​
Latitude Community SDK Documentation

The Latitude Rust SDK provides an easy way to interact with the Latitude API, allowing you to run documents and chat with AI models.

This SDK is community-maintained and not officially supported by Latitude. If you have any questions or requests, please reach out to the community on their GitHub repository.

To install the Latitude SDK: Crates.io

cargo add latitude-sdk

​
Getting Started

First, import the Client Struct from the SDK:

use latitude_sdk::{Client};

Then, create an instance of the Latitude class with your API key:

let client = Client::builder("your-api-key-here".into()).project_id(12345).build();

​
Running a Document

To run a document, use the run method:

use latitude_sdk::{models::document::RunDocument, Client};
use serde::{Deserialize, Serialize};

//only needed when the prompt has parameters
#[derive(Debug, Deserialize, Serialize, Default)]
struct Params {
    user_message: String,
}

#[tokio::main]
async fn main() {
    let client = Client::builder("<api_key>".into()).project_id(12345).build();

    //only needed when the prompt has parameters
    let params = Params {
        user_message: "".into(),
    };

    let document = RunDocument::builder()
        .path("<document_path>".to_owned())
        .parameters(params)
        .build()
        .unwrap();

    //in case the prompt has no parameters
    let document_without_params = RunDocument::<()>::builder()
        .path("<document_path>".to_owned())
        .build()
        .unwrap();

    let result = client.run(document).await.unwrap();
    let result_without_params = client.run(document_without_params).await.unwrap();
}

​
Chatting with an AI Model

The document run method previously described returns events which all contain a singular uuid field. This field can be used to further continue the conversation with the document, including the context from the document run. Here’s how to do it.

To continue a chat conversation, use the chat method:

The Chat Method currently only Supports Stream Response!

use latitude_sdk::models::chat::Chat;
use latitude_sdk::models::event::Event;
use latitude_sdk::models::message::{Message, Role};
use latitude_sdk::models::response::Response;
use latitude_sdk::Client;
use std::error::Error;

#[tokio::main]
async fn main() -> Result<(), Box<dyn Error>> {
    let client = Client::builder("<api_key>".into())
        .project_id(12345)
        .build();

    let chat = Chat::builder()
        .conversation_id("30cf5e50-978c-4c6b-9d61-e278e8d38692".into())
        .add_message(
            Message::builder()
                .role(Role::User)
                .add_content("text", "Another Joke!")
                .build()
                .unwrap(),
        )
        .stream()
        .build()
        .unwrap();

    match client.chat(chat).await {
        Ok(res) => match res {
            Response::Json(response) => {
                println!("JSON Response: {:?}", response);
            }
            Response::Stream(mut event_stream) => {
                while let Some(event) = event_stream.recv().await {
                    match event {
                        Event::LatitudeEvent(data) => {
                            println!("Latitude Event: {:?}", data);
                        }
                        Event::ProviderEvent(data) => {
                            println!("Provider Event: {:?}", data);
                        }
                        Event::UnknownEvent => todo!(),
                    }
                }
            }
        },
        Err(e) => eprintln!("Error executing the Chat: {:?}", e),
    }

    Ok(())
}

​
Stream Handling

The run and chat methods in this library both support streaming responses. When using these methods, the response can be handled as a stream of events in real-time. Here’s how you can work with streaming in Rust:

  • Stream Responses: By calling .stream() on your chat or run request, the library will return events as they are received. The stream can be iterated over asynchronously, allowing you to handle each event as it arrives.

Example:

let document_request = RunDocument::builder()
    .stream()
    .unwrap();

let chat_request = Chat::builder()
    .stream()
    .unwrap();

//chat or run request
match client.chat(chat).await {
    Ok(Response::Stream(mut event_stream)) => {
        while let Some(event) = event_stream.recv().await {
            match event {
                Event::LatitudeEvent(data) => {
                    println!("Latitude Event: {:?}", data);
                }
                Event::ProviderEvent(data) => {
                    println!("Provider Event: {:?}", data);
                }
                Event::UnknownEvent => {
                    println!("Received an unknown event.");
                }
            }
        }
    }
    _ => {}
}

​
Error Handling

Errors can occur during both the initialization of a chat or run request and while handling streamed events. It’s recommended to handle errors using Rust’s standard Result and match pattern.

  • Error Callback: If an error occurs, the library returns an Err variant containing the error details. It is good practice to handle errors immediately after calling run or chat.

Example:

match client.chat(chat).await {
    Ok(res) => match res {
        Response::Json(response) => {
            println!("JSON Response: {:?}", response);
        }
        Response::Stream(mut event_stream) => {
            while let Some(event) = event_stream.recv().await {
                match event {
                    Event::LatitudeEvent(data) => println!("Latitude Event: {:?}", data),
                    Event::ProviderEvent(data) => println!("Provider Event: {:?}", data),
                    Event::UnknownEvent => println!("Received an unknown event."),
                }
            }
        }
    },
    Err(e) => eprintln!("Error executing the Chat: {:?}", e),
}

In both cases, error messages can be logged or handled as needed. This ensures stability and gives insight into issues when handling streaming data.

​
Pushing a log to Latitude

You can push a log to Latitude in order to evaluate it, using the log method:

use latitude_sdk::models::log::Log;
use latitude_sdk::models::message::{Message, Role};
use latitude_sdk::Client;

#[tokio::main]
async fn main() {
    let client = Client::builder("<api_key>".into())
        .project_id(12345)
        .build();

    //do stuff with any OpenAI Lib you wish, i recommend using async-openai https://crates.io/crates/async-openai

    let log = Log::builder()
        .path("joker")
        .response("response-from-ai")
        .add_message(
            Message::builder()
                .role(Role::User)
                .add_content("text", "Please tell me a joke about doctors")
                .build()
                .unwrap(),
        )
        .build()
        .unwrap();

    let log = client.log(log).await.unwrap();
}

Logs follow OpenAI’s format. If you’re using a different method to run your prompts, you’ll need to format your logs accordingly.

If you include the assistant response in the optional response parameter, make sure to not include it in the log so it isn’t included twice.

​
Evaluating Conversations

To evaluate a conversation using configured evaluators, use the eval method:

use latitude_sdk::models::evaluate::Evaluation;
use latitude_sdk::Client;

#[tokio::main]
async fn main() {
    let client = Client::builder("<api_key>".into())
        .project_id(12345)
        .build();

    let eval = client
        .eval(
            "chat-uuid",
            Some(Evaluation {
                evaluation_uuids: vec![
                    Some("evaluator-uuid-1".into()),
                    Some("evaluator-uuid-2".into()),
                ],
            }),
        )
        .await
        .unwrap();
}

This allows you to evaluate a conversation at any point in time. Specially helpful when building agents that have multiple interactions with users, and you intend to evaluate the agent’s performance after the interaction is fully completed, or at particular points in time.

Vertex AIOverview