config/provider_catalog.py

"""Neutral provider catalog: IDs, credentials, defaults, proxy and capability metadata.

Adapter factories live in :mod:`providers.registry`; this module stays free of
provider implementation imports (see contract tests).
"""

from __future__ import annotations

from dataclasses import dataclass
from typing import Literal

TransportType = Literal["openai_chat", "anthropic_messages"]

# Default upstream base URLs (also re-exported via :mod:`providers.defaults`)
NVIDIA_NIM_DEFAULT_BASE = "https://integrate.api.nvidia.com/v1"
ZEN_DEFAULT_BASE = "https://opencode.ai/zen"
CEREBRAS_DEFAULT_BASE = "https://api.cerebras.ai/v1"
SILICON_DEFAULT_BASE = "https://api.siliconflow.com/v1"
GROQ_DEFAULT_BASE = "https://api.groq.com/openai/v1"


@dataclass(frozen=True, slots=True)
class ProviderDescriptor:
    """Metadata for building :class:`~providers.base.ProviderConfig` and factory wiring."""

    provider_id: str
    transport_type: TransportType
    capabilities: tuple[str, ...]
    credential_env: str | None = None
    credential_url: str | None = None
    credential_attr: str | None = None
    static_credential: str | None = None
    default_base_url: str | None = None
    base_url_attr: str | None = None
    proxy_attr: str | None = None


PROVIDER_CATALOG: dict[str, ProviderDescriptor] = {
    "nvidia_nim": ProviderDescriptor(
        provider_id="nvidia_nim",
        transport_type="openai_chat",
        credential_env="NVIDIA_NIM_API_KEY_QWEN",
        credential_url="https://build.nvidia.com/settings/api-keys",
        credential_attr="nvidia_nim_api_key_qwen",
        default_base_url=NVIDIA_NIM_DEFAULT_BASE,
        proxy_attr="nvidia_nim_proxy",
        capabilities=("chat", "streaming", "tools", "thinking", "rate_limit"),
    ),
    "zen": ProviderDescriptor(
        provider_id="zen",
        transport_type="openai_chat",
        credential_env="ZEN_API_KEY",
        credential_url="https://opencode.ai/settings",
        credential_attr="zen_api_key",
        default_base_url=ZEN_DEFAULT_BASE,
        base_url_attr="zen_base_url",
        capabilities=("chat", "streaming", "tools", "thinking"),
    ),
    "cerebras": ProviderDescriptor(
        provider_id="cerebras",
        transport_type="openai_chat",
        credential_env="CEREBRAS_API_KEY",
        credential_url="https://cerebras.ai/labs",
        credential_attr="cerebras_api_key",
        default_base_url=CEREBRAS_DEFAULT_BASE,
        capabilities=("chat", "streaming", "tools", "thinking"),
    ),
    "silicon": ProviderDescriptor(
        provider_id="silicon",
        transport_type="openai_chat",
        credential_env="SILICON_API_KEY",
        credential_url="https://siliconflow.cn",
        credential_attr="silicon_api_key",
        default_base_url=SILICON_DEFAULT_BASE,
        capabilities=("chat", "streaming", "tools", "thinking"),
    ),
    "groq": ProviderDescriptor(
        provider_id="groq",
        transport_type="openai_chat",
        credential_env="GROQ_API_KEY",
        credential_url="https://console.groq.com/keys",
        credential_attr="groq_api_key",
        default_base_url=GROQ_DEFAULT_BASE,
        capabilities=("chat", "streaming", "tools", "thinking"),
    ),
}

# Order matches docs; must match PROVIDER_CATALOG keys.
SUPPORTED_PROVIDER_IDS: tuple[str, ...] = (
    "nvidia_nim",
    "zen",
    "cerebras",
    "silicon",
    "groq",
)

if len(set(SUPPORTED_PROVIDER_IDS)) != len(SUPPORTED_PROVIDER_IDS):
    raise AssertionError("Duplicate provider ids in PROVIDER_CATALOG key order")