Move provider adapters (anthropic, bedrock, azure), platform adapters
(telegram, slack, discord, feishu, dingtalk, matrix), and terminal backends
(modal, daytona) out of core into plugins/ workspace members. Core references
them via the plugin registries (get_provider_namespace / get_provider_service /
get_tool_provider / get_credential_pool_hook) instead of direct imports.
- Provider/platform/terminal adapters relocated under plugins/; pyproject
extras reference workspace members; nix variants aggregate per-platform extras.
- Anthropic credential discovery + OAuth-masquerade guard live in the plugin's
credential_pool_hook; browser-open guarded by _can_open_graphical_browser.
- Vercel AI Gateway + Vercel Sandbox removed (upstream deletion); get_bedrock_model_ids
removed (replaced by bedrock_model_ids_or_none + discover_bedrock_models).
- Terminal backends resolve ModalEnvironment / DaytonaEnvironment lazily from
the plugin registry.
- uv.lock regenerated against the pluginified workspace.
Plugin test suites updated for the relocation: imports point at
hermes_agent_<plat>.adapter, caplog logger-name filters and monkeypatch targets
use the new module paths, and credential/rollback tests patch
registries.get_provider_service rather than the removed agent.*_adapter modules.
Verified: zero dead imports of relocated modules in core (import smoke test +
rename-map grep); nix develop succeeds; targeted plugin suites green
(bedrock, anthropic-auxiliary, matrix, dingtalk, feishu, credential_pool,
switch_model_rollback). Remaining full-suite failures are pre-existing on the
pre-merge tree (telegram setUpModule __code__) or environmental (voice/media/
PTY/network-dependent), not introduced here.
The opencode-go relay defaults max_tokens to 262144 when none is sent,
but Xiami mimo-v2.5-pro only supports 131072 completion tokens — every
request 400s with "max_tokens is too large: 262144" before the agent
can do anything.
Add a get_max_tokens(model) hook on ProviderProfile (default returns
default_max_tokens) so profiles fronting multiple upstreams can vary
the cap per-model. Wire chat_completions transport through the hook.
Override on OpenCodeGoProfile with mimo-v2.5-pro=131072.
Only mimo-v2.5-pro is capped — other opencode-go models (kimi, glm,
qwen, minimax, other mimo variants) unchanged.
The Kimi K2 branch added in the prior commit only emitted extra_body.thinking
and dropped reasoning_effort entirely. KimiProfile (api.moonshot.ai/v1) sends
both fields, and OpenCode Go proxies to the same Moonshot backend. Mirror that
shape on the Go path so /reasoning effort actually reaches Kimi.
- low/medium/high pass through verbatim
- xhigh/max clamp to high (Moonshot's max supported value)
- minimal / unknown effort → omit reasoning_effort, keep thinking on
- disabled / no config → unchanged
- DeepSeek branch unchanged
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.
