Auxiliary LLM calls routed through OpenRouter now carry the `session_id`
sticky routing key, so they pin to the same provider endpoint as the
conversation they belong to instead of routing independently.
`OpenRouterProfile.build_extra_body` sourced the key only from its explicit
argument. Auxiliary call sites — compression, title generation, vision,
web_extract, session_search, MoA slots — funnel through
`agent.auxiliary_client`, which has no session handle and never passes one.
Those calls sent NO sticky key at all, so OpenRouter fell back to hashing
the opening messages and each aux call could land on a different provider
than the conversation it served. Per OpenRouter's prompt-caching docs a
present `session_id` is used directly as the routing key and activates
stickiness on the first successful request rather than only after a cache
hit, so the missing key also delayed pinning for the main loop.
Resolve it from the ambient conversation contextvar first, explicit argument
as fallback — the same resolution the Nous Portal profile uses (f2f4df064d)
and the same one `nous_portal_tags` already used for the `conversation=` tag.
The ambient value is the session-lineage ROOT, so the key also stays stable
for installs that opt out of the default `compression.in_place: true` and
across delegate-subagent trees.
The xAI `x-grok-conv-id` cache-affinity header in `build_api_kwargs_extras`
had the identical gap and is fixed the same way; it remains model-gated to
`x-ai/grok-*` / `xai/grok-*`.
Fixes#70820. Credit to @webtecnica, who reported it and submitted #70883
first; that PR threaded `session_id` through `call_llm` to reach the same
goal. This takes the ambient-contextvar route instead because none of the
34 `call_llm` call sites currently pass a `session_id`, so the parameter
alone would not have changed any live request.
Tests: 2 cases in tests/providers/test_provider_profiles.py, both verified
to fail against the pre-fix behavior via a sabotage run. E2E-validated on a
real AIAgent driving the out-of-turn compaction path.
The /model interactive picker resolved a base_url from user credentials
but never passed it to ProviderProfile.fetch_models(), causing the
picker to always query the provider's hardcoded default endpoint
instead of the user's custom URL (e.g. a company litellm proxy).
- providers/base.py: add optional base_url parameter to fetch_models()
- hermes_cli/models.py: pass resolved base_url to fetch_models()
- Update all subclass overrides for signature compatibility
- Add 6 regression tests covering override, fallback, and integration
* fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models
Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over
OpenRouter ignore reasoning.effort and use adaptive thinking. #42991 correctly
stopped Hermes from sending a reasoning field to them (it 400s), but put nothing
in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter
path: the model always ran at its adaptive default (high) regardless of config.
OpenRouter honors the requested effort on the top-level verbosity field instead
(maps to Anthropic output_config.effort). Route the existing
reasoning_config[effort] there for these models while still never emitting a
reasoning field, preserving the #42991 fix. No new config arg — the value the
user already sets via agent.reasoning_effort now flows to verbosity.
- low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the
extended scale for Claude; verified live HTTP 200 + monotonic token spend).
- effort unset/none/disabled omits verbosity so the model keeps its default.
- native Anthropic transport already correct; unchanged.
Fixes#43432
* test(openrouter): cover real effort range (add minimal, frame max as passthrough)
Adversarial review noted the verbosity tests looped over 'max' — a value
parse_reasoning_effort can never produce — while omitting 'minimal', which it
can. Align the routing test with the real config range
(VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate
value-agnostic passthrough test that documents why xhigh/max must survive
verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the
extended scale for Claude).
* docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models
Document that reasoning_effort transparently maps to OpenRouter's verbosity
field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where
reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire-
only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
The previous fix (#42991) only omitted reasoning when it was being disabled.
But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with
thinking.type.disabled on EVERY tool-continuation turn even when reasoning is
enabled: chat_completions never replays signed thinking blocks, so the prior
assistant tool_call has no thinking, and OpenRouter resolves "reasoning
requested but history has none" by emitting thinking.type.disabled — which
these models reject. Result: first turn works, every turn after the first tool
call dies (HTTP 400, non-retryable).
OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the
model self-decides), so the reasoning field is pointless for them on every turn
and harmful on tool-replay turns. Omit it entirely → adaptive default.
- openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic
models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models
unchanged.
- tests: assert omission across enabled/disabled/effort variants; parity tests
switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no
longer carries a reasoning field.
Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with
reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
New Anthropic models without a recognized version substring (claude-fable-5
and future named/numbered releases) were classified as legacy and routed down
the manual-thinking path, which made OpenRouter emit thinking.type.disabled —
a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400.
Invert the brittle version-substring allowlists to default-to-modern (mirroring
_get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/
no-sampling contract, with an explicit legacy list for older families. Non-Claude
Anthropic-Messages models (minimax, qwen3, …) keep the manual path.
- anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort /
_forbids_sampling_params now default unknown Claude models to modern; legacy
families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS.
- openrouter profile: omit reasoning entirely (→ adaptive default) instead of
forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy
Anthropic + all non-Anthropic models still pass the disable form through.
- model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out).
Tests assert the invariant ("unknown Claude model -> modern contract; legacy
stays manual; non-Claude unaffected"), not specific model names.
OpenRouter supports a session_id field in extra_body that pins
multi-turn conversations to the same provider endpoint, enabling
prompt cache reuse across turns. The session_id was already threaded
through to build_extra_body() but never included in the returned dict.
Co-Authored-By: Claude Opus 4 (1M context) <noreply@anthropic.com>
Pick openrouter/pareto-code as your model and OpenRouter auto-routes each
request to the cheapest model meeting your coding-quality bar (ranked by
Artificial Analysis). The new openrouter.min_coding_score config key (0.0-1.0,
default 0.65) tunes the floor.
- hermes_cli/models.py: add openrouter/pareto-code to OPENROUTER_MODELS so
it shows up in the picker with a description
- hermes_cli/config.py: add openrouter.min_coding_score (default 0.65 — lands
on a mid-tier coder on the current Pareto frontier)
- plugins/model-providers/openrouter: emit extra_body.plugins =
[{id: pareto-router, min_coding_score: X}] when model is openrouter/pareto-code
AND the score is a valid float in [0.0, 1.0]
- agent/transports/chat_completions.py: same emission on the legacy flag
path (when no provider profile is loaded)
- run_agent.py: openrouter_min_coding_score kwarg + storage; plumbed into
both build_kwargs() invocations and the context-summary extra_body path
- cli.py: read openrouter.min_coding_score once at init, validate float in
[0,1], pass to AIAgent constructions (CLI + background-task paths)
- cron/scheduler.py, batch_runner.py, tools/delegate_tool.py,
tui_gateway/server.py: propagate the kwarg (mirrors providers_order
plumbing — subagents inherit, cron/batch read from config)
- tests: profile-level + transport-level coverage of the model gating,
unset/empty/out-of-range handling, and the legacy flag path
- docs: new 'OpenRouter Pareto Code Router' section in providers.md
Verified end-to-end against api.openrouter.ai: at score=0.65 we land on a
mid-tier coder, at omission we get the strongest. Score is silently dropped
on any model other than openrouter/pareto-code, so it's safe to leave set.
Pass session_id through to provider profile build_api_kwargs_extras so
the OpenRouter profile can attach an xAI cache-affinity header
(x-grok-conv-id: <session-id>) for x-ai/grok-* models. xAI prompt
cache requires server affinity via this header — without it the cache
is poisoned and Grok prompt-cache hit rates drop dramatically on
multi-turn sessions.
Carve-out of #22708 by Ninso112. The original PR bundled a /diff
slash command, a zsh completion fix (already on main via #22802),
and holographic memory null-guards. This salvage keeps just the
Grok header work — small, targeted, and well-tested. Other
contributors and changes preserved for separate review.
Closes#22705.
Every provider profile is now a self-contained plugin under
plugins/model-providers/<name>/, mirroring the plugins/platforms/
pattern established for IRC and Teams. The ProviderProfile ABC
stays in providers/; the per-provider profile data moves out.
- plugins/model-providers/<name>/__init__.py calls register_provider()
- plugins/model-providers/<name>/plugin.yaml declares kind: model-provider
- providers/__init__.py._discover_providers() lazily scans bundled plugins
then $HERMES_HOME/plugins/model-providers/<name>/ (user override path)
- User plugins with the same name override bundled ones (last-writer-wins
in register_provider)
- Legacy providers/<name>.py layout still supported for back-compat with
out-of-tree editable installs
- Hermes PluginManager: new kind=model-provider; skipped like memory
plugins (providers/ discovery owns them); standalone plugins with
register_provider+ProviderProfile in their __init__.py auto-coerce to
this kind (same heuristic as memory providers)
- skip_names extended to include 'model-providers' so the general
PluginManager doesn't double-scan the category
- 4 new tests in tests/providers/test_plugin_discovery.py covering
bundled discovery, user override, and general-loader isolation
- Docs updated: website/docs/developer-guide/adding-providers.md,
provider-runtime.md, providers/README.md, plugins/model-providers/README.md
No API break: auth.py / config.py / doctor.py / models.py / runtime_provider.py /
model_metadata.py / auxiliary_client.py / chat_completions.py / run_agent.py
all still consume providers via get_provider_profile() / list_providers() —
they just now see plugin-discovered entries instead of pkgutil-iterated ones.
Third parties can now drop a single directory into
~/.hermes/plugins/model-providers/<name>/ to add or override an inference
provider without touching the repo.
