Closes#4759, closes#4381.
Mutating actions (patch, edit, write_file, remove_file, delete) used to
refuse skills that lived under `skills.external_dirs` with 'Skill X is in
an external directory and cannot be modified. Copy it to your local skills
directory first.' Faced with that error, the agent would fall back to
action='create', which always writes under ~/.hermes/skills/ — producing
a silent duplicate of the external skill in the local store.
Fix: drop the read-only gate. `skills.external_dirs` is configured by the
user; if they pointed it at a directory, they already said 'these are my
skills, treat them the same.' Filesystem permissions handle the genuine
read-only case (write fails, agent sees the error).
- New _containing_skills_root() resolves whichever dir actually contains
the skill; _delete_skill uses it to bound empty-category cleanup so an
external root is never rmdir'd.
- _create_skill behavior is unchanged: new skills still land in local
SKILLS_DIR only. Fewer moving parts.
- Seven new TestExternalSkillMutations tests covering patch/edit/write_file/
remove_file/delete/create against a mocked two-root layout + a category
rmdir-safety check.
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.
