hermes-agent/plugins/model-providers
Repository files (latest commit first)
Filename Latest commit message Latest commit date
kshitijk4poor bf87ae620e feat(models): add gpt-6.1-sol (OpenRouter + direct OpenAI API)
OpenAI released GPT-6.1 Sol on 2026-09-29. It is live on OpenRouter
(openai/gpt-6.1-sol, -pro) and the direct API; not yet on Nous Portal
or on Codex OAuth for our test account.

- Catalog: OPENROUTER_MODELS + openai-api picker; kept off the nous
  list via _OPENROUTER_ONLY until the Portal serves it. Manifest regenerated.
- Context: 1.05M direct API, 272K Codex fallback; 272K compaction
  auto-raise on the Codex route.
- Pricing: $2/$10, cached input $0.10 (5%, not 6 Sol's 10%), cache write
  $2.50, 2x/1.5x above 272K; -pro aliased to the base row.
- Reasoning: none/minimal return 400 on 6.1 Sol (unlike 6 Sol), so it uses
  Astra's low..max ladder without Astra's account gating. OpenRouter's
  catalog wrongly lists 'none' for it, so the OpenRouter profile omits a
  disable when the OpenAI ladder has no 'none'.
2026-09-29 23:25:26 +05:30
..
actual refactor(model-providers): thinking-toggle XOR effort translation lives in agent.reasoning_effort; opencode-free imports it instead of borrowing via sys.modules 2026-09-13 05:19:48 -07:00
ai-gateway refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
alibaba fix(providers): declare catalog-only aliases on their plugin profiles 2026-09-13 12:52:43 -07:00
alibaba-coding-plan refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
anthropic Port from OpenHands/OpenHands#16758: Anthropic model catalogs no longer stop at the first page 2026-09-13 21:07:35 -07:00
arcee refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
azure-foundry refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
bedrock refactor(model-providers): fetch_models stubs drop to the ABC; commandcode and AI Gateway catalog GETs go through open_credentialed_url 2026-09-13 05:19:48 -07:00
commandcode refactor(commandcode): resolve the native DeepSeek profile through the registry 2026-09-13 19:46:52 +05:30
copilot refactor(copilot): one GitHub effort clamp for the profile and main agent 2026-09-26 22:21:47 +05:30
copilot-acp fix(copilot-acp): honor the fetch_models None-on-failure contract; bound the probe budget; refresh-proof the memo 2026-09-13 19:30:48 +05:30
custom fix: send reasoning_effort=medium on custom endpoints when agent.reasoning_effort is unset 2026-09-20 16:03:38 -07:00
deepinfra fix: trim DeepInfra reasoning salvage to the invariant and document it 2026-09-15 18:22:40 -07:00
deepseek fix(aux): structured-output rejection no longer kills fallback candidates or costs a doomed first request 2026-09-17 09:11:38 -07:00
fireworks refactor: make install-stamp.json the single runtime version identity 2026-09-23 11:41:01 -04:00
gemini refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
gmi refactor: make install-stamp.json the single runtime version identity 2026-09-23 11:41:01 -04:00
huggingface refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
kilocode refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
kimi-coding Merge remote-tracking branch 'origin/main' into ethie/pm-clean 2026-09-24 13:10:14 -04:00
meta-ai review-fix(comments): restore lost #NNNN rationale comments across non-test source (mechanical sweep, condensed, code unchanged) 2026-09-03 09:44:26 -07:00
minimax fix(minimax): register minimax-portal and minimax-global aliases for minimax-oauth (#107928) 2026-09-13 12:52:43 -07:00
nebius-token-factory refactor(model-providers): tighten router/openrouter bodies, drop __future__ annotations (py>=3.11) 2026-09-02 21:58:26 -07:00
nous fix(providers): stop forwarding provider_routing prefs to Nous Portal 2026-09-06 02:42:59 -07:00
novita refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
nvidia fix(providers): declare catalog-only aliases on their plugin profiles 2026-09-13 12:52:43 -07:00
ollama-cloud refactor(model-providers): tighten router/openrouter bodies, drop __future__ annotations (py>=3.11) 2026-09-02 21:58:26 -07:00
openai-codex refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
opencode-zen refactor: make install-stamp.json the single runtime version identity 2026-09-23 11:41:01 -04:00
openrouter feat(models): add gpt-6.1-sol (OpenRouter + direct OpenAI API) 2026-09-29 23:25:26 +05:30
qwen-oauth refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
router refactor: make install-stamp.json the single runtime version identity 2026-09-23 11:41:01 -04:00
stepfun refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
upstage fix(upstage): default new Solar models to the 512K context window 2026-09-17 17:10:58 +05:30
vertex fix(providers): declare catalog-only aliases on their plugin profiles 2026-09-13 12:52:43 -07:00
xai refactor: make install-stamp.json the single runtime version identity 2026-09-23 11:41:01 -04:00
xiaomi refactor(model-providers): pack keyword-only profile constructors (comments and multi-line values keep their lines) 2026-09-02 21:45:13 -07:00
zai refactor(model-providers): tighten router/openrouter bodies, drop __future__ annotations (py>=3.11) 2026-09-02 21:58:26 -07:00
README.md feat(providers): make all 33 providers pluggable under plugins/model-providers/ 2026-05-05 13:40:01 -07:00

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

  1. 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)
    
  2. 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.