mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-06-09 08:21:50 +00:00
03392b67d6
Salvaged from #40429; re-verified on main, tightened, tested. Co-authored-by: jimjsong <jimjsong@users.noreply.github.com>
126 lines
4.6 KiB
Python
126 lines
4.6 KiB
Python
"""OpenCode provider profiles (Zen + Go).
|
|
|
|
Both use per-model api_mode routing:
|
|
- OpenCode Zen: Claude → anthropic_messages, GPT-5/Codex → codex_responses,
|
|
everything else → chat_completions (this profile)
|
|
- OpenCode Go: MiniMax → anthropic_messages, GLM/Kimi → chat_completions
|
|
(this profile)
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import Any
|
|
|
|
from providers import register_provider
|
|
from providers.base import ProviderProfile
|
|
|
|
|
|
def _flat_model_name(model: str | None) -> str:
|
|
"""Return the bare OpenCode model ID, tolerating aggregator prefixes."""
|
|
return (model or "").strip().rsplit("/", 1)[-1].lower()
|
|
|
|
|
|
def _is_kimi_k2_model(model: str | None) -> bool:
|
|
return _flat_model_name(model).startswith("kimi-k2")
|
|
|
|
|
|
def _is_deepseek_thinking_model(model: str | None) -> bool:
|
|
m = _flat_model_name(model)
|
|
if m.startswith("deepseek-v") and not m.startswith("deepseek-v3"):
|
|
return True
|
|
return m == "deepseek-reasoner"
|
|
|
|
|
|
class OpenCodeGoProfile(ProviderProfile):
|
|
"""OpenCode Go - model-specific reasoning controls."""
|
|
|
|
# Per-model completion-token cap. The opencode-go relay's default is
|
|
# too large for mimo-v2.5-pro — it sends max_tokens=262144 but Xiaomi
|
|
# only supports 131072 completion tokens and 400s the request.
|
|
# Setting an explicit cap here prevents the relay default from being
|
|
# applied. Keys are normalized via _flat_model_name().
|
|
_MODEL_MAX_TOKENS: dict[str, int] = {
|
|
"mimo-v2.5-pro": 131072,
|
|
}
|
|
|
|
def get_max_tokens(self, model: str | None) -> int | None:
|
|
cap = self._MODEL_MAX_TOKENS.get(_flat_model_name(model))
|
|
if cap is not None:
|
|
return cap
|
|
return self.default_max_tokens
|
|
|
|
def build_api_kwargs_extras(
|
|
self, *, reasoning_config: dict | None = None, model: str | None = None, **context
|
|
) -> tuple[dict[str, Any], dict[str, Any]]:
|
|
extra_body: dict[str, Any] = {}
|
|
top_level: dict[str, Any] = {}
|
|
|
|
if _is_kimi_k2_model(model):
|
|
# Kimi K2 on OpenCode Go uses Moonshot's native wire shape:
|
|
# extra_body.thinking (binary toggle) + top-level reasoning_effort
|
|
# (low|medium|high). Mirrors the KimiProfile (api.moonshot.ai/v1).
|
|
if not isinstance(reasoning_config, dict):
|
|
# No config → leave server defaults alone.
|
|
return extra_body, top_level
|
|
|
|
enabled = reasoning_config.get("enabled") is not False
|
|
if not enabled:
|
|
extra_body["thinking"] = {"type": "disabled"}
|
|
return extra_body, top_level
|
|
|
|
effort = (reasoning_config.get("effort") or "").strip().lower()
|
|
if effort in {"xhigh", "max"}:
|
|
top_level["reasoning_effort"] = "high"
|
|
elif effort in {"low", "medium", "high"}:
|
|
top_level["reasoning_effort"] = effort
|
|
|
|
# Avoid "cannot specify both 'thinking' and 'reasoning_effort'" HTTP 400:
|
|
# only send extra_body["thinking"] when no reasoning_effort is set.
|
|
if "reasoning_effort" not in top_level:
|
|
extra_body["thinking"] = {"type": "enabled"}
|
|
return extra_body, top_level
|
|
|
|
if not _is_deepseek_thinking_model(model):
|
|
return extra_body, top_level
|
|
|
|
enabled = True
|
|
if isinstance(reasoning_config, dict) and reasoning_config.get("enabled") is False:
|
|
enabled = False
|
|
|
|
if not enabled:
|
|
extra_body["thinking"] = {"type": "disabled"}
|
|
return extra_body, top_level
|
|
|
|
if isinstance(reasoning_config, dict):
|
|
effort = (reasoning_config.get("effort") or "").strip().lower()
|
|
if effort in {"xhigh", "max"}:
|
|
top_level["reasoning_effort"] = "max"
|
|
elif effort in {"low", "medium", "high"}:
|
|
top_level["reasoning_effort"] = effort
|
|
|
|
# Avoid "cannot specify both 'thinking' and 'reasoning_effort'" HTTP 400:
|
|
# only send extra_body["thinking"] when no reasoning_effort is set.
|
|
if "reasoning_effort" not in top_level:
|
|
extra_body["thinking"] = {"type": "enabled"}
|
|
|
|
return extra_body, top_level
|
|
|
|
|
|
opencode_zen = ProviderProfile(
|
|
name="opencode-zen",
|
|
aliases=("opencode", "opencode_zen", "zen"),
|
|
env_vars=("OPENCODE_ZEN_API_KEY",),
|
|
base_url="https://opencode.ai/zen/v1",
|
|
default_aux_model="gemini-3-flash",
|
|
)
|
|
|
|
opencode_go = OpenCodeGoProfile(
|
|
name="opencode-go",
|
|
aliases=("opencode_go", "go", "opencode-go-sub"),
|
|
env_vars=("OPENCODE_GO_API_KEY",),
|
|
base_url="https://opencode.ai/zen/go/v1",
|
|
default_aux_model="glm-5",
|
|
)
|
|
|
|
register_provider(opencode_zen)
|
|
register_provider(opencode_go)
|