Documentation
¶
Overview ¶
Package llm contains Core's LLM abstraction. Provider-specific model clients and API keys belong outside Core; Core can either receive an injected Client or call a generic internal completion endpoint through RemoteClient.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Client ¶
Client 单轮 system+user → 文本完成。max_tokens 由实现侧决定。
调用方在持有 nil interface 时应走规则 fallback(详见 task.llmParse / skill.Benchmark)。 注意区分 typed nil:wire 处应用真 nil interface,不要把 (*concrete)(nil) 赋给本接口。
type OpenAIClient ¶
type OpenAIClient struct {
// contains filtered or unexported fields
}
OpenAIClient 调用任意 OpenAI 兼容接口（openai.com / Azure / 本地代理均可）。
自托管用户可通过 LLM_OPENAI_URL + LLM_OPENAI_API_KEY + LLM_OPENAI_MODEL 直接配置，无需私有代理服务。
func NewOpenAIClient ¶
func NewOpenAIClient(baseURL, apiKey, model string) *OpenAIClient
NewOpenAIClient 返回 OpenAI 兼容 client。baseURL 为空时返回 nil（调用方走规则 fallback）。
baseURL 示例：
type RemoteClient ¶
type RemoteClient struct {
// contains filtered or unexported fields
}
RemoteClient calls an internal LLM completion service. Core intentionally keeps provider-specific clients, keys, and model selection outside this repo.
func NewRemoteClient ¶
func NewRemoteClient(endpoint, internalToken string) *RemoteClient
NewRemoteClient returns nil when endpoint is empty so callers can keep the rule-based fallback path.
Source Files