origin-hash mechanic + reset usage
* feat(image_gen): upgrade Recraft V3 → V4 Pro, Nano Banana → Pro
Upstream asked for these two upgrades ASAP — the old entries show
stale models when newer, higher-quality versions are available on FAL.
Recraft V3 → Recraft V4 Pro
ID: fal-ai/recraft-v3 → fal-ai/recraft/v4/pro/text-to-image
Price: $0.04/image → $0.25/image (6x — V4 Pro is premium tier)
Schema: V4 dropped the required `style` enum entirely; defaults
handle taste now. Added `colors` and `background_color`
to supports for brand-palette control. `seed` is not
supported by V4 per the API docs.
Nano Banana → Nano Banana Pro
ID: fal-ai/nano-banana → fal-ai/nano-banana-pro
Price: $0.08/image → $0.15/image (1K); $0.30 at 4K
Schema: Aspect ratio family unchanged. Added `resolution`
(1K/2K/4K, default 1K for billing predictability),
`enable_web_search` (real-time info grounding, +$0.015),
and `limit_generations` (force exactly 1 image).
Architecture: Gemini 2.5 Flash → Gemini 3 Pro Image. Quality
and reasoning depth improved; slower (~6s → ~8s).
Migration: users who had the old IDs in `image_gen.model` will
fall through the existing 'unknown model → default' warning path
in `_resolve_fal_model()` and get the Klein 9B default on the next
run. Re-run `hermes tools` → Image Generation to pick the new
version. No silent cost-upgrade aliasing — the 2-6x price jump
on these tiers warrants explicit user re-selection.
Portal note: both new model IDs need to be allowlisted on the
Nous fal-queue-gateway alongside the previous 7 additions, or
users on Nous Subscription will see the 'managed gateway rejected
model' error we added previously (which is clear and
self-remediating, just noisy).
* docs: wrap '<1s' in backticks to unblock MDX compilation
Docusaurus's MDX parser treats unquoted '<' as the start of JSX, and
'<1s' fails because '1' isn't a valid tag-name start character. This
was broken on main since PR #11265 (never noticed because
docs-site-checks was failing on OTHER issues at the time and we
admin-merged through it).
Wrapping in backticks also gives the cell monospace styling which
reads more cleanly alongside the inline-code model ID in the same row.
The other '<1s' occurrence (line 52) is inside a fenced code block
and is already safe — code fences bypass MDX parsing.
* feat(mcp-oauth): scaffold MCPOAuthManager
Central manager for per-server MCP OAuth state. Provides
get_or_build_provider (cached), remove (evicts cache + deletes
disk), invalidate_if_disk_changed (mtime watch, core fix for
external-refresh workflow), and handle_401 (dedup'd recovery).
No behavior change yet — existing call sites still use
build_oauth_auth directly. Task 1 of 8 in the MCP OAuth
consolidation (fixes Cthulhu's BetterStack reliability issues).
* feat(mcp-oauth): add HermesMCPOAuthProvider with pre-flow disk watch
Subclasses the MCP SDK's OAuthClientProvider to inject a disk
mtime check before every async_auth_flow, via the central
manager. When a subclass instance is used, external token
refreshes (cron, another CLI instance) are picked up before
the next API call.
Still dead code: the manager's _build_provider still delegates
to build_oauth_auth and returns the plain OAuthClientProvider.
Task 4 wires this subclass in. Task 2 of 8.
* refactor(mcp-oauth): extract build_oauth_auth helpers
Decomposes build_oauth_auth into _configure_callback_port,
_build_client_metadata, _maybe_preregister_client, and
_parse_base_url. Public API preserved. These helpers let
MCPOAuthManager._build_provider reuse the same logic in Task 4
instead of duplicating the construction dance.
Also updates the SDK version hint in the warning from 1.10.0 to
1.26.0 (which is what we actually require for the OAuth types
used here). Task 3 of 8.
* feat(mcp-oauth): manager now builds HermesMCPOAuthProvider directly
_build_provider constructs the disk-watching subclass using the
helpers from Task 3, instead of delegating to the plain
build_oauth_auth factory. Any consumer using the manager now gets
pre-flow disk-freshness checks automatically.
build_oauth_auth is preserved as the public API for backwards
compatibility. The code path is now:
MCPOAuthManager.get_or_build_provider ->
_build_provider ->
_configure_callback_port
_build_client_metadata
_maybe_preregister_client
_parse_base_url
HermesMCPOAuthProvider(...)
Task 4 of 8.
* feat(mcp): wire OAuth manager + add _reconnect_event
MCPServerTask gains _reconnect_event alongside _shutdown_event.
When set, _run_http / _run_stdio exit their async-with blocks
cleanly (no exception), and the outer run() loop re-enters the
transport to rebuild the MCP session with fresh credentials.
This is the recovery path for OAuth failures that the SDK's
in-place httpx.Auth cannot handle (e.g. cron externally consumed
the refresh_token, or server-side session invalidation).
_run_http now asks MCPOAuthManager for the OAuth provider
instead of calling build_oauth_auth directly. Config-time,
runtime, and reconnect paths all share one provider instance
with pre-flow disk-watch active.
shutdown() defensively sets both events so there is no race
between reconnect and shutdown signalling.
Task 5 of 8.
* feat(mcp): detect auth failures in tool handlers, trigger reconnect
All 5 MCP tool handlers (tool call, list_resources, read_resource,
list_prompts, get_prompt) now detect auth failures and route
through MCPOAuthManager.handle_401:
1. If the manager says recovery is viable (disk has fresh tokens,
or SDK can refresh in-place), signal MCPServerTask._reconnect_event
to tear down and rebuild the MCP session with fresh credentials,
then retry the tool call once.
2. If no recovery path exists, return a structured needs_reauth
JSON error so the model stops hallucinating manual refresh
attempts (the 'let me curl the token endpoint' loop Cthulhu
pasted from Discord).
_is_auth_error catches OAuthFlowError, OAuthTokenError,
OAuthNonInteractiveError, and httpx.HTTPStatusError(401). Non-auth
exceptions still surface via the generic error path unchanged.
Task 6 of 8.
* feat(mcp-cli): route add/remove through manager, add 'hermes mcp login'
cmd_mcp_add and cmd_mcp_remove now go through MCPOAuthManager
instead of calling build_oauth_auth / remove_oauth_tokens
directly. This means CLI config-time state and runtime MCP
session state are backed by the same provider cache — removing
a server evicts the live provider, adding a server populates
the same cache the MCP session will read from.
New 'hermes mcp login <name>' command:
- Wipes both the on-disk tokens file and the in-memory
MCPOAuthManager cache
- Triggers a fresh OAuth browser flow via the existing probe
path
- Intended target for the needs_reauth error Task 6 returns
to the model
Task 7 of 8.
* test(mcp-oauth): end-to-end integration tests
Five new tests exercising the full consolidation with real file
I/O and real imports (no transport mocks):
1. external_refresh_picked_up_without_restart — Cthulhu's cron
workflow. External process writes fresh tokens to disk;
on the next auth flow the manager's mtime-watch flips
_initialized and the SDK re-reads from storage.
2. handle_401_deduplicates_concurrent_callers — 10 concurrent
handlers for the same failed token fire exactly ONE recovery
attempt (thundering-herd protection).
3. handle_401_returns_false_when_no_provider — defensive path
for unknown servers.
4. invalidate_if_disk_changed_handles_missing_file — pre-auth
state returns False cleanly.
5. provider_is_reused_across_reconnects — cache stickiness so
reconnects preserve the disk-watch baseline mtime.
Task 8 of 8 — consolidation complete.
