| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724 |
- """Static provider/model catalog tables (data only — no network).
- Curated per-provider model lists, the canonical provider registry, display groups and alias
- maps. Split out of ``hermes_cli.models``.
- """
- from __future__ import annotations
- from typing import Any, NamedTuple
- class CuratedFallbackModels(list[str]):
- """A curated list served because the provider's live catalog was unavailable. The disk cache
- treats it as a placeholder, never as the account's real catalog (#107391)."""
- # Fallback OpenRouter snapshot used when the live catalog is unavailable, as
- # ``(model_id, description shown in menus)``. ``:free`` SKUs are described "free".
- _OPENROUTER_DESCRIPTIONS = {
- "anthropic/claude-opus-5-fast": "2x price, higher output speed",
- "anthropic/claude-opus-4.8-fast": "2x price, higher output speed",
- "deepseek/deepseek-v4-pro-0813": "dated snapshot of v4-pro",
- "deepseek/deepseek-v4.1-flash-1017": "dated snapshot of v4.1-flash",
- "deepseek/deepseek-v4-flash-0731": "dated snapshot of v4-flash",
- "moonshotai/kimi-k3": "recommended",
- "z-ai/glm-5.3": "default",
- "z-ai/glm-5.3-flashx": "high-speed tier of glm-5.3-flash",
- "openrouter/pareto-code": "auto-routes to cheapest coder meeting openrouter.min_coding_score",
- "openai/gpt-6-astra-fast": "2x price, priority tier",
- "openai/gpt-6-astra-flex": "0.5x price, flex tier",
- "openai/gpt-6-astra-pro-fast": "2x price, priority tier",
- "openai/gpt-6-astra-pro-flex": "0.5x price, flex tier",
- "stealth/union-alpha": "free, stealth model",
- "stealth/space-bunny-alpha": "free, stealth model",
- }
- OPENROUTER_MODELS: list[tuple[str, str]] = [
- (mid, _OPENROUTER_DESCRIPTIONS.get(mid, "free" if mid.endswith(":free") else ""))
- for mid in (
- "anthropic/claude-fable-5.1", "anthropic/claude-fable-5", "anthropic/claude-opus-5.5",
- "anthropic/claude-opus-5", "anthropic/claude-opus-5-fast", "anthropic/claude-opus-4.8", "anthropic/claude-opus-4.8-fast",
- "anthropic/claude-sonnet-5.5", "anthropic/claude-sonnet-5", "anthropic/claude-haiku-5.5", "anthropic/claude-haiku-4.5",
- "openai/gpt-6-astra", "openai/gpt-6-astra-fast", "openai/gpt-6-astra-flex", "openai/gpt-6-astra-pro",
- "openai/gpt-6-astra-pro-fast", "openai/gpt-6-astra-pro-flex",
- "openai/gpt-6.1-sol", "openai/gpt-6.1-sol-pro", "openai/gpt-6.1-terra", "openai/gpt-6-sol", "openai/gpt-6-sol-pro",
- "openai/gpt-6-luna", "openai/gpt-6-luna-pro",
- "openai/gpt-5.5", "openai/gpt-5.5-pro", "openai/gpt-5.4-mini", "google/gemini-4-argon",
- "google/gemini-3.1-pro-preview",
- "google/gemini-3.8-flash", "google/gemini-3.7-flash", "x-ai/grok-4.7", "x-ai/grok-4.6",
- "deepseek/deepseek-v4-pro",
- "deepseek/deepseek-v4-pro-0813", "deepseek/deepseek-v4.1-flash", "deepseek/deepseek-v4.1-flash-1017",
- "deepseek/deepseek-v4-flash-0731",
- "qwen/qwen3.8-max-0902", "qwen/qwen3.8-flash", "moonshotai/kimi-k3", "minimax/minimax-m3", "z-ai/glm-5.3",
- "z-ai/glm-5.3-flash", "z-ai/glm-5.3-flashx", "z-ai/glm-5.2",
- "xiaomi/mimo-v2.6-pro", "xiaomi/mimo-v2.6-flash", "xiaomi/mimo-v2.6-pro-ultraspeed", "xiaomi/mimo-v2.5-pro", "tencent/hy4-preview",
- "tencent/hy3",
- "stepfun/step-3.7-flash", "nvidia/nemotron-3-super-120b-a12b", "meta/muse-spark-1.2",
- "meta/muse-spark-1.2-contributor", "meta/muse-spark-1.3", "meta/muse-spark-1.3-contributor", "sakana/fugu-ultra",
- "upstage/solar-pro4", "upstage/solar-mini4",
- "openrouter/pareto-code", "thinkingmachines/inkling:free", "thinkingmachines/inkling-small:free",
- "minimax/minimax-m3:free", "z-ai/glm-5.2:free", "poolside/laguna-s-2.1:free", "poolside/laguna-xs-2.1:free",
- "nvidia/nemotron-3-super-120b-a12b:free", "nvidia/nemotron-3-ultra-550b-a55b:free",
- "inclusionai/ling-3.0-flash:free", "nvidia/nemotron-3.5-lightning:free", "stealth/union-alpha",
- "stealth/space-bunny-alpha",
- )
- ]
- # OpenRouter entries the Nous Portal does not carry (routing/fast variants, free tier —
- # ``stealth/union-alpha`` is a $0 stealth SKU without the ``:free`` suffix).
- _OPENROUTER_ONLY = {
- "anthropic/claude-opus-5-fast", "anthropic/claude-opus-4.8-fast",
- "meta/muse-spark-1.2-contributor", "meta/muse-spark-1.3-contributor", "openrouter/pareto-code",
- "stealth/union-alpha",
- }
- # Fallback Vercel AI Gateway snapshot (open-weight first, then closed-source by family). Slugs
- # match Vercel's /v1/models catalog (``alibaba/`` for Qwen, ``zai/`` and ``xai/`` without hyphens).
- VERCEL_AI_GATEWAY_MODELS: list[tuple[str, str]] = [("moonshotai/kimi-k2.6", "recommended")] + [
- (mid, "") for mid in (
- "alibaba/qwen3.6-plus", "zai/glm-5.1", "minimax/minimax-m2.7", "anthropic/claude-sonnet-4.6",
- "anthropic/claude-opus-4.7", "anthropic/claude-opus-4.6", "anthropic/claude-haiku-4.5",
- "openai/gpt-5.4", "openai/gpt-5.4-mini", "openai/gpt-5.3-codex", "google/gemini-3.1-pro-preview",
- "google/gemini-3-flash", "google/gemini-3.1-flash-lite-preview", "xai/grok-4.20-reasoning",
- )
- ]
- def _codex_curated_models() -> list[str]:
- """openai-codex curated list from codex_models.py (DEFAULT_CODEX_MODELS + forward-compat
- synthesis) so the gateway /model picker and the CLI ``hermes model`` flow share one source."""
- from hermes_cli.codex_models import DEFAULT_CODEX_MODELS, _finalize_codex_models
- return _finalize_codex_models(list(DEFAULT_CODEX_MODELS))
- # Static xAI fallback when the models.dev disk cache is empty (fresh install, offline first run).
- # Mirrors the xAI-direct IDs from $HERMES_HOME/models_dev_cache.json; the cache overrides it on the
- # next refresh. Models xAI retired on 2026-05-15 (grok-4*, grok-4-fast*, grok-4-1-fast*,
- # grok-code-fast-1) are excluded — see docs.x.ai/developers/migration/may-15-retirement.
- _XAI_STATIC_FALLBACK: list[str] = [
- "grok-4.6", "grok-build-0.1", "grok-4.5", "grok-4.3", "grok-4.20-0309-reasoning",
- "grok-4.20-0309-non-reasoning", "grok-4.20-multi-agent-0309",
- ]
- # Callable via xAI OAuth but omitted from models.dev and /v1/models listings. grok-4.6 / grok-4.5
- # stay here until the models.dev disk cache refreshes.
- _XAI_CURATED_EXTRAS: list[str] = ["grok-4.6", "grok-4.5", "grok-composer-2.5-fast"]
- _XAI_TOP_MODEL = "grok-4.6"
- def _xai_promote_top(ids: list[str]) -> list[str]:
- """Pin the headline xAI model to the top of the curated list."""
- if _XAI_TOP_MODEL in ids:
- return [_XAI_TOP_MODEL] + [m for m in ids if m != _XAI_TOP_MODEL]
- return ids
- def _xai_merge_curated_extras(ids: list[str]) -> list[str]:
- """Append Hermes-curated xAI models missing from models.dev, right after the pinned headline."""
- out = list(ids)
- for extra in _XAI_CURATED_EXTRAS:
- if extra not in out:
- out.insert(1 if out and out[0] == _XAI_TOP_MODEL else len(out), extra)
- return out
- def _xai_finalize_catalog(ids: list[str]) -> list[str]:
- return _xai_promote_top(_xai_merge_curated_extras(ids))
- def _xai_curated_models() -> list[str]:
- """Offline curated floor for xAI / xAI OAuth pickers: $HERMES_HOME/models_dev_cache.json
- (no network), else ``_XAI_STATIC_FALLBACK``. Any failure falls through to the static list.
- Reads through the process-wide models.dev copy, so the parse is the one every later reader shares."""
- try:
- from agent.models_dev import fetch_models_dev
- data = fetch_models_dev(allow_network=False)
- xai = data.get("xai") if isinstance(data, dict) else None
- models = xai.get("models") if isinstance(xai, dict) else None
- if isinstance(models, dict) and models:
- ids = [mid for mid in models if isinstance(mid, str)]
- if ids:
- return _xai_finalize_catalog(sorted(ids))
- except Exception:
- pass
- return _xai_finalize_catalog(list(_XAI_STATIC_FALLBACK))
- _XAI_ROWS = ("xai", "xai-oauth")
- class _CuratedCatalog(dict):
- """``_PROVIDER_MODELS``: static rows, except xAI's, which come from the models.dev disk cache.
- That cache is a full parse of the whole catalog (tens of MiB resident), so it is read on the first
- lookup of an xAI row or a whole-table walk, never at import: a process that never asks for xAI
- never pays for it. Until then the xAI rows hold the static fallback, and they are filled in place,
- so a list handed out earlier sees the same ids. Overriding ``__iter__`` keeps ``dict(table)`` and
- ``{**table}`` on the slow path that reads values through ``__getitem__``."""
- _xai_ready = False
- def _ensure_xai(self) -> None:
- if self._xai_ready:
- return
- ids = _xai_curated_models()
- for row in _XAI_ROWS:
- current = dict.get(self, row)
- if isinstance(current, list):
- current[:] = ids
- self._xai_ready = True
- def __iter__(self):
- return dict.__iter__(self)
- def __getitem__(self, key):
- if key in _XAI_ROWS:
- self._ensure_xai()
- return dict.__getitem__(self, key)
- def get(self, key, default=None):
- if key in _XAI_ROWS:
- self._ensure_xai()
- return dict.get(self, key, default)
- def values(self):
- self._ensure_xai()
- return dict.values(self)
- def items(self):
- self._ensure_xai()
- return dict.items(self)
- def copy(self):
- self._ensure_xai()
- return dict(dict.items(self))
- # Native OpenAI Chat Completions (api.openai.com).
- _OPENAI_CHAT_MODELS = [
- "gpt-5.4", "gpt-5.4-mini", "gpt-5-mini", "gpt-5.3-codex", "gpt-5.2-codex", "gpt-4.1", "gpt-4o", "gpt-4o-mini",
- ]
- _MINIMAX_MODELS = ["MiniMax-M3", "MiniMax-M2.7", "MiniMax-M2.5", "MiniMax-M2.1", "MiniMax-M2"]
- _TENCENT_MODELS = ["hy4-preview", "hy3", "hy3-preview"]
- # Alibaba DashScope Coding platform (coding-intl): Qwen + third-party (GLM, Kimi, MiniMax, DeepSeek).
- # Classic DashScope keys should override DASHSCOPE_BASE_URL to
- # https://dashscope-intl.aliyuncs.com/compatible-mode/v1 (OpenAI-compat) or /apps/anthropic.
- _ALIBABA_MODELS = [
- "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash", "kimi-k2.5",
- "qwen3.5-plus", "qwen3-coder-plus", "qwen3-coder-next", "glm-5.2", "glm-5", "glm-4.7",
- "deepseek-v4-pro", "deepseek-v4-flash-0731", "MiniMax-M2.5",
- ]
- _ALIBABA_CODING_PLAN_MODELS = [
- "qwen3.7-plus", "qwen3.6-plus", "qwen3.5-plus", "qwen3-max-2026-01-23", "qwen3-coder-plus",
- "qwen3-coder-next", "kimi-k2.5", "glm-5", "glm-4.7", "MiniMax-M2.5",
- ]
- # Verified against a live Token Plan subscription (key tier ``sk-sp-...``).
- _ALIBABA_TOKEN_PLAN_MODELS = [
- "qwen3.8-max-0902", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash", "deepseek-v4-pro",
- "deepseek-v4-flash", "deepseek-v3.2", "kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "glm-5.2", "glm-5.1", "glm-5",
- ]
- # Curated per-provider lists. ``-cn`` twins share the international catalog on a domestic endpoint.
- _PROVIDER_MODELS: dict[str, list[str]] = _CuratedCatalog({
- "moa": ["default"],
- "nous": [mid for mid, _ in OPENROUTER_MODELS if mid not in _OPENROUTER_ONLY and not mid.endswith(":free")],
- # Used by /model counts and provider_model_ids fallback when /v1/models is unavailable.
- "openai": list(_OPENAI_CHAT_MODELS),
- "openai-api": [
- "gpt-6.1-sol", "gpt-6.1-sol-pro", "gpt-6-sol", "gpt-6-sol-pro", "gpt-6-luna", "gpt-6-luna-pro",
- "gpt-5.6-sol", "gpt-5.6-sol-pro", "gpt-5.6-terra", "gpt-5.6-terra-pro", "gpt-5.6-luna",
- "gpt-5.6-luna-pro", "gpt-5.5", "gpt-5.5-pro", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-nano",
- "gpt-5-mini", "gpt-5.3-codex", "gpt-4.1", "gpt-4o", "gpt-4o-mini",
- ],
- "openai-codex": _codex_curated_models(),
- "xai-oauth": _xai_finalize_catalog(list(_XAI_STATIC_FALLBACK)),
- "gemini": [
- # Argon second: [0] is the silent default when no model is configured, and a flagship never is.
- "gemini-3.8-flash", "gemini-4-argon", "gemini-3.7-flash",
- "gemini-3.1-pro-preview", "gemini-3-pro-preview", "gemini-3.6-flash", "gemini-3.1-flash-lite-preview",
- ],
- "zai": [
- "glm-5.3", "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5", "glm-5v-turbo", "glm-5-turbo",
- "glm-4.7", "glm-4.5", "glm-4.5-flash",
- ],
- "xai": _xai_finalize_catalog(list(_XAI_STATIC_FALLBACK)),
- # Nemotron flagships, then third-party agentic models hosted on build.nvidia.com.
- "nvidia": [
- "nvidia/nemotron-3.5-super-vl-120b-a12b", "nvidia/nemotron-3-ultra-550b-a55b", "nvidia/nemotron-3-super-120b-a12b",
- "nvidia/nemotron-3.5-lightning-30b-a3b", "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
- "z-ai/glm-5.3", "z-ai/glm-5.2", "moonshotai/kimi-k2.6", "minimaxai/minimax-m3",
- ],
- "kimi-coding": [
- "kimi-k3", "kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "kimi-for-coding", "kimi-for-coding-highspeed",
- "kimi-k2-thinking", "kimi-k2-thinking-turbo", "kimi-k2-turbo-preview", "kimi-k2-0905-preview",
- ],
- "kimi-coding-cn": [
- "kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5",
- "kimi-k2-thinking", "kimi-k2-turbo-preview", "kimi-k2-0905-preview",
- ],
- "stepfun": ["step-3.7-flash", "step-3.5-flash", "step-3.5-flash-2603"],
- "moonshot": [
- "kimi-k3", "kimi-k2.6", "kimi-k2.5", "kimi-k2-thinking", "kimi-k2-turbo-preview", "kimi-k2-0905-preview",
- ],
- "minimax": list(_MINIMAX_MODELS),
- "minimax-oauth": ["MiniMax-M3", "MiniMax-M2.7", "MiniMax-M2.7-highspeed"],
- "minimax-cn": list(_MINIMAX_MODELS),
- "anthropic": [
- "claude-fable-5.1", "claude-fable-5", "claude-opus-5-5", "claude-opus-5", "claude-sonnet-5",
- "claude-opus-4-8", "claude-opus-4-7", "claude-opus-4-6",
- "claude-sonnet-4-6", "claude-opus-4-5-20251101", "claude-sonnet-4-5-20250929",
- "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-haiku-4-5-20251001",
- ],
- "deepseek": ["deepseek-flash", "deepseek-v4-pro"],
- "xiaomi": [
- "mimo-v2.6-pro", "mimo-v2.6-flash", "mimo-v2.6-pro-ultraspeed",
- "mimo-v2.5-pro", "mimo-v2.5", "mimo-v2-pro", "mimo-v2-omni", "mimo-v2-flash",
- ],
- "tencent-tokenhub": list(_TENCENT_MODELS),
- "tencent-tokenplan": list(_TENCENT_MODELS),
- "arcee": ["trinity-large-thinking", "trinity-large-preview", "trinity-mini"],
- "gmi": [
- "zai-org/GLM-5.1-FP8", "deepseek-ai/DeepSeek-V3.2", "moonshotai/Kimi-K2.5",
- "google/gemini-3.1-flash-lite-preview", "anthropic/claude-sonnet-5",
- "anthropic/claude-sonnet-4.6", "openai/gpt-5.4",
- ],
- # Synced against opencode.ai/docs/zen + live GET /zen/v1/models. Zen/Go are
- # _LIVE_FIRST_PICKER_PROVIDERS, so this is a discovery floor: live entries lead in the picker
- # and stale curated names never pollute the top. "x-preview-f-free" = "Ox Alpha" stealth model.
- "opencode-zen": [
- "x-preview-f-free", "kimi-k3", "kimi-k2.5", "kimi-k2.6", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna",
- "gpt-5.5", "gpt-5.5-pro", "gpt-5.4-pro", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-nano", "gpt-5.3-codex",
- "gpt-5.3-codex-spark", "gpt-5.2", "gpt-5.2-codex", "gpt-5.1", "gpt-5.1-codex", "gpt-5.1-codex-max",
- "gpt-5.1-codex-mini", "gpt-5", "gpt-5-codex", "gpt-5-nano", "claude-fable-5", "claude-opus-5",
- "claude-sonnet-5", "claude-opus-4-8", "claude-opus-4-7", "claude-opus-4-6", "claude-opus-4-5",
- "claude-sonnet-4-6", "claude-sonnet-4-5", "claude-sonnet-4", "claude-haiku-4-5", "gemini-3.8-flash",
- "gemini-3.7-flash", "gemini-3.6-flash", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.1-pro", "gemini-3-flash",
- "grok-4.6", "grok-4.5", "grok-build-0.1", "muse-spark-1.2", "minimax-m3", "minimax-m2.7", "minimax-m2.5",
- "glm-5.3", "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5", "kimi-k2.7-code", "deepseek-v4-pro",
- "deepseek-v4-flash", "qwen3.6-plus", "qwen3.5-plus", "big-pickle", "mimo-v2.5-free",
- "nemotron-3-ultra-free", "nemotron-3.5-lightning-free",
- "muse-spark-1.2-contributor-free", "muse-spark-1.3-contributor-free",
- ],
- # Synced against opencode.ai/docs/go + live GET /zen/go/v1/models. Known-delisted models are
- # REMOVED (the live-first merge would otherwise keep offering a model that 401s): "ox-alpha-free"
- # — the Go-subscription twin of Zen's Ox Alpha — was delisted 2026-09-09.
- "opencode-go": [
- "kimi-k3", "kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "gpt-5.6-luna", "grok-4.5", "glm-5.3",
- "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5", "mimo-v2.5-pro", "mimo-v2.5", "mimo-v2-pro",
- "mimo-v2-omni", "minimax-m3", "minimax-m2.7", "minimax-m2.5", "deepseek-v4-pro",
- "deepseek-v4-flash", "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus",
- "qwen3.5-plus", "hy3", "hy3-preview", "muse-spark-1.2-contributor", "muse-spark-1.3-contributor",
- ],
- "kilocode": [
- "anthropic/claude-opus-4.6", "anthropic/claude-sonnet-4.6", "openai/gpt-5.4",
- "google/gemini-3-pro-preview", "google/gemini-3-flash-preview",
- ],
- "alibaba": list(_ALIBABA_MODELS),
- "alibaba-cn": list(_ALIBABA_MODELS),
- "alibaba-coding-plan": list(_ALIBABA_CODING_PLAN_MODELS),
- "alibaba-coding-plan-cn": list(_ALIBABA_CODING_PLAN_MODELS),
- "alibaba-token-plan": list(_ALIBABA_TOKEN_PLAN_MODELS),
- "alibaba-token-plan-cn": list(_ALIBABA_TOKEN_PLAN_MODELS),
- # Only agentic HF models that map to OpenRouter defaults.
- "huggingface": [
- "moonshotai/Kimi-K2.5", "Qwen/Qwen3.5-397B-A17B", "Qwen/Qwen3.5-35B-A3B",
- "deepseek-ai/DeepSeek-V3.2", "MiniMaxAI/MiniMax-M2.5", "zai-org/GLM-5",
- "XiaomiMiMo/MiMo-V2-Flash", "moonshotai/Kimi-K2-Thinking", "moonshotai/Kimi-K2.6",
- ],
- # Static fallback when live discovery (ListFoundationModels + ListInferenceProfiles) is
- # unavailable. Inference-profile IDs (us.*) because most models require them.
- "bedrock": [
- # [0] is the provider default (get_default_model_for_provider) — keep the cheaper Sonnet there.
- "us.anthropic.claude-sonnet-5", "us.anthropic.claude-opus-5-5", "us.anthropic.claude-sonnet-4-6",
- "us.anthropic.claude-opus-4-6-v1",
- "us.anthropic.claude-haiku-4-5-20251001-v1:0", "us.anthropic.claude-sonnet-4-5-20250929-v1:0",
- "openai.gpt-6-astra", "openai.gpt-6.1-sol", "openai.gpt-6.1-terra", "openai.gpt-6-sol", "openai.gpt-6-luna",
- "openai.gpt-5.5", "openai.gpt-5.6-sol", "openai.gpt-5.6-terra", "openai.gpt-5.6-luna", "openai.gpt-5.4",
- "us.amazon.nova-pro-v1:0", "us.amazon.nova-lite-v1:0", "us.amazon.nova-micro-v1:0", "deepseek.v3.2",
- "us.meta.llama4-maverick-17b-instruct-v1:0", "us.meta.llama4-scout-17b-instruct-v1:0",
- ],
- # Azure Foundry models depend on the user's endpoint configuration.
- "azure-foundry": [],
- # Vertex's OpenAI-compatible endpoint has no /models route, so without this the /model picker
- # only shows the configured model. IDs carry the "google/" publisher prefix Vertex expects
- # (see hermes_cli/model_setup_flows.py); validated live against a GCP project (global region).
- "vertex": [
- "google/gemini-3.8-flash", "google/gemini-4-argon", "google/gemini-3.7-flash", # [0] = silent default
- "google/gemini-3.1-pro-preview", "google/gemini-3-pro-preview", "google/gemini-3.6-flash",
- "google/gemini-3.5-flash", "google/gemini-3.5-flash-lite", "google/gemini-3-flash-preview",
- "google/gemini-3.1-flash-lite-preview", "google/gemini-3.1-flash-lite",
- ],
- "novita": [
- "moonshotai/kimi-k2.5", "minimax/minimax-m2.7", "zai-org/glm-5", "deepseek/deepseek-v3-0324",
- "deepseek/deepseek-r1-0528", "qwen/qwen3-235b-a22b-fp8",
- ],
- # Bare ids derived from the picker snapshot so both stay in sync.
- "ai-gateway": [mid for mid, _ in VERCEL_AI_GATEWAY_MODELS],
- })
- # ---------------------------------------------------------------------------
- # Canonical provider list — single source of truth for provider identity. Every code path that
- # lists, displays, or iterates providers (hermes model, /model, list_authenticated_providers)
- # derives from it. slug = internal ID (config.yaml, --provider); label = short display name;
- # tui_desc = longer description for the `hermes model` picker.
- # ---------------------------------------------------------------------------
- class ProviderEntry(NamedTuple):
- slug: str
- label: str
- tui_desc: str
- CANONICAL_PROVIDERS: list[ProviderEntry] = [ProviderEntry(*row) for row in (
- ("nous", "Nous Portal", "Nous Portal (Everything your agent needs, 300+ models with bundled tool use)"),
- ("fireworks", "Fireworks AI", "Fireworks AI (OpenAI-compatible direct model API)"),
- ("openrouter", "OpenRouter", "OpenRouter (Pay-per-use API aggregator)"),
- ("moa", "Mixture of Agents", "Mixture of Agents (named presets; aggregator acts after reference models)"),
- ("novita", "NovitaAI", "NovitaAI (Cloud: Model API, Agent Sandbox, GPU Cloud)"),
- ("lmstudio", "LM Studio", "LM Studio (Local desktop app with built-in model server)"),
- ("anthropic", "Anthropic", "Anthropic (Claude models via API key or Claude Code)"),
- ("openai-codex", "ChatGPT or Codex Subscription", "ChatGPT or Codex Subscription (Sign in with your ChatGPT account, uses Codex models)"),
- ("openai-api", "OpenAI API", "OpenAI API (api.openai.com, API key)"),
- ("alibaba", "Qwen Cloud", "Qwen Cloud / DashScope (Qwen + multi-provider)"),
- ("xai-oauth", "xAI Grok OAuth (SuperGrok / Premium+)", "xAI Grok OAuth (SuperGrok / Premium+ subscription)"),
- ("xiaomi", "Xiaomi MiMo", "Xiaomi MiMo (MiMo-V2.5 and V2 models: pro, omni, flash)"),
- ("tencent-tokenhub", "Tencent TokenHub", "Tencent TokenHub (Hy4 preview via tokenhub.tencentmaas.com)"),
- ("tencent-tokenplan", "Tencent TokenPlan", "Tencent TokenPlan (Hy4 preview via api.lkeap.cloud.tencent.com, Anthropic Messages)"),
- ("nvidia", "NVIDIA NIM", "NVIDIA NIM (Nemotron models via build.nvidia.com or local NIM)"),
- ("huggingface", "Hugging Face", "Hugging Face Inference Providers"),
- ("gemini", "Google AI Studio", "Google AI Studio (Native Gemini API)"),
- ("vertex", "Google Vertex AI", "Google Vertex AI (Gemini via GCP; OAuth2 service account or ADC, GCP billing/quotas)"),
- ("deepseek", "DeepSeek", "DeepSeek (V3, R1, coder, direct API)"), ("xai", "xAI", "xAI Grok (Direct API)"),
- ("zai", "Z.AI / GLM", "Z.AI / GLM (Zhipu direct API)"),
- ("kimi-coding", "Kimi / Kimi Coding Plan", "Kimi Coding Plan (api.kimi.com & Moonshot API)"),
- ("kimi-coding-cn", "Kimi / Moonshot (China)", "Kimi / Moonshot China (Domestic direct API)"),
- ("stepfun", "StepFun Step Plan", "StepFun Step Plan (Agent / coding models via Step Plan API)"),
- ("minimax", "MiniMax", "MiniMax (Global direct API)"),
- ("minimax-oauth", "MiniMax (OAuth)", "MiniMax via OAuth browser login (Coding Plan, minimax.io)"),
- ("minimax-cn", "MiniMax (China)", "MiniMax China (Domestic direct API)"),
- ("ollama-cloud", "Ollama Cloud", "Ollama Cloud (Cloud-hosted open models, ollama.com)"),
- ("arcee", "Arcee AI", "Arcee AI (Trinity models, direct API)"),
- ("gmi", "GMI Cloud", "GMI Cloud (Multi-model direct API)"),
- ("kilocode", "Kilo Code", "Kilo Code (Kilo Gateway API)"),
- ("opencode-zen", "OpenCode Zen", "OpenCode Zen (Curated models, pay-as-you-go)"),
- ("opencode-go", "OpenCode Go", "OpenCode Go (Open models subscription)"),
- ("bedrock", "AWS Bedrock", "AWS Bedrock (Claude, Nova, Llama, DeepSeek; IAM or API key)"),
- ("azure-foundry", "Azure Foundry", "Azure Foundry (OpenAI-style or Anthropic-style endpoint, your Azure AI deployment)"),
- ("ai-gateway", "Vercel AI Gateway", "Vercel AI Gateway (Multi-model aggregator)"),
- ("qwen-oauth", "Qwen OAuth (Portal)", "Qwen OAuth (Reuses local Qwen CLI login)"),
- )]
- # Auto-extend CANONICAL_PROVIDERS with providers registered under plugins/model-providers/<name>/
- # so a new provider reaches the picker, /model and every downstream consumer without edits here.
- # Admission is by slug only: every in-tree non-api-key profile (OAuth, external-process, cloud
- # SDK) already owns a hand-written row above, so the old auth_type skip set never excluded an
- # in-tree provider — it only hid out-of-tree plugins. Visibility is gated downstream by
- # credentials, not here: ``models._provider_has_credentials`` / ``_lap_canonical_rows`` route
- # through ``auth.get_auth_status`` (external_process → the binary resolves; OAuth → auth.json /
- # credential-pool entry), so an admitted row reads authenticated=False until the user signs in.
- _canonical_slugs = {p.slug for p in CANONICAL_PROVIDERS}
- _BUILTIN_SLUGS = frozenset(_canonical_slugs)
- def _plugin_provider_enters_picker(pp) -> bool:
- """Picker admission for a plugin model-provider profile: any slug without a built-in row.
- Every profile is admitted, hidden (pre-release) ones included: ``CANONICAL_PROVIDERS`` is also
- the typed-path registry (``provider:model`` parsing, labels), so a signed-in user must still be
- able to name one. The pickers that list rows filter with ``providers.provider_listed``."""
- return pp.name not in _canonical_slugs
- def sync_plugin_provider_catalog() -> int:
- """Admit every registered plugin provider without a built-in row; return how many were added.
- Runs at import and again from ``providers._sync_auth_registry`` whenever a profile is registered
- after this module was imported. The import-time pass alone observes a *partial* registry: a
- plugin whose own imports pull ``hermes_cli.models`` in mid-``_discover_providers()``, or a
- profile registered later at runtime, would otherwise never reach the picker, ``hermes model``,
- ``/model`` or the Desktop ``model.options`` list until restart — the catalog twin of the auth
- registry window (#102123). Idempotent by slug; built-in rows are never rewritten. A row lands
- after its ``picker_after`` slug when that row exists, else at the end.
- """
- try:
- from providers import list_providers
- profiles = list_providers()
- except Exception:
- return 0
- pending = [pp for pp in profiles if _plugin_provider_enters_picker(pp)]
- added = 0
- while pending:
- # Rows anchored on a still-pending row wait for it, so a chain keeps its declared order.
- names = {pp.name for pp in pending}
- ready = [pp for pp in pending if pp.picker_after not in names] or pending
- for pp in ready:
- _admit_plugin_provider(pp)
- added += 1
- pending = [pp for pp in pending if pp not in ready]
- return added
- def _admit_plugin_provider(pp) -> None:
- """Add one plugin profile to the picker tables: row (at ``picker_after``), label, group and the
- aggregator set. Its catalog aliases are read per home (:class:`_CatalogAliases`)."""
- label = pp.display_name or pp.name
- entry = ProviderEntry(pp.name, label, pp.description or f"{label} (direct API)")
- slugs = [p.slug for p in CANONICAL_PROVIDERS]
- anchor = pp.picker_after
- CANONICAL_PROVIDERS.insert(slugs.index(anchor) + 1 if anchor in slugs else len(slugs), entry)
- _canonical_slugs.add(pp.name)
- _PROVIDER_LABELS[pp.name] = label
- if pp.picker_group:
- gid, group_label, group_desc = pp.picker_group
- members = PROVIDER_GROUPS.setdefault(gid, (group_label, group_desc, []))[2]
- if pp.name not in members:
- members.append(pp.name)
- _SLUG_TO_GROUP[pp.name] = gid
- if pp.aggregator:
- _AGGREGATOR_PROVIDERS.add(pp.name)
- _PROVIDER_LABELS: dict[str, str] = {p.slug: p.label for p in CANONICAL_PROVIDERS}
- _PROVIDER_LABELS["custom"] = "Custom endpoint" # special case: not a named provider
- def listed_canonical_providers() -> list[ProviderEntry]:
- """``CANONICAL_PROVIDERS`` minus pre-release profiles the user has not opted into: what a provider
- LIST offers. Typed paths (``provider:model``, labels, ``--provider``) keep the full table."""
- from hermes_cli import models # the binding every list consumer read before this filter existed
- from providers import unlisted_provider_names
- hidden = unlisted_provider_names()
- return [p for p in models.CANONICAL_PROVIDERS if p.slug not in hidden]
- # ---------------------------------------------------------------------------
- # Provider groups — DISPLAY ONLY. Vendors with several slugs (global API, China API, OAuth plan,
- # ...) fold under one top-level row in the INTERACTIVE PICKERS (``hermes model``, setup wizard,
- # Telegram ``/model``). They do NOT change CANONICAL_PROVIDERS, slug identity, ``--provider``,
- # ``/model <provider:model>`` or any typed path — every member slug stays individually addressable.
- # ``group_providers()`` is the single fold used by all three surfaces.
- # group_id -> (display_label, group_description shown on the collapsed row, [member_slug, ...])
- # Member order is the order shown inside the group submenu; member detail lives in ``tui_desc``.
- # ---------------------------------------------------------------------------
- PROVIDER_GROUPS: dict[str, tuple[str, str, list[str]]] = {
- "kimi": ("Kimi / Moonshot", "Coding Plan, Moonshot global & China endpoints", ["kimi-coding", "kimi-coding-cn"]),
- "minimax": ("MiniMax", "Global, OAuth Coding Plan & China endpoints", ["minimax", "minimax-oauth", "minimax-cn"]),
- "xai": ("xAI Grok", "Direct API or SuperGrok / Premium+ OAuth", ["xai", "xai-oauth"]),
- "google": ("Google Gemini", "Google AI Studio (API key)", ["gemini"]),
- "openai": ("OpenAI", "ChatGPT/Codex subscription or direct OpenAI API", ["openai-codex", "openai-api"]),
- "qwen": ("Qwen", "Qwen Cloud / DashScope, Coding Plan, Token Plan & Qwen CLI OAuth", ["alibaba", "alibaba-cn", "alibaba-coding-plan", "alibaba-coding-plan-cn", "alibaba-token-plan", "alibaba-token-plan-cn", "qwen-oauth"]),
- "opencode": ("OpenCode", "Zen pay-as-you-go or Go subscription", ["opencode-zen", "opencode-go"]),
- "tencent": ("Tencent Hy", "Hy4 / Hy3 via TokenHub & TokenPlan", ["tencent-tokenhub", "tencent-tokenplan"]),
- }
- # Reverse index: member slug -> group_id.
- _SLUG_TO_GROUP: dict[str, str] = {
- slug: gid for gid, (_label, _desc, members) in PROVIDER_GROUPS.items() for slug in members
- }
- def provider_group_for_slug(slug: str) -> str:
- """Return the group_id a provider slug belongs to, or "" if ungrouped."""
- return _SLUG_TO_GROUP.get(str(slug or "").strip().lower(), "")
- def group_providers(slugs):
- """Fold a flat ordered slug iterable into picker rows by provider group (DISPLAY ONLY).
- A group row appears at the position of its FIRST present member, in input order; later
- members fold into it. Member order inside a group follows ``PROVIDER_GROUPS`` declaration,
- restricted to the members present in ``slugs``.
- """
- present = set(slugs)
- group_members = {
- gid: [m for m in members if m in present]
- for gid, (_label, _desc, members) in PROVIDER_GROUPS.items()
- }
- rows = []
- seen: set[str] = set()
- emitted_groups: set[str] = set()
- for slug in slugs:
- s = str(slug or "").strip().lower()
- if not s or s in seen:
- continue
- seen.add(s)
- gid = _SLUG_TO_GROUP.get(s, "")
- if not gid:
- rows.append({"kind": "single", "slug": s})
- continue
- if gid in emitted_groups:
- continue # already folded at the first member's position
- emitted_groups.add(gid)
- members = group_members.get(gid) or [s]
- if len(members) <= 1:
- rows.append({"kind": "single", "slug": members[0]})
- else:
- label, desc, _ = PROVIDER_GROUPS[gid]
- rows.append({"kind": "group", "group_id": gid, "label": label,
- "description": desc, "members": list(members)})
- return rows
- class _CatalogAliases(dict):
- """``_PROVIDER_ALIASES``: the built-in alias table plus the ``catalog_aliases`` of the bound home's
- plugin profiles, resolved per read and never copied in: two served homes may give one alias to
- different providers. A built-in alias always wins; among plugins the first registered does."""
- _memo: tuple[tuple, dict[str, str]] = ((), {})
- def _plugin_aliases(self) -> dict[str, str]:
- from providers import bound_registry_key, list_providers
- token = bound_registry_key()
- if self._memo[0] == token:
- return self._memo[1]
- found: dict[str, str] = {}
- for pp in list_providers():
- if pp.catalog_aliases and pp.name not in _BUILTIN_SLUGS:
- for alias in pp.aliases:
- if not dict.__contains__(self, alias):
- found.setdefault(alias, pp.name)
- self._memo = (token, found)
- return found
- def __getitem__(self, key: str) -> str:
- if dict.__contains__(self, key):
- return dict.__getitem__(self, key)
- return self._plugin_aliases()[key]
- def get(self, key: str, default: Any = None) -> Any:
- if dict.__contains__(self, key):
- return dict.__getitem__(self, key)
- return self._plugin_aliases().get(key, default)
- def __contains__(self, key: object) -> bool:
- return dict.__contains__(self, key) or key in self._plugin_aliases()
- def items(self) -> list[tuple[str, str]]:
- return [*dict.items(self), *self._plugin_aliases().items()]
- def keys(self) -> list[str]:
- return [key for key, _ in self.items()]
- def values(self) -> list[str]:
- return [value for _, value in self.items()]
- def __iter__(self):
- return iter(self.keys())
- def __len__(self) -> int:
- return len(self.items())
- _PROVIDER_ALIASES = _CatalogAliases((
- ("glm", "zai"), ("z-ai", "zai"), ("z.ai", "zai"), ("zhipu", "zai"), ("google", "gemini"),
- ("google-gemini", "gemini"), ("google-ai-studio", "gemini"), ("google-vertex", "vertex"), ("vertex-ai", "vertex"),
- ("gcp-vertex", "vertex"), ("vertexai", "vertex"), ("kimi", "kimi-coding"), ("moonshot", "kimi-coding"),
- ("kimi-cn", "kimi-coding-cn"), ("moonshot-cn", "kimi-coding-cn"), ("step", "stepfun"),
- ("stepfun-coding-plan", "stepfun"), ("arcee-ai", "arcee"), ("arceeai", "arcee"), ("gmi-cloud", "gmi"),
- ("gmicloud", "gmi"), ("fireworks-ai", "fireworks"), ("fw", "fireworks"), ("actual-computer", "actual"),
- ("actualcomputer", "actual"), ("aci", "actual"), ("nebius", "nebius-token-factory"),
- ("nebius-tokenfactory", "nebius-token-factory"), ("nebius-tf", "nebius-token-factory"),
- ("token-factory", "nebius-token-factory"), ("tokenfactory", "nebius-token-factory"),
- ("minimax-china", "minimax-cn"), ("minimax_cn", "minimax-cn"), ("minimax-portal", "minimax-oauth"),
- ("minimax-global", "minimax-oauth"), ("minimax_oauth", "minimax-oauth"), ("claude", "anthropic"),
- ("claude-code", "anthropic"), ("deep-seek", "deepseek"), ("opencode", "opencode-zen"), ("zen", "opencode-zen"),
- ("go", "opencode-go"), ("opencode-go-sub", "opencode-go"), ("aigateway", "ai-gateway"), ("vercel", "ai-gateway"),
- ("vercel-ai-gateway", "ai-gateway"), ("kilo", "kilocode"), ("kilo-code", "kilocode"),
- ("kilo-gateway", "kilocode"), ("dashscope", "alibaba"), ("aliyun", "alibaba"), ("qwen", "alibaba"),
- ("alibaba-cloud", "alibaba"), ("qwen-portal", "qwen-oauth"), ("hf", "huggingface"),
- ("hugging-face", "huggingface"), ("huggingface-hub", "huggingface"), ("novita-ai", "novita"),
- ("novitaai", "novita"), ("mimo", "xiaomi"), ("xiaomi-mimo", "xiaomi"), ("tencent", "tencent-tokenhub"),
- ("tokenhub", "tencent-tokenhub"), ("tencent-cloud", "tencent-tokenhub"), ("tencentmaas", "tencent-tokenhub"),
- ("tokenplan", "tencent-tokenplan"), ("tencent-lkeap", "tencent-tokenplan"), ("aws", "bedrock"),
- ("aws-bedrock", "bedrock"), ("amazon-bedrock", "bedrock"), ("amazon", "bedrock"), ("grok", "xai"),
- ("grok-oauth", "xai-oauth"), ("xai-oauth", "xai-oauth"), ("x-ai-oauth", "xai-oauth"),
- ("xai-grok-oauth", "xai-oauth"), ("x-ai", "xai"), ("x.ai", "xai"), ("nim", "nvidia"), ("nvidia-nim", "nvidia"),
- ("build-nvidia", "nvidia"), ("nemotron", "nvidia"), ("lmstudio", "lmstudio"), ("lm-studio", "lmstudio"),
- ("lm_studio", "lmstudio"), ("chatgpt", "openai-codex"), ("chatgpt-codex", "openai-codex"),
- ("ollama", "custom"), # bare "ollama" = local; use "ollama-cloud" for cloud
- ("ollama_cloud", "ollama-cloud"),
- # Local OpenAI-compatible servers route through the generic "custom" provider
- # (parity with hermes_cli.auth and hermes_cli.providers). Issue #62213. The llamacpp
- # aliases stay unmapped: they are the managed local runtime's picker id, and the model
- # validator must reach its staged-library branch before the custom one.
- ("local", "custom"), ("vllm", "custom"),
- ))
- # Offline/fresh-install fallback for the model Hermes silently lands on when the user never picked
- # one (GUI onboarding confirm card, empty ``model.default``, provider-set-but-model-missing). The
- # AUTHORITATIVE source is the remote catalog manifest, which labels exactly one entry per provider
- # ``"default": true`` (get_default_model_from_cache) so the default rotates without a release; this
- # MUST match the labeled entry in website/static/api/model-catalog.json. Deliberately a capable
- # low-cost model rather than the curated lists' entry [0]: aggregator lists are ordered
- # most-capable-first, so [0] is the priciest Anthropic flagship.
- PREFERRED_SILENT_DEFAULT_MODEL = "z-ai/glm-5.3"
- # Providers whose *silent* auto-default goes through the cost-safe catalog-labeled default
- # (``get_preferred_silent_default_model``) instead of curated entry [0]. Metered aggregators order
- # best-first, so [0] is the priciest flagship; a profile that sets a provider with no model would
- # otherwise silently bill the most expensive model (863 Opus requests before one user noticed).
- # Network-free (cache-only) on purpose — this is the hot resolution path. The *interactive* default
- # (GUI onboarding / ``hermes model``) uses the tier-aware ``get_recommended_default_model`` in
- # hermes_cli/web_server.py + ``partition_nous_models_by_tier``, which may hit the Portal.
- _SILENT_DEFAULT_PROVIDERS: frozenset[str] = frozenset({"nous", "openrouter"})
- # Retired model IDs kept for /model auto-detect only — not shown in pickers. DeepSeek cut these
- # off; model_normalize remaps them on the wire.
- _PROVIDER_RETIRED_ALIASES: dict[str, tuple[str, ...]] = {
- "deepseek": ("deepseek-chat", "deepseek-reasoner"),
- }
- # Plugin profiles with ``aggregator=True`` join at sync time (``sync_plugin_provider_catalog``).
- _AGGREGATOR_PROVIDERS: set[str] = {"nous", "openrouter", "ai-gateway", "kilocode"}
- sync_plugin_provider_catalog() # every table it extends is defined above
- # Subscription/OAuth providers whose catalogs RE-EXPOSE other vendors' models; tried only as a last
- # resort for bare short-alias resolution (after every native-vendor catalog) so they never hijack
- # an alias from the model's native vendor. None currently defined.
- _BORROWED_MODEL_PROVIDERS: frozenset[str] = frozenset()
- # Providers whose live /v1/models is the authoritative catalog: the picker merges live-first (live
- # entries lead, curated-only append). Every OTHER provider keeps curated-first so a deliberately
- # surfaced newest model stays on top when the live API lags. Zen/Go re-expose dozens of vendors
- # and rotate them often, so their stale curated entries must not pollute the top.
- _LIVE_FIRST_PICKER_PROVIDERS: frozenset[str] = frozenset({"opencode-zen", "opencode-go", "meta-ai"})
- # Models supporting OpenAI Priority Processing (service_tier="priority"; see
- # openai.com/api-priority-processing). Pattern-based: any OpenAI flagship (gpt-*, o1*, o3*, o4*).
- # Non-OpenAI endpoints (OpenRouter/opencode-zen and other proxies) strip service_tier, so false
- # positives are harmless. Codex-series models are excluded — the Codex Responses API doesn't
- # expose service_tier.
- _OPENAI_FAST_MODE_PREFIXES: tuple[str, ...] = ("gpt-", "o1", "o3", "o4")
- # OpenAI Ultrafast (service_tier="ultrafast", 6x Standard): GPT-6 Astra and GPT-6.1 Sol
- # (developers.openai.com/api/docs/guides/ultrafast-mode, 2026-10-08).
- # Exact wire slugs, matched after stripping the vendor prefix and the Hermes-side ``-900k`` alias.
- _OPENAI_ULTRAFAST_MODELS: frozenset[str] = frozenset({"gpt-6-astra", "gpt-6.1-sol"})
- # Providers where models.dev is authoritative: the curated list is an offline fallback plus custom
- # additions the registry lacks, merged fresh-first (curated-only names appended) for both the CLI
- # and the gateway /model picker. DELIBERATELY EXCLUDED: "openrouter" (curated list is a hand-picked
- # agentic subset of 400+ models — merging would dump everything), "nous" (curated list + Portal
- # /models are the subscription-tier source of truth), and providers with dedicated live-endpoint
- # branches or a profile catalog hook (anthropic, ai-gateway, ollama-cloud, custom, stepfun, openai-codex).
- _MODELS_DEV_PREFERRED: frozenset[str] = frozenset({
- "opencode-go", "opencode-zen", "kilocode", "fireworks", "mistral", "togetherai", "cohere",
- "perplexity", "groq", "nvidia", "huggingface", "zai", "gemini", "google", "xai", "xai-oauth",
- })
- # Azure Foundry model families that require the Responses API: Azure rejects /chat/completions
- # against them with ``400 "The requested operation is unsupported."`` (seen on gpt-5.3-codex while
- # gpt-4o on the same endpoint worked). Broad enough for vendor-renamed deployments (gpt-5.x-codex,
- # o1-preview), tight enough to leave GPT-4 / 3.5 / Llama / Mistral / Grok on chat completions.
- _AZURE_FOUNDRY_RESPONSES_PREFIXES = ("codex", "gpt-5", "o1", "o3", "o4")
|