Crate langchain_rust 

Observability

The crate is instrumented with tracing: every LLM call runs in an llm.generate span (provider, model, token usage, tool call count, error), embeddings in embed.documents / embed.query, LLMChain in chain.call, and agent runs in agent.run with an agent.plan span per step and a tool.call span per tool. Install any subscriber (tracing-subscriber, OpenTelemetry, Langfuse…) to see them. Without one, log events are forwarded to the log crate, so env_logger keeps working.

Re-exports

pub use url;

Modules

agent

chain

document_loaders

embedding

language_models

llm

memory

output_parsers

Output parsers.

prompt

schemas

semantic_router

text_splitter

tools

vectorstore

Macros

add_documents

fmt_message

fmt_message is a utility macro used to create a MessageOrTemplate::Message variant.

fmt_placeholder

fmt_placeholder is a utility macro used to create a MessageOrTemplate::MessagesPlaceholder variant.

fmt_template

fmt_template is a utility macro used to create a MessageOrTemplate::Template variant.

message_formatter

Example

prompt_args

prompt_args! is a utility macro used for creating a std::collections::HashMap<String, serde_json::Value>. This HashMap can then be passed as arguments to a function or method.

sequential_chain

similarity_search

template_fstring

template_fstring is a utility macro that creates a new PromptTemplate with FString as the template format.

template_jinja2

template_jinja2 is a utility macro that creates a new PromptTemplate with Jinja2 as the template format.