Crate langfuse 

Langfuse SDK for Rust — LLM observability, prompt management, and evaluation.

Quick Start

use langfuse::Langfuse;

let client = Langfuse::new(
    langfuse_core::LangfuseConfig::builder()
        .public_key("pk-lf-...")
        .secret_key("sk-lf-...")
        .build()
        .unwrap()
).unwrap();

Re-exports

pub use client::Langfuse;

pub use langfuse_tracing::context::get_current_observation_id;

pub use langfuse_tracing::context::get_current_trace_id;

pub use langfuse_tracing::context::propagate_as_baggage;

pub use langfuse_tracing::context::propagate_attributes;

pub use langfuse_tracing::context_apis::get_current_trace_url;

pub use langfuse_tracing::context_apis::score_current_span;

pub use langfuse_tracing::context_apis::score_current_span_with;

pub use langfuse_tracing::context_apis::score_current_trace;

pub use langfuse_tracing::context_apis::set_current_trace_as_public;

pub use langfuse_tracing::context_apis::update_current_generation;

pub use langfuse_tracing::context_apis::update_current_span;

pub use langfuse_tracing::embedding::LangfuseEmbedding;

pub use langfuse_tracing::generation::LangfuseGeneration;

pub use langfuse_tracing::observe::with_agent;

pub use langfuse_tracing::observe::with_chain;

pub use langfuse_tracing::observe::with_embedding;

pub use langfuse_tracing::observe::with_evaluator;

pub use langfuse_tracing::observe::with_generation;

pub use langfuse_tracing::observe::with_guardrail;

pub use langfuse_tracing::observe::with_observation;

pub use langfuse_tracing::observe::with_retriever;

pub use langfuse_tracing::observe::with_span;

pub use langfuse_tracing::observe::with_tool;

pub use langfuse_tracing::span::LangfuseSpan;

pub use langfuse_tracing::stream_wrapper::ObservingIterator;

pub use langfuse_tracing::stream_wrapper::ObservingStream;

Modules

client

The top-level Langfuse client that ties together all managers and the tracing pipeline.

datasets

Dataset management and experiment runner for the Langfuse SDK.

evaluation

Evaluation result types for dataset evaluators. Evaluation result types for dataset experiment evaluators.

http

HTTP client utilities: shared client builder and retry logic.

langfuse_tracing

OpenTelemetry-native tracing pipeline for Langfuse.

media

Media management — upload, fetch, and reference-token parsing.

observation

Observation types, levels, usage/cost details, and span attributes.

prompt

Prompt template types (text and chat).

prompts

Prompt management: fetching, caching, and template compilation.

score

Score types, values, and builder.

scoring

Score management with batched async flush.

Structs

ChatMessage

A chat message in a chat prompt template.

CostDetails

Cost details for LLM generations.

Evaluation

An evaluation result produced by an Evaluator.

LangfuseConfig

Configuration for the Langfuse client.

MediaContent

Content type and raw bytes for media.

MediaReference

Reference token format used in Langfuse for inline media. Format: @@@langfuseMedia:type=<content_type>|id=<media_id>|source=<source>@@@

PropagateAttributes

Attributes that propagate from a parent trace to child spans.

ScoreBody

Body for creating a score.

ScoreBodyBuilder

Builder for constructing a ScoreBody with a fluent API.

SpanUpdateParams

Parameters for batch-updating span attributes.

UsageDetails

Token usage details for LLM generations.

Enums

LangfuseError

Primary error type for all Langfuse operations.

ObservationType

The 10 observation types supported by Langfuse.

PromptType

The type of prompt template.

ScoreDataType

The data type of a score.

ScoreValue

A score value — can be numeric, categorical, or boolean.

SpanLevel

Severity level for spans.

Type Aliases

Result

Convenience result type for Langfuse operations.

Attribute Macros

observe

Instrument a function with Langfuse tracing.