Group the secrets import with time and webbrowser at the top of
run_hermes_oauth_login_pure(), matching the existing pattern.
Drop the _secrets alias — no name conflict in this scope.
The PKCE flow reused the code_verifier as the OAuth state parameter.
Per RFC 6749 §10.12 and RFC 7636, these serve different purposes:
state is an anti-CSRF token visible in the authorization URL; the
code_verifier must remain secret for the token exchange.
Generate an independent secrets.token_urlsafe(32) for state and
validate it on callback to provide actual CSRF protection.
Closes#10693
The long-lived prefix-cache layout split the system prompt into stable/
context/volatile blocks and re-derived them on every API call. The
volatile tier (timestamp + memory snapshot + USER profile) ticks per
turn, so the system message bytes mutated mid-conversation and broke
upstream prompt caches (OpenRouter, Nous Portal, Anthropic).
Diagnosed via live wire-format diffing: an 8-turn conversation showed
OLD layout flipping system block[1] sha mid-session at the minute
boundary, dropping cached_tokens to 0 on that turn (cumulative
66.6% vs 83.3% for the single-block layout). Hermes invariant:
history (system + all but the last 1-2 messages) must be static.
Fix: drop the long-lived layout entirely. Single layout everywhere —
system_and_3 with one cached system string built once on first turn,
replayed verbatim on every subsequent turn. Loses cross-session 1h
prefix caching for Claude (the feature that motivated the split), but
within-session caching now actually works on every provider.
Removed:
- run_agent.py: _use_long_lived_prefix_cache flag, _long_lived_cache_ttl,
_supports_long_lived_anthropic_cache method, the long-lived branch in
run_conversation, mark_tools_for_long_lived_cache call site
- agent/prompt_caching.py: apply_anthropic_cache_control_long_lived,
mark_tools_for_long_lived_cache, _mark_system_stable_block helper
- hermes_cli/config.py: prompt_caching.long_lived_prefix and
prompt_caching.long_lived_ttl config keys
- tests/agent/test_prompt_caching_live.py (entire file)
- tests/agent/test_prompt_caching.py: TestMarkToolsForLongLivedCache,
TestApplyAnthropicCacheControlLongLived
- tests/run_agent/test_anthropic_prompt_cache_policy.py:
TestSupportsLongLivedAnthropicCache
Targeted tests: 62/62 pass.
* feat(security): supply-chain advisory checker + lazy-install framework + tiered install fallback
Three coordinated mitigations for the Mini Shai-Hulud worm hitting
mistralai 2.4.6 on PyPI (2026-05-12) and for the next single-package
compromise that follows.
# What this PR makes true
1. Users with the poisoned mistralai 2.4.6 in their venv get a loud
detection banner with copy-pasteable remediation steps the moment
they run hermes (and on every gateway startup).
2. One quarantined / yanked PyPI package can no longer silently demote
a fresh install to 'core only' — the installer keeps every other
extra and tells the user which tier landed.
3. Future opt-in backends (Mistral, ElevenLabs, Honcho, etc.) can
lazy-install on first use under a strict allowlist, instead of
eagerly pulling everything at install time.
# Detection: hermes_cli/security_advisories.py
- ADVISORIES catalog (one entry currently: shai-hulud-2026-05 for
mistralai==2.4.6). Adding the next one is a single dataclass.
- detect_compromised() uses importlib.metadata.version() — no pip
dependency, works in uv venvs that lack pip.
- Banner cache (~/.hermes/cache/advisory_banner_seen) rate-limits
the startup banner to once per 24h per advisory.
- Acks persisted to security.acked_advisories in config.yaml; never
re-banner after ack.
- Wired into:
* hermes doctor — runs first, prints full remediation block
* hermes doctor --ack <id> — dismisses an advisory
* cli.py interactive run() and single-query branches — short
stderr banner pointing at hermes doctor
* gateway/run.py startup — operator-visible warning in gateway.log
# Lazy-install framework: tools/lazy_deps.py
- LAZY_DEPS allowlist maps namespaced feature keys (tts.elevenlabs,
memory.honcho, provider.bedrock, etc.) to pip specs.
- ensure(feature) installs missing deps in the active venv via the
uv → pip → ensurepip ladder (matches tools_config._pip_install).
- Strict spec safety regex rejects URLs, file paths, shell metas,
pip flag injection, control chars — only PyPI-by-name accepted.
- Gated on security.allow_lazy_installs (default true) plus the
HERMES_DISABLE_LAZY_INSTALLS env var for restricted/audited envs.
- Migrated three backends as proof of pattern:
* tools/tts_tool.py — _import_elevenlabs() calls ensure first
* plugins/memory/honcho/client.py — get_honcho_client lazy-installs
* tts.mistral / stt.mistral entries pre-registered for when PyPI
restores mistralai
# Installer fallback tiers
scripts/install.sh, scripts/install.ps1, setup-hermes.sh:
- Centralised _BROKEN_EXTRAS list (currently: mistral). Edit one
array when a transitive breaks; users keep every other extra.
- New 'all minus known-broken' tier between [all] and the existing
PyPI-only-extras tier. Only kicks in when [all] fails resolve.
- All three tiers explicit: every fallback announces which tier
landed and prints a re-run hint when not on Tier 1.
- install.ps1 and install.sh both regenerate their tier specs from
the same _BROKEN_EXTRAS array so updates stay in sync.
Side effect: install.ps1 Tier 2 spec previously hardcoded 'mistral'
in its extra list — bug fixed by the refactor (mistral is filtered
out).
# Config
hermes_cli/config.py — DEFAULT_CONFIG.security gains:
- acked_advisories: [] (advisory IDs the user has dismissed)
- allow_lazy_installs: True (security gate for ensure())
No config version bump needed — both keys nest under existing
security: block, and load_config's deep-merge picks up DEFAULT_CONFIG
defaults for users with older configs.
# Tests
tests/hermes_cli/test_security_advisories.py — 23 tests covering:
- detect_compromised matches/non-matches, wildcard frozenset
- ack persistence, idempotence, blank rejection, config-failure path
- banner cache rate limiting + 24h re-banner + ack-stops-banner
- short_banner_lines / full_remediation_text / render_doctor_section /
gateway_log_message
- shipped catalog well-formedness invariant
tests/tools/test_lazy_deps.py — 40 tests covering:
- spec safety: 11 safe parametrized + 18 unsafe parametrized
- allowlist: unknown-feature rejection, namespace.name shape,
every shipped spec passes the safety regex
- security gating: config flag, env var, default, fail-open
- ensure() happy/sad paths: already-satisfied, install success,
pip stderr surfaced on failure, install-succeeds-but-still-missing
- is_available, feature_install_command
Combined: 63 new tests, all passing under scripts/run_tests.sh.
# Validation
- scripts/run_tests.sh tests/hermes_cli/test_security_advisories.py
tests/tools/test_lazy_deps.py → 63/63 passing
- scripts/run_tests.sh tests/hermes_cli/test_doctor.py
tests/hermes_cli/test_doctor_command_install.py
tests/tools/test_tts_mistral.py tests/tools/test_transcription_tools.py
tests/tools/test_transcription_dotenv_fallback.py → 165/165 passing
- scripts/run_tests.sh tests/hermes_cli/ tests/tools/ →
9191 passed, 8 pre-existing failures (verified on origin/main
before this change)
- bash -n on install.sh and setup-hermes.sh → OK
- py_compile on all modified .py files → OK
- End-to-end smoke test of detect_compromised + render_doctor_section
+ gateway_log_message with mocked installed version → produces
copy-pasteable remediation output
# Community
Full advisory + remediation steps:
website/docs/community/security-advisories/shai-hulud-mistralai-2026-05.md
Short-form post drafts (Discord, GitHub pinned issue, README banner):
scripts/community-announcement-shai-hulud.md
Refs: PR #24205 (mistral disabled), Socket Security advisory
<https://socket.dev/blog/mini-shai-hulud-worm-pypi>
* build(deps): pin every direct dep to ==X.Y.Z (no ranges)
Companion to the supply-chain advisory work: replace every >=/</~= range
in pyproject.toml's [project.dependencies] and [project.optional-dependencies]
with an exact ==X.Y.Z pin sourced from uv.lock.
Why: ranges allow PyPI to ship a fresh version of any direct dep at any
time without a code review on our side. With ranges, the malicious
mistralai 2.4.6 release would have been pulled by every fresh
'pip install -e .[all]' for the hours between upload and PyPI's
quarantine — exactly the install window we got hit on. Exact pins close
that window: the only way a new package version reaches a user is via
an intentional update on our end.
What the user-facing change is: nothing, behavior-wise. Every package
resolves to the same version it was already resolving to via uv.lock —
the pins just remove the resolver's freedom to pick a different one.
Cost: any user installing Hermes alongside another package that requires
a newer pin gets a resolver conflict. Acceptable for our isolated-venv
install path; documented in the new comment block.
Build-system requires line (setuptools>=61.0) is intentionally left
as a range — pinning the build backend would block fresh pip from
bootstrapping the build on architectures where that exact wheel isn't
available.
mistral extra (mistralai==2.3.0) is pinned but stays out of [all]
(per PR #24205). 'uv lock' regeneration will fail until PyPI restores
mistralai; lockfile regeneration is gated behind that, NOT on every PR.
LAZY_DEPS in tools/lazy_deps.py also moved to exact pins so the lazy-
install pathway can never resolve a different version than the one
declared in pyproject.toml.
Validation:
- Cross-checked all 77 pinned direct deps in pyproject.toml against
uv.lock — every pin matches the resolved version exactly.
- Cross-checked all LAZY_DEPS specs against uv.lock — same.
- 'uv pip install -e .[all] --dry-run' resolves 205 packages cleanly.
- tests/tools/test_lazy_deps.py + tests/hermes_cli/test_security_advisories.py
→ 63/63 passing (every shipped spec passes the safety regex).
- Doctor + TTS + transcription targeted suite → 146/146 passing.
* build(deps): hash-verify transitives via uv.lock; remove unresolvable [mistral] extra
You asked: 'what about the dependencies the dependencies rely on?' —
correctly noting that exact-pinning direct deps in pyproject.toml does
NOT cover the transitive graph. `pip install` and `uv pip install` both
re-resolve transitives fresh from PyPI at install time, so a compromised
transitive (e.g. `httpcore` if it got worm-poisoned tomorrow) would
still hit our users even with every direct dep exact-pinned.
# What this commit fixes
1. **Both real installer scripts now prefer `uv sync --locked` as Tier 0.**
uv.lock records SHA256 hashes for every transitive — a compromised
package with a different hash gets REJECTED. Falls through to the
existing `uv pip install` cascade if the lockfile is missing or
stale, with a loud warning that the fallback path does NOT
hash-verify transitives. Previously only `setup-hermes.sh` (the dev
path) used the lockfile; `scripts/install.sh` and `scripts/install.ps1`
(the paths fresh users actually run) skipped it.
2. **Removed the `[mistral]` extra entirely.** The `mistralai` PyPI
project is fully quarantined right now — every version returns 404,
so any pin we wrote was unresolvable, which broke `uv lock --check`
in CI. Restoration is documented in pyproject.toml as a 5-step
checklist (verify, re-add extra, re-enable in 4 modules, regenerate
lock, optionally re-add to [all]).
3. **Regenerated uv.lock.** 262 packages, mistralai/eval-type-backport/
jsonpath-python pruned. `uv lock --check` now passes.
# Defense-in-depth view
| Layer | Where | Protects against |
|----------------------------|-------------------|-------------------------------------------|
| Exact pins in pyproject | direct deps | new mistralai 2.4.6-style direct compromise |
| uv.lock + `--locked` install | transitive graph | transitive worm injection |
| Tier-0 hash-verified path | install.sh / .ps1 | actually USE the lockfile in fresh installs |
| `uv lock --check` CI gate | every PR | drift between pyproject and lockfile |
| `hermes_cli/security_advisories.py` | runtime | cleanup for users who already got hit |
The exact pinning + hash verification together close the supply-chain
gap. Without the lockfile path, exact pins alone are theater.
# Validation
- `uv lock --check` → passes (262 packages resolved, no drift).
- `bash -n` on install.sh + setup-hermes.sh → OK.
- 209/209 tests passing across new + adjacent test files
(test_lazy_deps.py, test_security_advisories.py, test_doctor.py,
test_tts_mistral.py, test_transcription_tools.py).
- TOML parse OK.
* chore: remove community announcement drafts (PR body covers it)
* build(deps): lazy-install every opt-in backend (anthropic, search, terminal, platforms, dashboard)
Extends the lazy-install framework to cover everything that's not used by
every hermes session. Base install drops from ~60 packages to 45.
Moved out of core dependencies = []:
- anthropic (only when provider=anthropic native, not via aggregators)
- exa-py, firecrawl-py, parallel-web (search backends; only when picked)
- fal-client (image gen; only when picked)
- edge-tts (default TTS but still optional)
New extras in pyproject.toml: [anthropic] [exa] [firecrawl] [parallel-web]
[fal] [edge-tts]. All added to [all].
New LAZY_DEPS entries: provider.anthropic, search.{exa,firecrawl,parallel},
tts.edge, image.fal, memory.hindsight, platform.{telegram,discord,matrix},
terminal.{modal,daytona,vercel}, tool.dashboard.
Each import site now calls ensure() before importing the SDK. Where the
module had a top-level try/except (telegram, discord, fastapi), the
graceful-fallback pattern was extended to lazy-install on first
check_*_requirements() call and re-bind module globals.
Updated test_windows_native_support.py tzdata check from snapshot
(>=2023.3 literal) to invariant (any version + win32 marker).
Validation:
- Base install: 45 packages (was ~60); 6 newly-extracted packages absent
- uv lock --check: passes (262 packages, no drift)
- 209/209 lazy_deps + advisory + doctor + tts/transcription tests passing
- py_compile clean on all 12 modified modules
Cuts input cost for first-turn Claude requests by ~85-90% on subsequent
sessions within an hour. Tools array (~13k tokens for default toolset) +
stable system prefix (~5-8k tokens) get a 1h cache_control marker; the
volatile suffix (memory, USER profile, timestamp, session id) sits in a
separate non-cached block at the end so it doesn't poison the cross-session
prefix when it changes.
Provider gate: Claude on native Anthropic (incl. OAuth subscription),
OpenRouter, and Nous Portal (which proxies to OpenRouter). All other
providers keep today's system_and_3 layout unchanged.
Layout (4 cache_control breakpoints, Anthropic max):
1. tools[-1] -> 1h (cross-session)
2. system content[0] -> 1h (cross-session, stable prefix)
3. messages[-2] -> 5m (within-session rolling)
4. messages[-1] -> 5m (within-session rolling)
Within-session rolling shrinks from 3 messages to 2 to free the breakpoint
budget. On Claude with realistic tool loadouts the long-lived tier carries
the bulk of cross-session value anyway.
System prompt is now always assembled cache-friendly: stable identity /
guidance / skills / platform hints first, then session-stable context
files (AGENTS.md, .cursorrules), then per-call volatile content. Old
single-string callers see the same logical content (same join order),
just reordered so volatile lives at the end.
Config knobs (defaults shown):
prompt_caching:
cache_ttl: "5m" # rolling-window TTL (unchanged)
long_lived_prefix: true # opt-out switch
long_lived_ttl: "1h" # cross-session prefix TTL
Live E2E (tests/agent/test_prompt_caching_live.py, gated on
OPENROUTER_API_KEY) on anthropic/claude-haiku-4.5 with default toolset:
Call 1 (cold): cache_write=13,415 cache_read=0
Call 2 (NEW agent + msg): cache_write=391 cache_read=13,025
Cross-session reuse: 97.09%
Implementation:
* agent/prompt_caching.py: new apply_anthropic_cache_control_long_lived()
+ mark_tools_for_long_lived_cache(); existing apply_anthropic_cache_control()
preserved verbatim for the fallback path.
* agent/anthropic_adapter.py: convert_tools_to_anthropic() now forwards
cache_control onto each Anthropic-format tool dict.
* run_agent.py: _build_system_prompt_parts() returns the 3-tier dict;
_build_system_prompt() joins them (backward compatible).
_supports_long_lived_anthropic_cache() policy added next to the existing
_anthropic_prompt_cache_policy() (which now also recognises Nous Portal
Claude — pre-existing gap fixed in passing).
_build_api_kwargs() resolves tools_for_api once and propagates the
marker through all four build paths (anthropic_messages, bedrock,
codex_responses, profile/legacy chat completions).
Long-lived flag plumbed into the runtime snapshot/restore + model-switch
+ fallback-promotion paths.
Tests:
* tests/agent/test_prompt_caching.py: +8 tests (TestMarkToolsForLongLivedCache,
TestApplyAnthropicCacheControlLongLived).
* tests/run_agent/test_anthropic_prompt_cache_policy.py: +9 tests
(TestSupportsLongLivedAnthropicCache matrix across 8 endpoint classes
+ a fallback-target case).
* tests/agent/test_prompt_caching_live.py: new live E2E (skipif when
OPENROUTER_API_KEY is unset; runs outside the hermetic suite).
* Targeted suites: 327/327 pass (caching/adapter/policy/builder).
* tests/agent/ + tests/run_agent/: 3992 pass, 17 skip, 1 pre-existing
flake (test_async_httpx_del_neuter::test_same_key_replaces_stale_loop_entry,
verified failing on pristine origin/main).
Replace with for all literal-tuple
membership tests. Set lookup is O(1) vs O(n) for tuple — consistent
micro-optimization across the codebase.
608 instances fixed via `ruff --fix --unsafe-fixes`, 0 remaining.
133 files, +626/-626 (net zero).
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.
Per https://platform.claude.com/docs/en/build-with-claude/fast-mode:
"Fast mode is currently supported on Opus 4.6 only. Sending speed: fast
with an unsupported model returns an error."
Pre-fix, _is_anthropic_fast_model() returned True for any claude-* model,
so /fast on Opus 4.7 (or Sonnet/Haiku) would persist agent.service_tier=fast
in config.yaml and the adapter would inject extra_body["speed"] = "fast"
on every subsequent request. Opus 4.7 returns:
HTTP 400: 'claude-opus-4-7' does not support the `speed` parameter.
This wedged sessions across model upgrades (a user who ran /fast on Opus 4.6
and later switched the default model to 4.7 hit a hard 400 on every turn
until they manually edited config.yaml).
Changes:
- _is_anthropic_fast_model: gate on "opus-4-6" / "opus-4.6" only
- anthropic_adapter: add _supports_fast_mode predicate as defensive guard
so stale request_overrides on an unsupported model are dropped silently
instead of 400'ing
- Tests: flip the assertions that mirrored the bug (Sonnet/Haiku/Opus 4.7
asserting fast-mode support) to match the documented API contract
Extends the existing _normalize_tool_input_schema to also drop top-level
union keywords that Anthropic's tool schema validator rejects with HTTP 400.
Several upstream and plugin tools ship schemas with a top-level oneOf/
allOf/anyOf (common for Pydantic discriminated unions). The existing
strip_nullable_unions pass only handles anyOf-with-null patterns; a
non-null top-level union keyword sails through and hits the API.
Salvage of #16471 — approach folded into the existing normalize helper
rather than introducing a parallel _sanitize_input_schema function, to
avoid two schema-munging code paths running against the same input.
Co-authored-by: Grey0202 <grey0202@users.noreply.github.com>
DashScope's Anthropic-compatible endpoint enforces max_tokens ∈ [1, 65536].
Adding "qwen3" to _ANTHROPIC_OUTPUT_LIMITS prevents 400 errors that were
misclassified as context overflow, triggering premature compression.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Providers like Google Vertex, Azure, and Amazon Bedrock reject API
requests with duplicate tool names (HTTP 400: 'Tool names must be
unique'). The upstream injection paths in run_agent.py already dedup
after PR #17335, but two API-boundary functions pass tools through
without checking:
- agent/auxiliary_client.py: _build_call_kwargs() (all non-Anthropic
providers in chat_completions mode)
- agent/anthropic_adapter.py: convert_tools_to_anthropic() (Anthropic
Messages API path)
Add defensive dedup guards at both sites. Duplicates are dropped with
a warning log, converting a hard 400 failure into a recoverable
condition. This is intentionally conservative — the root-cause dedup
in run_agent.py is the primary defense; these guards add resilience
against future injection-path regressions.
Includes 8 new tests covering unique passthrough, duplicate removal,
empty/None edge cases.
Closes#18478
Keep context-1m-2025-08-07 in OAuth requests by default so 1M-capable
subscriptions retain full context. When Anthropic rejects a request with
400 'long context beta is not yet available for this subscription',
disable the beta for the rest of the session, rebuild the client, and
retry once.
Addresses #17680 (thanks @JayGwod for the clean reproduction) without
forcing every OAuth user off the 1M context window.
Changes:
- agent/error_classifier.py: new FailoverReason.oauth_long_context_beta_forbidden;
pattern matches 400 + 'long context beta' + 'not yet available'. Narrow
enough that the existing 429 tier-gate pattern keeps its own reason.
- agent/anthropic_adapter.py: _common_betas_for_base_url,
build_anthropic_client, build_anthropic_kwargs gain drop_context_1m_beta
kwarg. Default=False (1M stays). OAuth OAUTH_ONLY_BETAS unchanged.
- agent/transports/anthropic.py: build_kwargs forwards the flag.
- run_agent.py: self._oauth_1m_beta_disabled flag, retry-once guard,
recovery branch next to the image-shrink path. _rebuild_anthropic_client
honors the flag. The main build_kwargs call site threads it through for
fast-mode extra_headers.
- hermes_cli/doctor.py, hermes_cli/models.py: sibling OAuth /v1/models
probes get the same reactive retry — previously they'd falsely report
the Anthropic API as unreachable for affected subscriptions.
Tests: 2190 tests/agent/ + 94 adjacent integration tests pass. New unit
tests cover the classifier pattern (including the collision guard against
the 429 tier-gate) and the drop_context_1m_beta adapter behavior (default
keeps 1M, flag strips only 1M while preserving every other beta).
DeepSeek's /anthropic endpoint requires thinking blocks to be replayed
in multi-turn conversations for reasoning continuity. The existing code
classified api.deepseek.com as a generic third-party endpoint and stripped
ALL thinking blocks, causing HTTP 400 from DeepSeek.
Fix: add _is_deepseek_anthropic_endpoint() detector (following the Kimi
precedent) and a dedicated branch that strips only signed Anthropic blocks
while preserving unsigned ones synthesised from reasoning_content.
This follows the exact same pattern as the Kimi exemption (issue #13848)
and does not change behavior for any other third-party endpoint (Azure,
Bedrock, MiniMax, etc.).
FixesNousResearch/hermes-agent#16748
The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was
matched on `https://api.kimi.com/coding` only. Users configuring a
custom Kimi-compatible gateway (or an official Moonshot host) with
`api_mode: anthropic_messages` fall through to the generic third-party
path, which strips thinking blocks AND still sends
`thinking={enabled,...}` → upstream rejects with HTTP 400
"reasoning_content is missing in assistant tool call message at index N"
on the next request after a tool call.
Replace `_is_kimi_coding_endpoint` callers (history replay + thinking
kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also
matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and
Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`,
…) for custom / proxied endpoints. Keeps the UA-header check in
`build_anthropic_client` URL-only — the `claude-code/0.1.0` header is
an official-Kimi contract.
Plumbs optional `model` through `convert_messages_to_anthropic` so
the unsigned reasoning_content→thinking block synthesised for Kimi's
history validation survives the third-party signature-stripping pass
on custom hosts too.
Closes#17057.
The normalize_model_name() function unconditionally converted dots to
hyphens in all model names. This caused non-Anthropic models (e.g.
gpt-5.4) to be mangled to gpt-5-4 when routed through the Anthropic
adapter path, resulting in HTTP 404 from the backend.
Now only applies dot-to-hyphen conversion for models starting with
"claude-" or "anthropic/", which are the actual Anthropic model IDs.
Fixes NousResearch/hermes-agent#17171
Related: #7421, #13061, #16417
* 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>
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>
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.
* fix(anthropic): remove Claude Code fingerprinting from OAuth Messages API path
OAuth requests now identify as Hermes on the wire. Removed:
- "You are Claude Code, Anthropic's official CLI for Claude." system
prompt prepend
- Hermes Agent → Claude Code / Nous Research → Anthropic
system-prompt substitutions
- mcp_ tool-name prefix on outgoing tool schemas + message history
- Matching mcp_ strip on inbound tool_use blocks (strip_tool_prefix path
removed from AnthropicTransport.normalize_response, + all 5 call
sites in run_agent.py and auxiliary_client.py)
- user-agent: claude-cli/<v> (external, cli) and x-app: cli headers on
the Messages API client
Added:
- OAuth path strips context-1m-2025-08-07 — Anthropic rejects OAuth
requests carrying it with HTTP 400 'This authentication style is
incompatible with the long context beta header.'
Kept (auth plumbing, not identity spoofing):
- _is_oauth_token classifier and is_oauth flag threading
- Bearer vs x-api-key auth routing
- _OAUTH_ONLY_BETAS (claude-code-20250219, oauth-2025-04-20) — backend
requires these on the OAuth-gated Messages endpoint
- _OAUTH_CLIENT_ID (Claude Code's) — Anthropic doesn't issue OAuth
creds to third parties; this is the only way the login flow works
- claude-cli/<v> User-Agent on the OAuth token exchange + refresh
endpoints at platform.claude.com/v1/oauth/token — bare requests get
Cloudflare 1010 blocked
Verified live against api.anthropic.com with a fresh sk-ant-oat01-*
token:
- claude-haiku-4-5 simple message: HTTP 200, 'OK' response
- claude-haiku-4-5 tool call: HTTP 200, stop_reason=tool_use, tool
named 'terminal' (no mcp_ prefix) round-tripped correctly
- Outgoing wire: no user-agent, no x-app, real Hermes identity in
system prompt, real tool name in schema
Closes/supersedes #16820 (mcp_ PascalCase normalization patch — no longer
needed since the mcp_ round-trip is gone).
* fix(anthropic): resolve_anthropic_token() reads credential pool first
Close the gap where ~/.hermes/auth.json → credential_pool.anthropic
(where hermes login + dashboard PKCE flow write OAuth tokens) was not
in resolve_anthropic_token()'s source list.
Before: users who authed via hermes login got the token written into
the pool, but legacy fallback code paths (auxiliary_client, models
catalog fetch, explicit-runtime path) that call resolve_anthropic_token()
saw None and raised 'No Anthropic credentials found' — even though the
token was sitting in auth.json.
New priority 1: pool.select() with env-sourced entries skipped. Skipping
env:* entries preserves the existing env-var priority logic further
down the chain (static env OAuth → refreshable Claude Code upgrade via
_prefer_refreshable_claude_code_token).
Surfaced while writing the hermes-agent-dev skill playbook for
'finding a live OAuth token for an E2E test'.
---------
Co-authored-by: teknium1 <teknium@users.noreply.github.com>
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.
On AWS Bedrock (and Azure AI Foundry), Claude Opus 4.6/4.7 and Sonnet 4.6
are capped at 200K context unless the request carries the
`context-1m-2025-08-07` beta header. On native Anthropic (api.anthropic.com)
1M went GA so the header is a harmless no-op, but Bedrock/Azure still gate
it as beta as of 2026-04.
Hermes was advertising 1M in model_metadata.py (`claude-opus-4-7: 1000000`)
while silently sending a request without the beta — so Bedrock users saw
a 200K ceiling with no error message, and no config knob unblocked it.
Claude Code sends this header by default, which is why the same Bedrock
credentials worked there.
- Add `context-1m-2025-08-07` to `_COMMON_BETAS` (alongside interleaved
thinking and fine-grained tool streaming).
- Strip it in `_common_betas_for_base_url` for MiniMax bearer-auth
endpoints — they host their own models, not Claude, so Anthropic beta
headers are irrelevant and could risk rejection.
- Attach `_COMMON_BETAS` as `default_headers` on the AnthropicBedrock
client. Previously that constructor passed no betas at all, so native
Anthropic had the 1M unlock via default_headers but Bedrock didn't.
- Fast-mode per-request `extra_headers` already rebuilds from
`_common_betas_for_base_url`, so it picks up the 1M beta automatically.
Reported by user 'Rodmar' on Discord: Bedrock Opus 4.7 stuck at 200K while
same credentials worked in Claude Code.
The AIAgent.flush_memories pre-compression save, the gateway
_flush_memories_for_session, and everything feeding them are
obsolete now that the background memory/skill review handles
persistent memory extraction.
Problems with flush_memories:
- Pre-dates the background review loop. It was the only memory-save
path when introduced; the background review now fires every 10 user
turns on CLI and gateway alike, which is far more frequent than
compression or session reset ever triggered flush.
- Blocking and synchronous. Pre-compression flush ran on the live agent
before compression, blocking the user-visible response.
- Cache-breaking. Flush built a temporary conversation prefix
(system prompt + memory-only tool list) that diverged from the live
conversation's cached prefix, invalidating prompt caching. The
gateway variant spawned a fresh AIAgent with its own clean prompt
for each finalized session — still cache-breaking, just in a
different process.
- Redundant. Background review runs in the live conversation's
session context, gets the same content, writes to the same memory
store, and doesn't break the cache. Everything flush_memories
claimed to preserve is already covered.
What this removes:
- AIAgent.flush_memories() method (~248 LOC in run_agent.py)
- Pre-compression flush call in _compress_context
- flush_memories call sites in cli.py (/new + exit)
- GatewayRunner._flush_memories_for_session + _async_flush_memories
(and the 3 call sites: session expiry watcher, /new, /resume)
- 'flush_memories' entry from DEFAULT_CONFIG auxiliary tasks,
hermes tools UI task list, auxiliary_client docstrings
- _memory_flush_min_turns config + init
- #15631's headroom-deduction math in
_check_compression_model_feasibility (headroom was only needed
because flush dragged the full main-agent system prompt along;
the compression summariser sends a single user-role prompt so
new_threshold = aux_context is safe again)
- The dedicated test files and assertions that exercised
flush-specific paths
What this renames (with read-time backcompat on sessions.json):
- SessionEntry.memory_flushed -> SessionEntry.expiry_finalized.
The session-expiry watcher still uses the flag to avoid re-running
finalize/eviction on the same expired session; the new name
reflects what it now actually gates. from_dict() reads
'expiry_finalized' first, falls back to the legacy 'memory_flushed'
key so existing sessions.json files upgrade seamlessly.
Supersedes #15631 and #15638.
Tested: 383 targeted tests pass across run_agent/, agent/, cli/,
and gateway/ session-boundary suites. No behavior regressions —
background memory review continues to handle persistent memory
extraction on both CLI and gateway.
Bedrock model IDs use dots as namespace separators (anthropic.claude-opus-4-7,
us.anthropic.claude-sonnet-4-5-v1:0), not version separators.
normalize_model_name() was unconditionally converting all dots to hyphens,
producing invalid IDs that Bedrock rejects with HTTP 400/404.
This affected both the main agent loop (partially mitigated by
_anthropic_preserve_dots in run_agent.py) and all auxiliary client calls
(compression, session_search, vision, etc.) which go through
_AnthropicCompletionsAdapter and never pass preserve_dots=True.
Fix: add _is_bedrock_model_id() to detect Bedrock namespace prefixes
(anthropic., us., eu., ap., jp., global.) and skip dot-to-hyphen
conversion for these IDs regardless of the preserve_dots flag.
Bug 3 — Stale OAuth token not detected in 'hermes model':
- _model_flow_anthropic used 'has_creds = bool(existing_key)' which treats
any non-empty token (including expired OAuth tokens) as valid.
- Added existing_is_stale_oauth check: if the only credential is an OAuth
token (sk-ant- prefix) with no valid cc_creds fallback, mark it stale
and force the re-auth menu instead of silently accepting a broken token.
Bug 4 — macOS Keychain credentials never read:
- Claude Code >=2.1.114 migrated from ~/.claude/.credentials.json to the
macOS Keychain under service 'Claude Code-credentials'.
- Added _read_claude_code_credentials_from_keychain() using the 'security'
CLI tool; read_claude_code_credentials() now tries Keychain first then
falls back to JSON file.
- Non-Darwin platforms return None from Keychain read immediately.
Tests:
- tests/agent/test_anthropic_keychain.py: 11 cases covering Darwin-only
guard, security command failures, JSON parsing, fallback priority.
- tests/hermes_cli/test_anthropic_model_flow_stale_oauth.py: 8 cases
covering stale OAuth detection, API key passthrough, cc_creds fallback.
Refs: #12905
FixesNousResearch/hermes-agent#9813
Root cause: _is_oauth_token() only recognized sk-ant-* and eyJ* patterns,
but Claude Code OAuth tokens from CLAUDE_CODE_OAUTH_TOKEN use cc- prefix
Fix: Add cc- prefix detection so these tokens route through Bearer auth
NormalizedResponse and ToolCall now have backward-compat properties
so the agent loop can read them directly without the shim:
ToolCall: .type, .function (returns self), .call_id, .response_item_id
NormalizedResponse: .reasoning_content, .reasoning_details,
.codex_reasoning_items
This eliminates the 35-line shim and its 4 call sites in run_agent.py.
Also changes flush_memories guard from hasattr(response, 'choices')
to self.api_mode in ('chat_completions', 'bedrock_converse') so it
works with raw boto3 dicts too.
WS1 items 3+4 of Cycle 2 (#14418).
3-layer chain (transport → v2 → v1) was collapsed to 2-layer in PR 7.
This collapses the remaining 2-layer (transport → v1 → NR mapping in
transport) to 1-layer: v1 now returns NormalizedResponse directly.
Before: adapter returns (SimpleNamespace, finish_reason) tuple,
transport unpacks and maps to NormalizedResponse (22 lines).
After: adapter returns NormalizedResponse, transport is a
1-line passthrough.
Also updates ToolCall construction — adapter now creates ToolCall
dataclass directly instead of SimpleNamespace(id, type, function).
WS1 item 1 of Cycle 2 (#14418).
Consolidate 4 per-transport lazy singleton helpers (_get_anthropic_transport,
_get_codex_transport, _get_chat_completions_transport, _get_bedrock_transport)
into one generic _get_transport(api_mode) with a shared dict cache.
Collapse the 65-line main normalize block (3 api_mode branches, each with
its own SimpleNamespace shim) into 7 lines: one _get_transport() call +
one _nr_to_assistant_message() shared shim. The shim extracts provider_data
fields (codex_reasoning_items, reasoning_details, call_id, response_item_id)
into the SimpleNamespace shape downstream code expects.
Wire chat_completions and bedrock_converse normalize through their transports
for the first time — these were previously falling into the raw
response.choices[0].message else branch.
Remove 8 dead codex adapter imports that have zero callers after PRs 1-6.
Transport lifecycle improvements:
- Eagerly warm transport cache at __init__ (surfaces import errors early)
- Invalidate transport cache on api_mode change (switch_model, fallback
activation, fallback restore, transport recovery) — prevents stale
transport after mid-session provider switch
run_agent.py: -32 net lines (11,988 -> 11,956).
PR 7 of the provider transport refactor.
Port from openclaw/openclaw#66664. The build_anthropic_kwargs call site
used 'max_tokens or _get_anthropic_max_output(model)', which correctly
falls back when max_tokens is 0 or None (falsy) but lets negative ints
(-1, -500), fractional floats (0.5, 8192.7), NaN, and infinity leak
through to the Anthropic API. Anthropic rejects these with HTTP 400
('max_tokens: must be greater than or equal to 1'), turning a local
config error into a surprise mid-conversation failure.
Add two resolver helpers matching OpenClaw's:
_resolve_positive_anthropic_max_tokens — returns int(value) only if
value is a finite positive number; excludes bools, strings, NaN,
infinity, sub-one positives (floor to 0).
_resolve_anthropic_messages_max_tokens — prefers a positive requested
value, else falls back to the model's output ceiling; raises
ValueError only if no positive budget can be resolved.
The context-window clamp at the call site (max_tokens > context_length)
is preserved unchanged — it handles oversized values; the new resolver
handles non-positive values. These concerns are now cleanly separated.
Tests: 17 new cases covering positive/zero/negative ints, fractional
floats (both >1 and <1), NaN, infinity, booleans, strings, None, and
integration via build_anthropic_kwargs.
Refs: openclaw/openclaw#66664
Follow-up to the cherry-picked PR #13897 fix. Three issues found:
1. CRITICAL: The thinking block synthesised from reasoning_content was
immediately stripped by the third-party signature management code
(Kimi is classified as _is_third_party_anthropic_endpoint). Added a
Kimi-specific carve-out that preserves unsigned thinking blocks while
still stripping Anthropic-signed blocks Kimi can't validate.
2. Empty-string reasoning_content was silently dropped because the
truthiness check ('if reasoning_content and ...') evaluates to False
for ''. Changed to 'isinstance(reasoning_content, str)' so the
tier-3 fallback from _copy_reasoning_content_for_api (which injects
'' for Kimi tool-call messages with no reasoning) actually produces
a thinking block.
3. The thinking block was appended AFTER tool_use blocks. Anthropic
protocol requires thinking -> text -> tool_use ordering. Changed to
blocks.insert(0, ...) to prepend.
FixesNousResearch/hermes-agent#13848
Kimi's /coding endpoint speaks the Anthropic Messages protocol but has its
own thinking semantics: when thinking is enabled, Kimi validates message
history and requires every prior assistant tool-call message to carry
OpenAI-style reasoning_content.
The Anthropic path never populated that field, and
convert_messages_to_anthropic strips all Anthropic thinking blocks on
third-party endpoints — so the request failed with HTTP 400:
"thinking is enabled but reasoning_content is missing in assistant
tool call message at index N"
Now, when an assistant message contains tool_calls and a
reasoning_content string, we append a {"type": "thinking", ...} block
to the Anthropic content so Kimi can validate the history. This only
affects assistant messages with tool_calls + reasoning_content; plain
text assistant messages are unchanged.
Kimi's /coding endpoint speaks the Anthropic Messages protocol but has
its own thinking semantics: when thinking.enabled is sent, Kimi validates
the history and requires every prior assistant tool-call message to carry
OpenAI-style reasoning_content. The Anthropic path never populates that
field, and convert_messages_to_anthropic strips Anthropic thinking blocks
on third-party endpoints — so after one tool-calling turn the next request
fails with:
HTTP 400: thinking is enabled but reasoning_content is missing in
assistant tool call message at index N
Kimi on chat_completions handles thinking via extra_body in
ChatCompletionsTransport (#13503). On the Anthropic route, drop the
parameter entirely and let Kimi drive reasoning server-side.
build_anthropic_kwargs now gates the reasoning_config -> thinking block
on not _is_kimi_coding_endpoint(base_url).
Tests: 8 new parametric tests cover /coding, /coding/v1, /coding/anthropic,
/coding/ (trailing slash), explicit disabled, other third-party endpoints
still getting thinking (MiniMax), native Anthropic unaffected, and the
non-/coding Kimi root route.
- Add _is_kimi_coding_endpoint() to detect Kimi coding API
- Place Kimi check BEFORE _requires_bearer_auth to ensure User-Agent header is set
- Without this header, Kimi returns 403 on /coding/v1/messages
- Fixes kimi-2.5, kimi-for-coding, kimi-k2.6-code-preview all returning 403
Add agent/transports/types.py with three shared dataclasses:
- NormalizedResponse: content, tool_calls, finish_reason, reasoning, usage, provider_data
- ToolCall: id, name, arguments, provider_data (per-tool-call protocol metadata)
- Usage: prompt_tokens, completion_tokens, total_tokens, cached_tokens
Add normalize_anthropic_response_v2() to anthropic_adapter.py — wraps the
existing v1 function and maps its output to NormalizedResponse. One call site
in run_agent.py (the main normalize branch) uses v2 with a back-compat shim
to SimpleNamespace for downstream code.
No ABC, no registry, no streaming, no client lifecycle. Those land in PR 3
with the first concrete transport (AnthropicTransport).
46 new tests:
- test_types.py: dataclass construction, build_tool_call, map_finish_reason
- test_anthropic_normalize_v2.py: v1-vs-v2 regression tests (text, tools,
thinking, mixed, stop reasons, mcp prefix stripping, edge cases)
Part of the provider transport refactor (PR 2 of 9).
Follow-up on top of mvanhorn's cherry-picked commit. Original PR only
wired request_timeout_seconds into the explicit-creds OpenAI branch at
run_agent.py init; router-based implicit auth, native Anthropic, and the
fallback chain were still hardcoded to SDK defaults.
- agent/anthropic_adapter.py: build_anthropic_client() accepts an optional
timeout kwarg (default 900s preserved when unset/invalid).
- run_agent.py: resolve per-provider/per-model timeout once at init; apply
to Anthropic native init + post-refresh rebuild + stale/interrupt
rebuilds + switch_model + _restore_primary_runtime + the OpenAI
implicit-auth path + _try_activate_fallback (with immediate client
rebuild so the first fallback request carries the configured timeout).
- tests: cover anthropic adapter kwarg honoring; widen mock signatures
to accept the new timeout kwarg.
- docs/example: clarify that the knob now applies to every transport,
the fallback chain, and rebuilds after credential rotation.
Regression from #11161 (Claude Opus 4.7 migration, commit 0517ac3e).
The Opus 4.7 migration changed `ADAPTIVE_EFFORT_MAP["xhigh"]` from "max"
(the pre-migration alias) to "xhigh" to preserve the new 4.7 effort level
as distinct from max. This is correct for 4.7, but Opus/Sonnet 4.6 only
expose 4 levels (low/medium/high/max) — sending "xhigh" there now 400s:
BadRequestError [HTTP 400]: This model does not support effort
level 'xhigh'. Supported levels: high, low, max, medium.
Users who set reasoning_effort=xhigh as their default (xhigh is the
recommended default for coding/agentic on 4.7 per the Anthropic migration
guide) now 400 every request the moment they switch back to a 4.6 model
via `/model` or config. Verified live against the Anthropic API on
`anthropic==0.94.0`.
Fix: make the mapping model-aware. Add `_supports_xhigh_effort()`
predicate (matches 4-7/4.7 substrings, mirroring the existing
`_supports_adaptive_thinking` / `_forbids_sampling_params` pattern).
On pre-4.7 adaptive models, downgrade xhigh→max (the strongest effort
those models accept, restoring pre-migration behavior). On 4.7+, keep
xhigh as a distinct level.
Per Anthropic's migration guide, xhigh is 4.7-only:
https://platform.claude.com/docs/en/about-claude/models/migration-guide
> Opus 4.7 effort levels: max, xhigh (new), high, medium, low.
> Opus 4.6 effort levels: max, high, medium, low.
SDK typing confirms: `anthropic.types.OutputConfigParam.effort: Literal[
"low", "medium", "high", "max"]` (v0.94.0 not yet updated for xhigh).
## Test plan
Verified live on macOS 15.5 / anthropic==0.94.0:
claude-opus-4-6 + effort=xhigh → output_config.effort=max → 200 OK
claude-opus-4-7 + effort=xhigh → output_config.effort=xhigh → 200 OK
claude-opus-4-6 + effort=max → output_config.effort=max → 200 OK
claude-opus-4-7 + effort=max → output_config.effort=max → 200 OK
`tests/agent/test_anthropic_adapter.py` — 120 pass (replaced 1 bugged
test that asserted the broken behavior, added 1 for 4.7 preservation).
Full adapter suite: 120 passed in 1.05s.
Broader suite (agent + run_agent + cli/gateway reasoning): 2140 passed
(2 pre-existing failures on clean upstream/main, unrelated).
## Platforms
Tested on macOS 15.5. No platform-specific code paths touched.
Claude Opus 4.7 introduced several breaking API changes that the current
codebase partially handled but not completely. This patch finishes the
migration per the official migration guide at
https://platform.claude.com/docs/en/about-claude/models/migration-guideFixesNousResearch/hermes-agent#11137
Breaking-change coverage:
1. Adaptive thinking + output_config.effort — 4.7 is now recognized by
_supports_adaptive_thinking() (extends previous 4.6-only gate).
2. Sampling parameter stripping — 4.7 returns 400 for any non-default
temperature / top_p / top_k. build_anthropic_kwargs drops them as a
safety net; the OpenAI-protocol auxiliary path (_build_call_kwargs)
and AnthropicCompletionsAdapter.create() both early-exit before
setting temperature for 4.7+ models. This keeps flush_memories and
structured-JSON aux paths that hardcode temperature from 400ing
when the aux model is flipped to 4.7.
3. thinking.display = "summarized" — 4.7 defaults display to "omitted",
which silently hides reasoning text from Hermes's CLI activity feed
during long tool runs. Restoring "summarized" preserves 4.6 UX.
4. Effort level mapping — xhigh now maps to xhigh (was xhigh→max, which
silently over-efforted every coding/agentic request). max is now a
distinct ceiling per Anthropic's 5-level effort model.
5. New stop_reason values — refusal and model_context_window_exceeded
were silently collapsed to "stop" (end_turn) by the adapter's
stop_reason_map. Now mapped to "content_filter" and "length"
respectively, matching upstream finish-reason handling already in
bedrock_adapter.
6. Model catalogs — claude-opus-4-7 added to the Anthropic provider
list, anthropic/claude-opus-4.7 added at top of OpenRouter fallback
catalog (recommended), claude-opus-4-7 added to model_metadata
DEFAULT_CONTEXT_LENGTHS (1M, matching 4.6 per migration guide).
7. Prefill docstrings — run_agent.AIAgent and BatchRunner now document
that Anthropic Sonnet/Opus 4.6+ reject a trailing assistant-role
prefill (400).
8. Tests — 4 new tests in test_anthropic_adapter covering display
default, xhigh preservation, max on 4.7, refusal / context-overflow
stop_reason mapping, plus the sampling-param predicate. test_model_metadata
accepts 4.7 at 1M context.
Tested on macOS 15.5 (darwin). 119 tests pass in
tests/agent/test_anthropic_adapter.py, 1320 pass in tests/agent/.
Aligns MiniMax provider with official API documentation. Fixes 6 bugs:
transport mismatch (openai_chat -> anthropic_messages), credential leak
in switch_model(), prompt caching sent to non-Anthropic endpoints,
dot-to-hyphen model name corruption, trajectory compressor URL routing,
and stale doctor health check.
Also corrects context window (204,800), thinking support (manual mode),
max output (131,072), and model catalog (M2 family only on /anthropic).
Source: https://platform.minimax.io/docs/api-reference/text-anthropic-api
Co-authored-by: kshitijk4poor <kshitijk4poor@users.noreply.github.com>
_is_oauth_token() returned True for any key not starting with 'sk-ant-api',
which means MiniMax and Alibaba API keys were falsely treated as Anthropic
OAuth tokens. This triggered the Claude Code compatibility path:
- All tool names prefixed with mcp_ (e.g. mcp_terminal, mcp_web_search)
- System prompt injected with 'You are Claude Code' identity
- 'Hermes Agent' replaced with 'Claude Code' throughout
Fix: Make _is_oauth_token() positively identify Anthropic OAuth tokens by
their key format instead of using a broad catch-all:
- sk-ant-* (but not sk-ant-api-*) -> setup tokens, managed keys
- eyJ* -> JWTs from Anthropic OAuth flow
- Everything else -> False (MiniMax, Alibaba, etc.)
Reported by stefan171.
Automated dead code audit using vulture + coverage.py + ast-grep intersection,
confirmed by Opus deep verification pass. Every symbol verified to have zero
production callers (test imports excluded from reachability analysis).
Removes ~1,534 lines of dead production code across 46 files and ~1,382 lines
of stale test code. 3 entire files deleted (agent/builtin_memory_provider.py,
hermes_cli/checklist.py, tests/hermes_cli/test_setup_model_selection.py).
Co-authored-by: alt-glitch <balyan.sid@gmail.com>
Extends the /fast command to support Anthropic's Fast Mode beta in addition
to OpenAI Priority Processing. When enabled on Claude Opus 4.6, adds
speed:"fast" and the fast-mode-2026-02-01 beta header to API requests for
~2.5x faster output token throughput.
Changes:
- hermes_cli/models.py: Add _ANTHROPIC_FAST_MODE_MODELS registry,
model_supports_fast_mode() now recognizes Claude Opus 4.6,
resolve_fast_mode_overrides() returns {speed: fast} for Anthropic
vs {service_tier: priority} for OpenAI
- agent/anthropic_adapter.py: Add _FAST_MODE_BETA constant,
build_anthropic_kwargs() accepts fast_mode=True which injects
speed:fast + beta header via extra_headers (skipped for third-party
Anthropic-compatible endpoints like MiniMax)
- run_agent.py: Pass fast_mode to build_anthropic_kwargs in the
anthropic_messages path of _build_api_kwargs()
- cli.py: Update _handle_fast_command with provider-aware messaging
(shows 'Anthropic Fast Mode' vs 'Priority Processing')
- hermes_cli/commands.py: Update /fast description to mention both
providers
- tests: 13 new tests covering Anthropic model detection, override
resolution, CLI availability, routing, adapter kwargs, and
third-party endpoint safety
