Crate lc_core 

Core abstractions for LangChainRust.

This crate provides the foundational traits and types:

• Runnable: Base execution interface
• BaseLanguageModel: LLM abstraction
• BaseChatModel: Chat model interface
• BaseTool, Tool: Tool abstraction
• RunnableConfig: Execution configuration with callbacks
• Output parsers, structured output, caching, token counting, batch API

Re-exports

pub use json_parse::parse_llm_json;

pub use json_parse::parse_llm_json_with_retry;

pub use json_parse::LlmJsonParseError;

pub use judge::structured_call;

pub use judge::truncate;

pub use judge::StructuredJudgeError;

pub use language_models::predict_tools;

pub use language_models::BaseChatModel;

pub use language_models::BaseLanguageModel;

pub use language_models::MultimodalError;

pub use language_models::MultimodalModel;

pub use language_models::PredictToolsError;

pub use language_models::StreamChunk;

pub use language_models::TokenUsage;

pub use output_parsers::BaseOutputParser;

pub use output_parsers::CommaSeparatedListOutputParser;

pub use output_parsers::JsonOutputParser;

pub use output_parsers::OutputParserError;

pub use output_parsers::OutputParserResult;

pub use output_parsers::StrOutputParser;

pub use output_parsers::StructuredOutputParser;

pub use output_parsers::TypedOutputParser;

pub use runnables::into_runnable_any;

pub use runnables::CancellationToken;

pub use runnables::LcelError;

pub use runnables::RetryConfig;

pub use runnables::RetryOn;

pub use runnables::Runnable;

pub use runnables::RunnableAny;

pub use runnables::RunnableAnyWrapper;

pub use runnables::RunnableAssign;

pub use runnables::RunnableBinding;

pub use runnables::RunnableBranch;

pub use runnables::RunnableConfig;

pub use runnables::RunnableConfigurable;

pub use runnables::RunnableConfigurableFields;

pub use runnables::RunnableExt;

pub use runnables::RunnableLambda;

pub use runnables::RunnableParallel;

pub use runnables::RunnablePassthrough;

pub use runnables::RunnablePick;

pub use runnables::RunnableRetry;

pub use runnables::RunnableSequence;

pub use runnables::RunnableWithFallbacks;

pub use structured_output::stream_structured_output;

pub use structured_output::with_structured_output;

pub use structured_output::PartialJsonError;

pub use structured_output::PartialJsonParser;

pub use structured_output::StreamingStructuredOutputExt;

pub use structured_output::StructuredOutputError;

pub use structured_output::StructuredOutputExt;

pub use tools::BaseTool;

pub use tools::FunctionDefinition;

pub use tools::StructuredOutput;

pub use tools::Tool;

pub use tools::ToolDefinition;

pub use tools::ToolError;

pub use tools::ToolRegistry;

Modules

batch

Batch API client for OpenAI and Anthropic batch processing.

cache

LLM call cache

json_parse

Tolerant JSON parsing for LLM outputs.

judge

Shared LLM judge infrastructure: lets an LLM return a verdict via structured arguments.

language_models

Language model base traits.

math

Shared math utilities.

output_parsers

Output parsers for extracting and transforming LLM response text.

router_llm

Model routing, fallback and load balancing.

runnables

Runnable module - LangChain Expression Language (LCEL) core

ssrf

SSRF protection — a single shared implementation, no copies allowed.

structured_output

Structured output utilities for extracting typed data from LLM responses.

token_counter

Token counter and cost tracking

tools

Tool abstractions for agent function calling.

Structs

FunctionCall

Function call inside a ToolCall

ToolCall

Tool call from LLM response

ToolCallBuilder

Builder for constructing a ToolCall field by field.

ToolCallResult

Tool call result to send back to LLM

Enums

CoreError

Unified error type for the lc-core crate.

Functions

other_error

Helper to convert any error into CoreError::Other. Use this instead of ? when the error type is not a known CoreError variant.