336b11584c
Every model reasons and intends that reasoning to be replayed, but the chat-completions wire echoed reasoning_content only to DeepSeek/Kimi/MiMo/ Ollama (plus a hardcoded Portal stealth model) and reasoning_details only to OpenRouter; turn_context popped `reasoning` everywhere, so Ollama's /v1 (reads only `reasoning`) never saw any. - message_sanitization owns one route->carrier table (_REASONING_CARRIER_ROUTES): strict schemas (Mistral, Groq, Cerebras, SambaNova) get nothing, Fireworks reasoning_content only, OpenRouter-format gateways (OpenRouter, Kilo, Vercel AI Gateway, Nous Portal) all three, Novita/MiniMax /v1/TokenHub + reasoning_details, local/Ollama/Upstage + reasoning, everyone else reasoning_content. - The " " pad stays limited to the must-echo tier (DeepSeek, Kimi, MiMo); Ollama leaves it (no echo requirement) and the stealth hardcode is gone. - Reactive recovery: a 4xx unknown-field rejection that NAMES a reasoning key this request carried is recorded per (provider, host, model) and the retry re-shapes api_messages without it. Generic upstream errors never count; a recorded key falls through, so no loop. History is untouched. - Nous Portal reads reasoning_details again: the #118182 "replay budget" does not exist in the Portal and does not reproduce live. - Gemini thought_signature survives on Gemini hosts/providers even when the model id is an alias. - Vendor keep-reasoning flags (docs-only): GLM thinking.clear_thinking=false, DashScope preserve_thinking / clear_thinking, Kimi K2.x thinking.keep=all. - stale_thinking_reaches_wire follows the same table, so compaction trigger and tail walks agree with what is sent.
Model Provider Plugins
Each subdirectory is a self-contained provider profile plugin. The
directory layout mirrors plugins/platforms/:
plugins/model-providers/
├── openrouter/
│ ├── __init__.py # registers the ProviderProfile
│ └── plugin.yaml # manifest: name, kind, version, description
├── anthropic/
│ ├── __init__.py
│ └── plugin.yaml
└── ...
How discovery works
providers/__init__.py._discover_providers() scans this directory (and
$HERMES_HOME/plugins/model-providers/) the first time anything calls
get_provider_profile() or list_providers(). Each __init__.py is
imported and expected to call providers.register_provider(profile).
User plugins at $HERMES_HOME/plugins/model-providers/<name>/ override
bundled plugins of the same name — last-writer-wins in
register_provider(). Drop a file there to replace a built-in.
Adding a new provider
-
Create
plugins/model-providers/<your_provider>/__init__.py:from providers import register_provider from providers.base import ProviderProfile my_provider = ProviderProfile( name="your-provider", aliases=("alias1", "alias2"), display_name="Your Provider", description="One-line description shown in the setup picker", signup_url="https://your-provider.example.com/keys", env_vars=("YOUR_PROVIDER_API_KEY", "YOUR_PROVIDER_BASE_URL"), base_url="https://api.your-provider.example.com/v1", default_aux_model="your-cheap-model", ) register_provider(my_provider) -
Create
plugins/model-providers/<your_provider>/plugin.yaml:name: your-provider-profile kind: model-provider version: 1.0.0 description: Short sentence about the provider author: Your Name
Nothing else needs to change. auth.py, config.py, models.py,
doctor.py, model_metadata.py, runtime_provider.py, and the
chat_completions transport all auto-wire from the registry.
Non-trivial profiles
Override the ProviderProfile hooks in a subclass for per-provider
quirks — see plugins/model-providers/openrouter/__init__.py for
build_extra_body and build_api_kwargs_extras examples, and
plugins/model-providers/gemini/__init__.py for thinking_config
translation.