Tracing Swiftide with Langfuse

This guide demonstrates how to integrate Langfuse into your Swiftide workflows to monitor, debug, and evaluate your LLM applications.

What is Swiftide?: Swiftide is a Rust library for building LLM applications. It provides simple primitives for prompt completion, streaming indexing and querying pipelines, and agents that can use tools or call other agents. It has integrations with popular LLMs and storage providers, a modular API, and first-class support for tracing and Langfuse. Documentation is available at swiftide.rs.

Install Dependencies

Enable the langfuse feature when adding Swiftide to your Cargo.toml:

swiftide = { version = "0.31", features = ["langfuse", "openai"] }

Also add tracing-subscriber to configure tracing:

tracing-subscriber = "0.3"

Configure Environment & Langfuse Credentials

Set the required environment variables for Langfuse. You can obtain these keys from the Langfuse Cloud project settings page or by self-hosting Langfuse.

export LANGFUSE_PUBLIC_KEY=pk-lf-...
export LANGFUSE_SECRET_KEY=sk-lf-...
# optional, defaults to http://localhost:3000
export LANGFUSE_URL=https://cloud.langfuse.com

Instrumenting Swiftide with Langfuse

Swiftide integrates with tracing. To send traces to Langfuse, set up the LangfuseLayer alongside your normal tracing layers.

use tracing::level_filters::LevelFilter;
use tracing_subscriber::{
    EnvFilter, Layer as _, layer::SubscriberExt as _, util::SubscriberInitExt as _,
};
use swiftide::langfuse::LangfuseLayer;
 
let fmt_layer = tracing_subscriber::fmt::layer()
    .compact()
    .with_target(false)
    .boxed();
 
let langfuse_layer = LangfuseLayer::default()
    .with_filter(LevelFilter::DEBUG)
    .boxed();
 
let registry = tracing_subscriber::registry()
    .with(EnvFilter::from_default_env())
    .with(vec![fmt_layer, langfuse_layer]);
 
registry.init();

Once this is set up, any instrumented spans or Swiftide operations will be reported to Langfuse.

Hello World Example

Here is a minimal Swiftide program with Langfuse enabled. It sends a simple prompt to OpenAI through Swiftide and logs the trace to Langfuse.

//! This is an example of using the langfuse integration with Swiftide.
//!
//! Langfuse is a platform for tracking and monitoring LLM usage and performance.
//!
//! When the feature `langfuse` is enabled, Swiftide can report tracing information,
//! usage, inputs, and outputs to langfuse.
//!
//! For this to work, you need to set the LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY
//! to the appropriate values. You can also set the LANGFUSE_URL environment variable
//! to overwrite the default URL (http://localhost:3000).
use anyhow::Result;
use swiftide::traits::SimplePrompt;
use tracing::level_filters::LevelFilter;
use tracing_subscriber::{
    EnvFilter, Layer as _, layer::SubscriberExt as _, util::SubscriberInitExt as _,
};
 
#[tokio::main]
async fn main() -> Result<()> {
    println!("Hello, langfuse!");
 
    let fmt_layer = tracing_subscriber::fmt::layer()
        .compact()
        .with_target(false)
        .boxed();
 
    let langfuse_layer = swiftide::langfuse::LangfuseLayer::default()
        .with_filter(LevelFilter::DEBUG)
        .boxed();
 
    let registry = tracing_subscriber::registry()
        .with(EnvFilter::from_default_env())
        .with(vec![fmt_layer, langfuse_layer]);
 
    registry.init();
 
    prompt_openai().await?;
 
    Ok(())
}
 
#[tracing::instrument]
async fn prompt_openai() -> Result<()> {
    let openai = swiftide::integrations::openai::OpenAI::builder()
        .default_prompt_model("gpt-5")
        .build()
        .unwrap();
 
    let paris = openai
        .prompt("What is the capital of France?".into())
        .await?;
 
    println!("The capital of France is {paris}");
 
    Ok(())
}

Running this program will create a Langfuse trace for the prompt, including input, output, usage, and metadata.

Learn More

For more details on using Swiftide, check out the following resources:

Strands Agents Vercel AI SDK

Was this page helpful?

Support