| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081 |
- """Kimi / Moonshot provider profiles (chat_completions path; sk-kimi-* keys are
- redirected to api.kimi.com/coding by core)."""
- from typing import Any
- from urllib.parse import urlparse
- from agent.reasoning_effort import KIMI_K3_EFFORTS, KIMI_K3_OVERRIDES, thinking_toggle_extras
- from hermes_cli.version_info import hermes_user_agent
- from providers import register_provider
- from providers.base import OMIT_TEMPERATURE, LazyHeaders, ProviderProfile
- _HEADERS = {
- "HTTP-Referer": "https://hermes-agent.nousresearch.com",
- "X-Title": "Hermes Agent",
- "User-Agent": lambda: hermes_user_agent("HermesAgent"),
- # Exclude brotli: httpx's brotlicffi backend has a streaming decode bug on
- # Moonshot's content-encoding: br SSE responses (#28043, #48428, #59556).
- # gzip sidesteps it while still compressing the transfer.
- "Accept-Encoding": "gzip",
- }
- def _is_confirmed_kimi_coding_url(base_url: str) -> bool:
- """True only for Kimi Code's canonical HTTPS API surfaces."""
- try:
- p = urlparse(base_url)
- port = p.port
- except ValueError:
- return False
- return (
- p.scheme.lower() == "https" and (p.hostname or "").lower() == "api.kimi.com" and port in (None, 443)
- and p.username is None and p.password is None
- and p.path.rstrip("/") in {"/coding", "/coding/v1"} and not p.query and not p.fragment
- )
- class KimiProfile(ProviderProfile):
- """Kimi/Moonshot — temperature omitted, thinking xor reasoning_effort."""
- def fetch_models(
- self, *, api_key: str | None = None, base_url: str | None = None, timeout: float = 8.0
- ) -> list[str] | None:
- """Use Kimi Code's OpenAI-compatible surface for model discovery; the bare
- ``k3`` slug is only served there, so it is filtered off other endpoints."""
- effective_base = (base_url or self.base_url or "").rstrip("/")
- confirmed_coding_endpoint = _is_confirmed_kimi_coding_url(effective_base)
- if confirmed_coding_endpoint and urlparse(effective_base).path.rstrip("/") == "/coding":
- effective_base += "/v1"
- models = super().fetch_models(api_key=api_key, base_url=effective_base or None, timeout=timeout)
- if models is None or confirmed_coding_endpoint:
- return models
- return [model for model in models if model.strip().lower() != "k3"]
- def build_api_kwargs_extras(
- self, *, reasoning_config: dict | None = None, model: str | None = None, **context
- ) -> tuple[dict[str, Any], dict[str, Any]]:
- """Moonshot treats extra_body.thinking and reasoning_effort as mutually
- exclusive (400 on both): send effort when requested, else the toggle. K2.x keeps
- replayed prior-turn reasoning_content only with ``thinking.keep: "all"`` (default null
- ignores it; platform.kimi.ai/docs/guide/use-thinking-models); K3 always keeps it."""
- extra_body, top_level = thinking_toggle_extras(reasoning_config, KIMI_K3_EFFORTS, KIMI_K3_OVERRIDES)
- thinking = extra_body.get("thinking")
- if isinstance(thinking, dict) and thinking.get("type") == "enabled" and "k2" in (model or "").lower():
- extra_body["thinking"] = {**thinking, "keep": "all"}
- return extra_body, top_level
- def _kimi(name: str, aliases: tuple, env_vars: tuple, base_url: str) -> KimiProfile:
- return KimiProfile(
- name=name, aliases=aliases, env_vars=env_vars, base_url=base_url,
- fixed_temperature=OMIT_TEMPERATURE, default_max_tokens=32000,
- default_headers=LazyHeaders(_HEADERS), default_aux_model="kimi-k2-turbo-preview",
- )
- kimi = _kimi("kimi-coding", ("kimi", "moonshot", "kimi-for-coding"), ("KIMI_API_KEY", "KIMI_CODING_API_KEY"),
- "https://api.moonshot.ai/v1")
- kimi_cn = _kimi("kimi-coding-cn", ("kimi-cn", "moonshot-cn"), ("KIMI_CN_API_KEY",), "https://api.moonshot.cn/v1")
- register_provider(kimi)
- register_provider(kimi_cn)
|