Re-exports§
pub use anthropic::AnthropicClient;pub use anthropic::AnthropicFactory;pub use client::LlmClient;pub use deepseek::DeepSeekFactory;pub use deepseek::DEFAULT_BASE_URL as DEEPSEEK_DEFAULT_BASE_URL;pub use gemini::GeminiClient;pub use gemini::GeminiFactory;pub use minimax::MiniMaxClient;pub use openai_compat::OpenAiClient;pub use prompt_block::flatten_blocks;pub use prompt_block::CachePolicy;pub use prompt_block::PromptBlock;pub use quota_tracker::QuotaTracker;pub use rate_limit_info::extract_rate_limit_info;pub use rate_limit_info::format_rate_limit_message;pub use rate_limit_info::format_using_overage;pub use rate_limit_info::LlmProvider as RateLimitProvider;pub use rate_limit_info::QuotaStatus;pub use rate_limit_info::RateLimitInfo;pub use rate_limit_info::RateLimitMessage;pub use rate_limit_info::RateLimitSeverity;pub use rate_limit_info::RateLimitWindow;pub use rate_limiter::RateLimiter;pub use registry::LlmProviderCatalogEntry;pub use registry::LlmProviderFactory;pub use registry::LlmRegistry;pub use registry::MiniMaxFactory;pub use registry::OpenAiFactory;pub use retry::parse_retry_after_ms;pub use retry::with_retry;pub use retry::LlmError;pub use stream::collect_stream;pub use stream::default_stream_from_chat;pub use stream::StreamChunk;pub use token_counter::AnthropicTokenCounter;pub use token_counter::TiktokenCounter;pub use token_counter::TokenCounter;pub use types::Attachment;pub use types::AttachmentData;pub use types::CacheUsage;pub use types::ChatMessage;pub use types::ChatRequest;pub use types::ChatResponse;pub use types::ChatRole;pub use types::FinishReason;pub use types::ResponseContent;pub use types::TokenUsage;pub use types::ToolCall;pub use types::ToolChoice;pub use types::ToolDef;
Modules§
- anthropic
- Anthropic Messages API client (Claude).
- anthropic_
auth - Anthropic auth sources: API key, setup-token (Bearer), and OAuth subscription bundle with refresh.
- client
- deepseek
- DeepSeek connector.
- gemini
- Google Gemini API client (
generateContent). - minimax
- MiniMax LLM client with two wire formats:
- minimax_
auth - MiniMax auth sources: static API key and Token Plan OAuth bundle.
- openai_
compat - prompt_
block - Structured system-prompt assembly with explicit cache breakpoints.
- quota_
tracker - rate_
limit_ info - Transversal rate-limit header extraction and human-readable message generation for all LLM providers.
- rate_
limiter - registry
- retry
- stream
- Incremental streaming primitives for
LlmClient. - telemetry
- Streaming telemetry for
nexo-llm. - text_
sanitize - Strip lone UTF-16 surrogate code points before sending text to the
Anthropic API. Mirrors OpenClaw’s
sanitizeTransportPayloadText. - token_
counter - Pre-flight token counting for LLM requests.
- types