rig/providers/

hyperbolic.rs

//! Hyperbolic Inference API client and Rig integration
//!
//! # Example
//! ```
//! use rig::providers::hyperbolic;
//!
//! let client = hyperbolic::Client::new("YOUR_API_KEY");
//!
//! let llama_3_1_8b = client.completion_model(hyperbolic::LLAMA_3_1_8B);
//! ```
use super::openai::{AssistantContent, send_compatible_streaming_request};

use crate::client::{self, Capabilities, Capable, DebugExt, Nothing, Provider, ProviderBuilder};
use crate::client::{BearerAuth, ProviderClient};
use crate::http_client::{self, HttpClientExt};
use crate::streaming::StreamingCompletionResponse;

use crate::providers::openai;
use crate::{
    OneOrMany,
    completion::{self, CompletionError, CompletionRequest},
    json_utils,
    providers::openai::Message,
};
use serde::{Deserialize, Serialize};

// ================================================================
// Main Hyperbolic Client
// ================================================================
const HYPERBOLIC_API_BASE_URL: &str = "https://api.hyperbolic.xyz";

#[derive(Debug, Default, Clone, Copy)]
pub struct HyperbolicExt;
#[derive(Debug, Default, Clone, Copy)]
pub struct HyperbolicBuilder;

type HyperbolicApiKey = BearerAuth;

impl Provider for HyperbolicExt {
    type Builder = HyperbolicBuilder;

    const VERIFY_PATH: &'static str = "/models";

    fn build<H>(
        _: &crate::client::ClientBuilder<
            Self::Builder,
            <Self::Builder as crate::client::ProviderBuilder>::ApiKey,
            H,
        >,
    ) -> http_client::Result<Self> {
        Ok(Self)
    }
}

impl<H> Capabilities<H> for HyperbolicExt {
    type Completion = Capable<CompletionModel<H>>;
    type Embeddings = Nothing;
    type Transcription = Nothing;
    #[cfg(feature = "image")]
    type ImageGeneration = Capable<ImageGenerationModel<H>>;
    #[cfg(feature = "audio")]
    type AudioGeneration = Capable<AudioGenerationModel<H>>;
}

impl DebugExt for HyperbolicExt {}

impl ProviderBuilder for HyperbolicBuilder {
    type Output = HyperbolicExt;
    type ApiKey = HyperbolicApiKey;

    const BASE_URL: &'static str = HYPERBOLIC_API_BASE_URL;
}

pub type Client<H = reqwest::Client> = client::Client<HyperbolicExt, H>;
pub type ClientBuilder<H = reqwest::Client> = client::ClientBuilder<HyperbolicBuilder, String, H>;

impl ProviderClient for Client {
    type Input = HyperbolicApiKey;

    /// Create a new Hyperbolic client from the `HYPERBOLIC_API_KEY` environment variable.
    /// Panics if the environment variable is not set.
    fn from_env() -> Self {
        let api_key = std::env::var("HYPERBOLIC_API_KEY").expect("HYPERBOLIC_API_KEY not set");
        Self::new(&api_key).unwrap()
    }

    fn from_val(input: Self::Input) -> Self {
        Self::new(input).unwrap()
    }
}

#[derive(Debug, Deserialize)]
struct ApiErrorResponse {
    message: String,
}

#[derive(Debug, Deserialize)]
#[serde(untagged)]
enum ApiResponse<T> {
    Ok(T),
    Err(ApiErrorResponse),
}

#[derive(Debug, Deserialize)]
pub struct EmbeddingData {
    pub object: String,
    pub embedding: Vec<f64>,
    pub index: usize,
}

#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Usage {
    pub prompt_tokens: usize,
    pub total_tokens: usize,
}

impl std::fmt::Display for Usage {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        write!(
            f,
            "Prompt tokens: {} Total tokens: {}",
            self.prompt_tokens, self.total_tokens
        )
    }
}

// ================================================================
// Hyperbolic Completion API
// ================================================================

/// Meta Llama 3.1b Instruct model with 8B parameters.
pub const LLAMA_3_1_8B: &str = "meta-llama/Meta-Llama-3.1-8B-Instruct";
/// Meta Llama 3.3b Instruct model with 70B parameters.
pub const LLAMA_3_3_70B: &str = "meta-llama/Llama-3.3-70B-Instruct";
/// Meta Llama 3.1b Instruct model with 70B parameters.
pub const LLAMA_3_1_70B: &str = "meta-llama/Meta-Llama-3.1-70B-Instruct";
/// Meta Llama 3 Instruct model with 70B parameters.
pub const LLAMA_3_70B: &str = "meta-llama/Meta-Llama-3-70B-Instruct";
/// Hermes 3 Instruct model with 70B parameters.
pub const HERMES_3_70B: &str = "NousResearch/Hermes-3-Llama-3.1-70b";
/// Deepseek v2.5 model.
pub const DEEPSEEK_2_5: &str = "deepseek-ai/DeepSeek-V2.5";
/// Qwen 2.5 model with 72B parameters.
pub const QWEN_2_5_72B: &str = "Qwen/Qwen2.5-72B-Instruct";
/// Meta Llama 3.2b Instruct model with 3B parameters.
pub const LLAMA_3_2_3B: &str = "meta-llama/Llama-3.2-3B-Instruct";
/// Qwen 2.5 Coder Instruct model with 32B parameters.
pub const QWEN_2_5_CODER_32B: &str = "Qwen/Qwen2.5-Coder-32B-Instruct";
/// Preview (latest) version of Qwen model with 32B parameters.
pub const QWEN_QWQ_PREVIEW_32B: &str = "Qwen/QwQ-32B-Preview";
/// Deepseek R1 Zero model.
pub const DEEPSEEK_R1_ZERO: &str = "deepseek-ai/DeepSeek-R1-Zero";
/// Deepseek R1 model.
pub const DEEPSEEK_R1: &str = "deepseek-ai/DeepSeek-R1";

/// A Hyperbolic completion object.
///
/// For more information, see this link: <https://docs.hyperbolic.xyz/reference/create_chat_completion_v1_chat_completions_post>
#[derive(Debug, Deserialize, Serialize)]
pub struct CompletionResponse {
    pub id: String,
    pub object: String,
    pub created: u64,
    pub model: String,
    pub choices: Vec<Choice>,
    pub usage: Option<Usage>,
}

impl From<ApiErrorResponse> for CompletionError {
    fn from(err: ApiErrorResponse) -> Self {
        CompletionError::ProviderError(err.message)
    }
}

impl TryFrom<CompletionResponse> for completion::CompletionResponse<CompletionResponse> {
    type Error = CompletionError;

    fn try_from(response: CompletionResponse) -> Result<Self, Self::Error> {
        let choice = response.choices.first().ok_or_else(|| {
            CompletionError::ResponseError("Response contained no choices".to_owned())
        })?;

        let content = match &choice.message {
            Message::Assistant {
                content,
                tool_calls,
                ..
            } => {
                let mut content = content
                    .iter()
                    .map(|c| match c {
                        AssistantContent::Text { text } => completion::AssistantContent::text(text),
                        AssistantContent::Refusal { refusal } => {
                            completion::AssistantContent::text(refusal)
                        }
                    })
                    .collect::<Vec<_>>();

                content.extend(
                    tool_calls
                        .iter()
                        .map(|call| {
                            completion::AssistantContent::tool_call(
                                &call.id,
                                &call.function.name,
                                call.function.arguments.clone(),
                            )
                        })
                        .collect::<Vec<_>>(),
                );
                Ok(content)
            }
            _ => Err(CompletionError::ResponseError(
                "Response did not contain a valid message or tool call".into(),
            )),
        }?;

        let choice = OneOrMany::many(content).map_err(|_| {
            CompletionError::ResponseError(
                "Response contained no message or tool call (empty)".to_owned(),
            )
        })?;

        let usage = response
            .usage
            .as_ref()
            .map(|usage| completion::Usage {
                input_tokens: usage.prompt_tokens as u64,
                output_tokens: (usage.total_tokens - usage.prompt_tokens) as u64,
                total_tokens: usage.total_tokens as u64,
            })
            .unwrap_or_default();

        Ok(completion::CompletionResponse {
            choice,
            usage,
            raw_response: response,
        })
    }
}

#[derive(Debug, Deserialize, Serialize)]
pub struct Choice {
    pub index: usize,
    pub message: Message,
    pub finish_reason: String,
}

#[derive(Debug, Serialize, Deserialize)]
pub(super) struct HyperbolicCompletionRequest {
    model: String,
    pub messages: Vec<Message>,
    #[serde(flatten, skip_serializing_if = "Option::is_none")]
    temperature: Option<f64>,
    #[serde(flatten, skip_serializing_if = "Option::is_none")]
    pub additional_params: Option<serde_json::Value>,
}

impl TryFrom<(&str, CompletionRequest)> for HyperbolicCompletionRequest {
    type Error = CompletionError;

    fn try_from((model, req): (&str, CompletionRequest)) -> Result<Self, Self::Error> {
        if req.tool_choice.is_some() {
            tracing::warn!("WARNING: `tool_choice` not supported on Hyperbolic");
        }

        if !req.tools.is_empty() {
            tracing::warn!("WARNING: `tools` not supported on Hyperbolic");
        }

        let mut full_history: Vec<Message> = match &req.preamble {
            Some(preamble) => vec![Message::system(preamble)],
            None => vec![],
        };

        if let Some(docs) = req.normalized_documents() {
            let docs: Vec<Message> = docs.try_into()?;
            full_history.extend(docs);
        }

        let chat_history: Vec<Message> = req
            .chat_history
            .clone()
            .into_iter()
            .map(|message| message.try_into())
            .collect::<Result<Vec<Vec<Message>>, _>>()?
            .into_iter()
            .flatten()
            .collect();

        full_history.extend(chat_history);

        Ok(Self {
            model: model.to_string(),
            messages: full_history,
            temperature: req.temperature,
            additional_params: req.additional_params,
        })
    }
}

#[derive(Clone)]
pub struct CompletionModel<T = reqwest::Client> {
    client: Client<T>,
    /// Name of the model (e.g.: deepseek-ai/DeepSeek-R1)
    pub model: String,
}

impl<T> CompletionModel<T> {
    pub fn new(client: Client<T>, model: impl Into<String>) -> Self {
        Self {
            client,
            model: model.into(),
        }
    }

    pub fn with_model(client: Client<T>, model: &str) -> Self {
        Self {
            client,
            model: model.into(),
        }
    }
}

impl<T> completion::CompletionModel for CompletionModel<T>
where
    T: HttpClientExt + Clone + Default + std::fmt::Debug + Send + 'static,
{
    type Response = CompletionResponse;
    type StreamingResponse = openai::StreamingCompletionResponse;

    type Client = Client<T>;

    fn make(client: &Self::Client, model: impl Into<String>) -> Self {
        Self::new(client.clone(), model)
    }

    #[cfg_attr(feature = "worker", worker::send)]
    async fn completion(
        &self,
        completion_request: CompletionRequest,
    ) -> Result<completion::CompletionResponse<CompletionResponse>, CompletionError> {
        let preamble = completion_request.preamble.clone();
        let request =
            HyperbolicCompletionRequest::try_from((self.model.as_ref(), completion_request))?;
        let body = serde_json::to_vec(&request)?;

        let span = if tracing::Span::current().is_disabled() {
            info_span!(
                target: "rig::completions",
                "chat",
                gen_ai.operation.name = "chat",
                gen_ai.provider.name = "hyperbolic",
                gen_ai.request.model = self.model,
                gen_ai.system_instructions = preamble,
                gen_ai.response.id = tracing::field::Empty,
                gen_ai.response.model = tracing::field::Empty,
                gen_ai.usage.output_tokens = tracing::field::Empty,
                gen_ai.usage.input_tokens = tracing::field::Empty,
                gen_ai.input.messages = serde_json::to_string(&request.messages)?,
                gen_ai.output.messages = tracing::field::Empty,
            )
        } else {
            tracing::Span::current()
        };

        let req = self
            .client
            .post("/v1/chat/completions")?
            .body(body)
            .map_err(http_client::Error::from)?;

        let async_block = async move {
            let response = self.client.send::<_, bytes::Bytes>(req).await?;

            let status = response.status();
            let response_body = response.into_body().into_future().await?.to_vec();

            if status.is_success() {
                match serde_json::from_slice::<ApiResponse<CompletionResponse>>(&response_body)? {
                    ApiResponse::Ok(response) => {
                        tracing::info!(target: "rig",
                            "Hyperbolic completion token usage: {:?}",
                            response.usage.clone().map(|usage| format!("{usage}")).unwrap_or("N/A".to_string())
                        );

                        response.try_into()
                    }
                    ApiResponse::Err(err) => Err(CompletionError::ProviderError(err.message)),
                }
            } else {
                Err(CompletionError::ProviderError(
                    String::from_utf8_lossy(&response_body).to_string(),
                ))
            }
        };

        async_block.instrument(span).await
    }

    #[cfg_attr(feature = "worker", worker::send)]
    async fn stream(
        &self,
        completion_request: CompletionRequest,
    ) -> Result<StreamingCompletionResponse<Self::StreamingResponse>, CompletionError> {
        let preamble = completion_request.preamble.clone();
        let mut request =
            HyperbolicCompletionRequest::try_from((self.model.as_ref(), completion_request))?;

        let span = if tracing::Span::current().is_disabled() {
            info_span!(
                target: "rig::completions",
                "chat_streaming",
                gen_ai.operation.name = "chat_streaming",
                gen_ai.provider.name = "hyperbolic",
                gen_ai.request.model = self.model,
                gen_ai.system_instructions = preamble,
                gen_ai.response.id = tracing::field::Empty,
                gen_ai.response.model = tracing::field::Empty,
                gen_ai.usage.output_tokens = tracing::field::Empty,
                gen_ai.usage.input_tokens = tracing::field::Empty,
                gen_ai.input.messages = serde_json::to_string(&request.messages)?,
                gen_ai.output.messages = tracing::field::Empty,
            )
        } else {
            tracing::Span::current()
        };

        let params = json_utils::merge(
            request.additional_params.unwrap_or(serde_json::json!({})),
            serde_json::json!({"stream": true, "stream_options": {"include_usage": true} }),
        );

        request.additional_params = Some(params);

        let body = serde_json::to_vec(&request)?;

        let req = self
            .client
            .post("/v1/chat/completions")?
            .body(body)
            .map_err(http_client::Error::from)?;

        send_compatible_streaming_request(self.client.http_client().clone(), req)
            .instrument(span)
            .await
    }
}

// =======================================
// Hyperbolic Image Generation API
// =======================================

#[cfg(feature = "image")]
pub use image_generation::*;

#[cfg(feature = "image")]
#[cfg_attr(docsrs, doc(cfg(feature = "image")))]
mod image_generation {
    use super::{ApiResponse, Client};
    use crate::http_client::HttpClientExt;
    use crate::image_generation;
    use crate::image_generation::{ImageGenerationError, ImageGenerationRequest};
    use crate::json_utils::merge_inplace;
    use base64::Engine;
    use base64::prelude::BASE64_STANDARD;
    use serde::Deserialize;
    use serde_json::json;

    pub const SDXL1_0_BASE: &str = "SDXL1.0-base";
    pub const SD2: &str = "SD2";
    pub const SD1_5: &str = "SD1.5";
    pub const SSD: &str = "SSD";
    pub const SDXL_TURBO: &str = "SDXL-turbo";
    pub const SDXL_CONTROLNET: &str = "SDXL-ControlNet";
    pub const SD1_5_CONTROLNET: &str = "SD1.5-ControlNet";

    #[derive(Clone)]
    pub struct ImageGenerationModel<T> {
        client: Client<T>,
        pub model: String,
    }

    impl<T> ImageGenerationModel<T> {
        pub(crate) fn new(client: Client<T>, model: impl Into<String>) -> Self {
            Self {
                client,
                model: model.into(),
            }
        }

        pub fn with_model(client: Client<T>, model: &str) -> Self {
            Self {
                client,
                model: model.into(),
            }
        }
    }

    #[derive(Clone, Deserialize)]
    pub struct Image {
        image: String,
    }

    #[derive(Clone, Deserialize)]
    pub struct ImageGenerationResponse {
        images: Vec<Image>,
    }

    impl TryFrom<ImageGenerationResponse>
        for image_generation::ImageGenerationResponse<ImageGenerationResponse>
    {
        type Error = ImageGenerationError;

        fn try_from(value: ImageGenerationResponse) -> Result<Self, Self::Error> {
            let data = BASE64_STANDARD
                .decode(&value.images[0].image)
                .expect("Could not decode image.");

            Ok(Self {
                image: data,
                response: value,
            })
        }
    }

    impl<T> image_generation::ImageGenerationModel for ImageGenerationModel<T>
    where
        T: HttpClientExt + Clone + Default + std::fmt::Debug + Send + 'static,
    {
        type Response = ImageGenerationResponse;

        type Client = Client<T>;

        fn make(client: &Self::Client, model: impl Into<String>) -> Self {
            Self::new(client.clone(), model)
        }

        #[cfg_attr(feature = "worker", worker::send)]
        async fn image_generation(
            &self,
            generation_request: ImageGenerationRequest,
        ) -> Result<image_generation::ImageGenerationResponse<Self::Response>, ImageGenerationError>
        {
            let mut request = json!({
                "model_name": self.model,
                "prompt": generation_request.prompt,
                "height": generation_request.height,
                "width": generation_request.width,
            });

            if let Some(params) = generation_request.additional_params {
                merge_inplace(&mut request, params);
            }

            let body = serde_json::to_vec(&request)?;

            let request = self
                .client
                .post("/v1/image/generation")?
                .header("Content-Type", "application/json")
                .body(body)
                .map_err(|e| ImageGenerationError::HttpError(e.into()))?;

            let response = self.client.send::<_, bytes::Bytes>(request).await?;

            let status = response.status();
            let response_body = response.into_body().into_future().await?.to_vec();

            if !status.is_success() {
                return Err(ImageGenerationError::ProviderError(format!(
                    "{status}: {}",
                    String::from_utf8_lossy(&response_body)
                )));
            }

            match serde_json::from_slice::<ApiResponse<ImageGenerationResponse>>(&response_body)? {
                ApiResponse::Ok(response) => response.try_into(),
                ApiResponse::Err(err) => Err(ImageGenerationError::ResponseError(err.message)),
            }
        }
    }
}

// ======================================
// Hyperbolic Audio Generation API
// ======================================
#[cfg(feature = "audio")]
pub use audio_generation::*;
use tracing::{Instrument, info_span};

#[cfg(feature = "audio")]
#[cfg_attr(docsrs, doc(cfg(feature = "image")))]
mod audio_generation {
    use super::{ApiResponse, Client};
    use crate::audio_generation;
    use crate::audio_generation::{AudioGenerationError, AudioGenerationRequest};
    use crate::http_client::{self, HttpClientExt};
    use base64::Engine;
    use base64::prelude::BASE64_STANDARD;
    use bytes::Bytes;
    use serde::Deserialize;
    use serde_json::json;

    #[derive(Clone)]
    pub struct AudioGenerationModel<T> {
        client: Client<T>,
        pub language: String,
    }

    #[derive(Clone, Deserialize)]
    pub struct AudioGenerationResponse {
        audio: String,
    }

    impl TryFrom<AudioGenerationResponse>
        for audio_generation::AudioGenerationResponse<AudioGenerationResponse>
    {
        type Error = AudioGenerationError;

        fn try_from(value: AudioGenerationResponse) -> Result<Self, Self::Error> {
            let data = BASE64_STANDARD
                .decode(&value.audio)
                .expect("Could not decode audio.");

            Ok(Self {
                audio: data,
                response: value,
            })
        }
    }

    impl<T> audio_generation::AudioGenerationModel for AudioGenerationModel<T>
    where
        T: HttpClientExt + Clone + Default + std::fmt::Debug + Send + 'static,
    {
        type Response = AudioGenerationResponse;
        type Client = Client<T>;

        fn make(client: &Self::Client, language: impl Into<String>) -> Self {
            Self {
                client: client.clone(),
                language: language.into(),
            }
        }

        #[cfg_attr(feature = "worker", worker::send)]
        async fn audio_generation(
            &self,
            request: AudioGenerationRequest,
        ) -> Result<audio_generation::AudioGenerationResponse<Self::Response>, AudioGenerationError>
        {
            let request = json!({
                "language": self.language,
                "speaker": request.voice,
                "text": request.text,
                "speed": request.speed
            });

            let body = serde_json::to_vec(&request)?;

            let req = self
                .client
                .post("/v1/audio/generation")?
                .body(body)
                .map_err(http_client::Error::from)?;

            let response = self.client.send::<_, Bytes>(req).await?;
            let status = response.status();
            let response_body = response.into_body().into_future().await?.to_vec();

            if !status.is_success() {
                return Err(AudioGenerationError::ProviderError(format!(
                    "{status}: {}",
                    String::from_utf8_lossy(&response_body)
                )));
            }

            match serde_json::from_slice::<ApiResponse<AudioGenerationResponse>>(&response_body)? {
                ApiResponse::Ok(response) => response.try_into(),
                ApiResponse::Err(err) => Err(AudioGenerationError::ProviderError(err.message)),
            }
        }
    }
}