rig_core/providers/

deepseek.rs

//! DeepSeek API client and Rig integration
//!
//! # Example
//! ```no_run
//! use rig_core::{client::CompletionClient, providers::deepseek};
//!
//! # fn run() -> Result<(), Box<dyn std::error::Error>> {
//! let client = deepseek::Client::new("DEEPSEEK_API_KEY")?;
//!
//! let deepseek_chat = client.completion_model(deepseek::DEEPSEEK_CHAT);
//! # Ok(())
//! # }
//! ```

use bytes::Bytes;
use http::Request;
use tracing::{Instrument, Level, enabled, info_span};

use crate::client::{
    self, BearerAuth, Capabilities, Capable, DebugExt, ModelLister, Nothing, Provider,
    ProviderBuilder, ProviderClient,
};
use crate::completion::GetTokenUsage;
use crate::http_client::{self, HttpClientExt};
use crate::message::{Document, DocumentSourceKind};
use crate::model::{Model, ModelList, ModelListingError};
use crate::providers::internal::openai_chat_completions_compatible::{
    self, CompatibleChoiceData, CompatibleChunk, CompatibleFinishReason, CompatibleStreamProfile,
};
use crate::{
    OneOrMany,
    completion::{self, CompletionError, CompletionRequest},
    json_utils, message,
    wasm_compat::{WasmCompatSend, WasmCompatSync},
};
use serde::{Deserialize, Serialize};

use super::openai::StreamingToolCall;

// ================================================================
// Main DeepSeek Client
// ================================================================
const DEEPSEEK_API_BASE_URL: &str = "https://api.deepseek.com";

#[derive(Debug, Default, Clone, Copy)]
pub struct DeepSeekExt;
#[derive(Debug, Default, Clone, Copy)]
pub struct DeepSeekExtBuilder;

type DeepSeekApiKey = BearerAuth;

impl Provider for DeepSeekExt {
    type Builder = DeepSeekExtBuilder;
    const VERIFY_PATH: &'static str = "/user/balance";
}

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

impl DebugExt for DeepSeekExt {}

impl ProviderBuilder for DeepSeekExtBuilder {
    type Extension<H>
        = DeepSeekExt
    where
        H: HttpClientExt;
    type ApiKey = DeepSeekApiKey;

    const BASE_URL: &'static str = DEEPSEEK_API_BASE_URL;

    fn build<H>(
        _builder: &client::ClientBuilder<Self, Self::ApiKey, H>,
    ) -> http_client::Result<Self::Extension<H>>
    where
        H: HttpClientExt,
    {
        Ok(DeepSeekExt)
    }
}

pub type Client<H = reqwest::Client> = client::Client<DeepSeekExt, H>;
pub type ClientBuilder<H = crate::markers::Missing> =
    client::ClientBuilder<DeepSeekExtBuilder, String, H>;

impl ProviderClient for Client {
    type Input = DeepSeekApiKey;
    type Error = crate::client::ProviderClientError;

    // If you prefer the environment variable approach:
    fn from_env() -> Result<Self, Self::Error> {
        let api_key = crate::client::required_env_var("DEEPSEEK_API_KEY")?;
        let mut client_builder = Self::builder();
        client_builder.headers_mut().insert(
            http::header::CONTENT_TYPE,
            http::HeaderValue::from_static("application/json"),
        );
        let client_builder = client_builder.api_key(&api_key);
        client_builder.build().map_err(Into::into)
    }

    fn from_val(input: Self::Input) -> Result<Self, Self::Error> {
        Self::new(input).map_err(Into::into)
    }
}

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

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

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

/// The response shape from the DeepSeek API
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct CompletionResponse {
    // We'll match the JSON:
    pub choices: Vec<Choice>,
    pub usage: Usage,
    // you may want other fields
}

#[derive(Clone, Debug, Serialize, Deserialize, Default)]
pub struct Usage {
    pub completion_tokens: u32,
    pub prompt_tokens: u32,
    pub prompt_cache_hit_tokens: u32,
    pub prompt_cache_miss_tokens: u32,
    pub total_tokens: u32,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub completion_tokens_details: Option<CompletionTokensDetails>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub prompt_tokens_details: Option<PromptTokensDetails>,
}

impl GetTokenUsage for Usage {
    fn token_usage(&self) -> crate::completion::Usage {
        crate::providers::internal::completion_usage(
            self.prompt_tokens as u64,
            self.completion_tokens as u64,
            self.total_tokens as u64,
            self.prompt_tokens_details
                .as_ref()
                .and_then(|details| details.cached_tokens)
                .map(u64::from)
                .unwrap_or(0),
        )
    }
}

#[derive(Clone, Debug, Serialize, Deserialize, Default)]
pub struct CompletionTokensDetails {
    #[serde(skip_serializing_if = "Option::is_none")]
    pub reasoning_tokens: Option<u32>,
}

#[derive(Clone, Debug, Serialize, Deserialize, Default)]
pub struct PromptTokensDetails {
    #[serde(skip_serializing_if = "Option::is_none")]
    pub cached_tokens: Option<u32>,
}

#[derive(Clone, Debug, Serialize, Deserialize, PartialEq)]
pub struct Choice {
    pub index: usize,
    pub message: Message,
    pub logprobs: Option<serde_json::Value>,
    pub finish_reason: String,
}

#[derive(Debug, Serialize, Deserialize, PartialEq, Clone)]
#[serde(tag = "role", rename_all = "lowercase")]
pub enum Message {
    System {
        content: String,
        #[serde(skip_serializing_if = "Option::is_none")]
        name: Option<String>,
    },
    User {
        content: String,
        #[serde(skip_serializing_if = "Option::is_none")]
        name: Option<String>,
    },
    Assistant {
        content: String,
        #[serde(skip_serializing_if = "Option::is_none")]
        name: Option<String>,
        #[serde(
            default,
            deserialize_with = "json_utils::null_or_vec",
            skip_serializing_if = "Vec::is_empty"
        )]
        tool_calls: Vec<ToolCall>,
        /// only exists on `deepseek-reasoner` model at time of addition
        #[serde(skip_serializing_if = "Option::is_none")]
        reasoning_content: Option<String>,
    },
    #[serde(rename = "tool")]
    ToolResult {
        tool_call_id: String,
        content: String,
    },
}

impl Message {
    pub fn system(content: &str) -> Self {
        Message::System {
            content: content.to_owned(),
            name: None,
        }
    }
}

impl From<message::ToolResult> for Message {
    fn from(tool_result: message::ToolResult) -> Self {
        let content = match tool_result.content.first() {
            message::ToolResultContent::Text(text) => text.text,
            message::ToolResultContent::Image(_) => String::from("[Image]"),
        };

        Message::ToolResult {
            tool_call_id: tool_result.id,
            content,
        }
    }
}

impl From<message::ToolCall> for ToolCall {
    fn from(tool_call: message::ToolCall) -> Self {
        Self {
            id: tool_call.id,
            // TODO: update index when we have it
            index: 0,
            r#type: ToolType::Function,
            function: Function {
                name: tool_call.function.name,
                arguments: tool_call.function.arguments,
            },
        }
    }
}

impl TryFrom<message::Message> for Vec<Message> {
    type Error = message::MessageError;

    fn try_from(message: message::Message) -> Result<Self, Self::Error> {
        match message {
            message::Message::System { content } => Ok(vec![Message::System {
                content,
                name: None,
            }]),
            message::Message::User { content } => {
                // extract tool results
                let mut messages = vec![];

                let tool_results = content
                    .clone()
                    .into_iter()
                    .filter_map(|content| match content {
                        message::UserContent::ToolResult(tool_result) => {
                            Some(Message::from(tool_result))
                        }
                        _ => None,
                    })
                    .collect::<Vec<_>>();

                messages.extend(tool_results);

                let text_content: String = content
                    .into_iter()
                    .filter_map(|content| match content {
                        message::UserContent::Text(text) => Some(text.text),
                        message::UserContent::Document(Document {
                            data:
                                DocumentSourceKind::Base64(content)
                                | DocumentSourceKind::String(content),
                            ..
                        }) => Some(content),
                        _ => None,
                    })
                    .collect::<Vec<_>>()
                    .join("\n");

                if !text_content.is_empty() {
                    messages.push(Message::User {
                        content: text_content,
                        name: None,
                    });
                }

                Ok(messages)
            }
            message::Message::Assistant { content, .. } => {
                let mut text_content = String::new();
                let mut reasoning_content = String::new();
                let mut tool_calls = Vec::new();

                for item in content.iter() {
                    match item {
                        message::AssistantContent::Text(text) => {
                            text_content.push_str(text.text());
                        }
                        message::AssistantContent::Reasoning(reasoning) => {
                            reasoning_content.push_str(&reasoning.display_text());
                        }
                        message::AssistantContent::ToolCall(tool_call) => {
                            tool_calls.push(ToolCall::from(tool_call.clone()));
                        }
                        _ => {}
                    }
                }

                let reasoning = if reasoning_content.is_empty() {
                    None
                } else {
                    Some(reasoning_content)
                };

                Ok(vec![Message::Assistant {
                    content: text_content,
                    name: None,
                    tool_calls,
                    reasoning_content: reasoning,
                }])
            }
        }
    }
}

#[derive(Debug, Serialize, Deserialize, PartialEq, Clone)]
pub struct ToolCall {
    pub id: String,
    pub index: usize,
    #[serde(default)]
    pub r#type: ToolType,
    pub function: Function,
}

#[derive(Debug, Serialize, Deserialize, PartialEq, Clone)]
pub struct Function {
    pub name: String,
    #[serde(with = "json_utils::stringified_json")]
    pub arguments: serde_json::Value,
}

#[derive(Default, Debug, Serialize, Deserialize, PartialEq, Clone)]
#[serde(rename_all = "lowercase")]
pub enum ToolType {
    #[default]
    Function,
}

#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct ToolDefinition {
    pub r#type: String,
    pub function: completion::ToolDefinition,
}

impl From<crate::completion::ToolDefinition> for ToolDefinition {
    fn from(tool: crate::completion::ToolDefinition) -> Self {
        Self {
            r#type: "function".into(),
            function: tool,
        }
    }
}

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,
                reasoning_content,
                ..
            } => {
                let mut content = if content.trim().is_empty() {
                    vec![]
                } else {
                    vec![completion::AssistantContent::text(content)]
                };

                content.extend(
                    tool_calls
                        .iter()
                        .map(|call| {
                            completion::AssistantContent::tool_call(
                                &call.id,
                                &call.function.name,
                                call.function.arguments.clone(),
                            )
                        })
                        .collect::<Vec<_>>(),
                );

                if let Some(reasoning_content) = reasoning_content {
                    content.push(completion::AssistantContent::reasoning(reasoning_content));
                }

                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 = completion::Usage {
            input_tokens: response.usage.prompt_tokens as u64,
            output_tokens: response.usage.completion_tokens as u64,
            total_tokens: response.usage.total_tokens as u64,
            cached_input_tokens: response
                .usage
                .prompt_tokens_details
                .as_ref()
                .and_then(|d| d.cached_tokens)
                .map(|c| c as u64)
                .unwrap_or(0),
            cache_creation_input_tokens: 0,
            tool_use_prompt_tokens: 0,
            reasoning_tokens: 0,
        };

        Ok(completion::CompletionResponse {
            choice,
            usage,
            raw_response: response,
            message_id: None,
        })
    }
}

#[derive(Debug, Serialize, Deserialize)]
pub(super) struct DeepseekCompletionRequest {
    model: String,
    pub messages: Vec<Message>,
    #[serde(skip_serializing_if = "Option::is_none")]
    temperature: Option<f64>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    tools: Vec<ToolDefinition>,
    #[serde(skip_serializing_if = "Option::is_none")]
    tool_choice: Option<crate::providers::openrouter::ToolChoice>,
    #[serde(flatten, skip_serializing_if = "Option::is_none")]
    pub additional_params: Option<serde_json::Value>,
}

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

    fn try_from((model, req): (&str, CompletionRequest)) -> Result<Self, Self::Error> {
        let chat_history = req.chat_history_with_documents();
        if req.output_schema.is_some() {
            tracing::warn!("Structured outputs currently not supported for DeepSeek");
        }
        let model = req.model.clone().unwrap_or_else(|| model.to_string());
        let mut full_history: Vec<Message> = match &req.preamble {
            Some(preamble) => vec![Message::system(preamble)],
            None => vec![],
        };

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

        full_history.extend(chat_history);

        let tool_choice = req
            .tool_choice
            .clone()
            .map(crate::providers::openrouter::ToolChoice::try_from)
            .transpose()?;

        Ok(Self {
            model: model.to_string(),
            messages: full_history,
            temperature: req.temperature,
            tools: req
                .tools
                .clone()
                .into_iter()
                .map(ToolDefinition::from)
                .collect::<Vec<_>>(),
            tool_choice,
            additional_params: req.additional_params,
        })
    }
}

/// The struct implementing the `CompletionModel` trait
#[derive(Clone)]
pub struct CompletionModel<T = reqwest::Client> {
    pub client: Client<T>,
    pub model: String,
}

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

    type Client = Client<T>;

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

    async fn completion(
        &self,
        completion_request: CompletionRequest,
    ) -> Result<
        completion::CompletionResponse<CompletionResponse>,
        crate::completion::CompletionError,
    > {
        let span = if tracing::Span::current().is_disabled() {
            info_span!(
                target: "rig::completions",
                "chat",
                gen_ai.operation.name = "chat",
                gen_ai.provider.name = "deepseek",
                gen_ai.request.model = self.model,
                gen_ai.system_instructions = tracing::field::Empty,
                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.usage.cache_read.input_tokens = tracing::field::Empty,
            )
        } else {
            tracing::Span::current()
        };

        span.record("gen_ai.system_instructions", &completion_request.preamble);

        let request =
            DeepseekCompletionRequest::try_from((self.model.as_ref(), completion_request))?;

        if enabled!(Level::TRACE) {
            tracing::trace!(target: "rig::completions",
                "DeepSeek completion request: {}",
                serde_json::to_string_pretty(&request)?
            );
        }

        let body = serde_json::to_vec(&request)?;
        let req = self
            .client
            .post("/chat/completions")?
            .body(body)
            .map_err(|e| CompletionError::HttpError(e.into()))?;

        async move {
            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() {
                match serde_json::from_slice::<ApiResponse<CompletionResponse>>(&response_body)? {
                    ApiResponse::Ok(response) => {
                        let span = tracing::Span::current();
                        span.record("gen_ai.usage.input_tokens", response.usage.prompt_tokens);
                        span.record(
                            "gen_ai.usage.output_tokens",
                            response.usage.completion_tokens,
                        );
                        span.record(
                            "gen_ai.usage.cache_read.input_tokens",
                            response
                                .usage
                                .prompt_tokens_details
                                .as_ref()
                                .and_then(|d| d.cached_tokens)
                                .unwrap_or(0),
                        );
                        if enabled!(Level::TRACE) {
                            tracing::trace!(target: "rig::completions",
                                "DeepSeek completion response: {}",
                                serde_json::to_string_pretty(&response)?
                            );
                        }
                        response.try_into()
                    }
                    ApiResponse::Err(err) => Err(CompletionError::ProviderError(err.message)),
                }
            } else {
                Err(CompletionError::ProviderError(
                    String::from_utf8_lossy(&response_body).to_string(),
                ))
            }
        }
        .instrument(span)
        .await
    }

    async fn stream(
        &self,
        completion_request: CompletionRequest,
    ) -> Result<
        crate::streaming::StreamingCompletionResponse<Self::StreamingResponse>,
        CompletionError,
    > {
        let preamble = completion_request.preamble.clone();
        let mut request =
            DeepseekCompletionRequest::try_from((self.model.as_ref(), completion_request))?;

        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);

        if enabled!(Level::TRACE) {
            tracing::trace!(target: "rig::completions",
                "DeepSeek streaming completion request: {}",
                serde_json::to_string_pretty(&request)?
            );
        }

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

        let req = self
            .client
            .post("/chat/completions")?
            .body(body)
            .map_err(|e| CompletionError::HttpError(e.into()))?;

        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 = "deepseek",
                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.usage.cache_read.input_tokens = tracing::field::Empty,
            )
        } else {
            tracing::Span::current()
        };

        tracing::Instrument::instrument(
            send_compatible_streaming_request(self.client.clone(), req),
            span,
        )
        .await
    }
}

#[derive(Deserialize, Debug)]
pub struct StreamingDelta {
    #[serde(default)]
    content: Option<String>,
    #[serde(default, deserialize_with = "json_utils::null_or_vec")]
    tool_calls: Vec<StreamingToolCall>,
    reasoning_content: Option<String>,
}

#[derive(Deserialize, Debug)]
struct StreamingChoice {
    delta: StreamingDelta,
}

#[derive(Deserialize, Debug)]
struct StreamingCompletionChunk {
    id: Option<String>,
    model: Option<String>,
    choices: Vec<StreamingChoice>,
    usage: Option<Usage>,
}

#[derive(Clone, Deserialize, Serialize, Debug)]
pub struct StreamingCompletionResponse {
    pub usage: Usage,
}

impl GetTokenUsage for StreamingCompletionResponse {
    fn token_usage(&self) -> crate::completion::Usage {
        self.usage.token_usage()
    }
}

#[derive(Clone, Copy)]
struct DeepSeekCompatibleProfile;

impl CompatibleStreamProfile for DeepSeekCompatibleProfile {
    type Usage = Usage;
    type Detail = ();
    type FinalResponse = StreamingCompletionResponse;

    fn normalize_chunk(
        &self,
        data: &str,
    ) -> Result<Option<CompatibleChunk<Self::Usage, Self::Detail>>, CompletionError> {
        let data = match serde_json::from_str::<StreamingCompletionChunk>(data) {
            Ok(data) => data,
            Err(error) => {
                tracing::debug!(
                    "Couldn't parse SSE payload as StreamingCompletionChunk: {:?}",
                    error
                );
                return Ok(None);
            }
        };

        Ok(Some(
            openai_chat_completions_compatible::normalize_first_choice_chunk(
                data.id,
                data.model,
                data.usage,
                &data.choices,
                |choice| CompatibleChoiceData {
                    finish_reason: CompatibleFinishReason::Other,
                    text: choice.delta.content.clone(),
                    reasoning: choice.delta.reasoning_content.clone(),
                    tool_calls: openai_chat_completions_compatible::tool_call_chunks(
                        &choice.delta.tool_calls,
                    ),
                    details: Vec::new(),
                },
            ),
        ))
    }

    fn build_final_response(&self, usage: Self::Usage) -> Self::FinalResponse {
        StreamingCompletionResponse { usage }
    }

    fn uses_distinct_tool_call_eviction(&self) -> bool {
        true
    }

    fn emits_complete_single_chunk_tool_calls(&self) -> bool {
        true
    }
}

pub async fn send_compatible_streaming_request<T>(
    http_client: T,
    req: Request<Vec<u8>>,
) -> Result<
    crate::streaming::StreamingCompletionResponse<StreamingCompletionResponse>,
    CompletionError,
>
where
    T: HttpClientExt + Clone + 'static,
{
    openai_chat_completions_compatible::send_compatible_streaming_request(
        http_client,
        req,
        DeepSeekCompatibleProfile,
    )
    .await
}

#[derive(Debug, Deserialize)]
struct ListModelsResponse {
    data: Vec<ListModelEntry>,
}

#[derive(Debug, Deserialize)]
struct ListModelEntry {
    id: String,
    owned_by: String,
}

impl From<ListModelEntry> for Model {
    fn from(value: ListModelEntry) -> Self {
        let mut model = Model::from_id(value.id);
        model.owned_by = Some(value.owned_by);
        model
    }
}

/// [`ModelLister`] implementation for the DeepSeek API (`GET /models`).
#[derive(Clone)]
pub struct DeepSeekModelLister<H = reqwest::Client> {
    client: Client<H>,
}

impl<H> ModelLister<H> for DeepSeekModelLister<H>
where
    H: HttpClientExt + WasmCompatSend + WasmCompatSync + 'static,
{
    type Client = Client<H>;

    fn new(client: Self::Client) -> Self {
        Self { client }
    }

    async fn list_all(&self) -> Result<ModelList, ModelListingError> {
        let path = "/models";
        let req = self.client.get(path)?.body(http_client::NoBody)?;
        let response = self
            .client
            .send::<_, Vec<u8>>(req)
            .await
            .map_err(|error| match error {
                http_client::Error::InvalidStatusCodeWithMessage(status, message) => {
                    ModelListingError::api_error_with_context(
                        "DeepSeek",
                        path,
                        status.as_u16(),
                        message.as_bytes(),
                    )
                }
                other => ModelListingError::from(other),
            })?;

        if !response.status().is_success() {
            let status_code = response.status().as_u16();
            let body = response.into_body().await?;
            return Err(ModelListingError::api_error_with_context(
                "DeepSeek",
                path,
                status_code,
                &body,
            ));
        }

        let body = response.into_body().await?;
        let api_resp: ListModelsResponse = serde_json::from_slice(&body).map_err(|error| {
            ModelListingError::parse_error_with_context("DeepSeek", path, &error, &body)
        })?;

        let models = api_resp.data.into_iter().map(Model::from).collect();

        Ok(ModelList::new(models))
    }
}

// ================================================================
// DeepSeek Completion API
// ================================================================
#[deprecated(
    note = "The model names `deepseek-chat` and `deepseek-reasoner` will be deprecated on 2026/07/24. \
    For compatibility, they correspond to the non-thinking mode and thinking mode of `deepseek-v4-flash`, \
    respectively."
)]
pub const DEEPSEEK_CHAT: &str = "deepseek-chat";
#[deprecated(
    note = "The model names `deepseek-chat` and `deepseek-reasoner` will be deprecated on 2026/07/24. \
    For compatibility, they correspond to the non-thinking mode and thinking mode of `deepseek-v4-flash`, \
    respectively."
)]
pub const DEEPSEEK_REASONER: &str = "deepseek-reasoner";
pub const DEEPSEEK_V4_FLASH: &str = "deepseek-v4-flash";
pub const DEEPSEEK_V4_PRO: &str = "deepseek-v4-pro";

// Tests
#[cfg(test)]
mod tests {
    use super::*;
    use crate::client::ModelListingClient;
    use crate::test_utils::RecordingHttpClient;

    #[test]
    fn test_deserialize_vec_choice() {
        let data = r#"[{
            "finish_reason": "stop",
            "index": 0,
            "logprobs": null,
            "message":{"role":"assistant","content":"Hello, world!"}
            }]"#;

        let choices: Vec<Choice> = serde_json::from_str(data).unwrap();
        assert_eq!(choices.len(), 1);
        match &choices.first().unwrap().message {
            Message::Assistant { content, .. } => assert_eq!(content, "Hello, world!"),
            _ => panic!("Expected assistant message"),
        }
    }

    #[test]
    fn test_deserialize_deepseek_response() {
        let data = r#"{
            "choices":[{
                "finish_reason": "stop",
                "index": 0,
                "logprobs": null,
                "message":{"role":"assistant","content":"Hello, world!"}
            }],
            "usage": {
                "completion_tokens": 0,
                "prompt_tokens": 0,
                "prompt_cache_hit_tokens": 0,
                "prompt_cache_miss_tokens": 0,
                "total_tokens": 0
            }
        }"#;

        let jd = &mut serde_json::Deserializer::from_str(data);
        let result: Result<CompletionResponse, _> = serde_path_to_error::deserialize(jd);
        match result {
            Ok(response) => match &response.choices.first().unwrap().message {
                Message::Assistant { content, .. } => assert_eq!(content, "Hello, world!"),
                _ => panic!("Expected assistant message"),
            },
            Err(err) => {
                panic!("Deserialization error at {}: {}", err.path(), err);
            }
        }
    }

    #[test]
    fn test_deserialize_example_response() {
        let data = r#"
        {
            "id": "e45f6c68-9d9e-43de-beb4-4f402b850feb",
            "object": "chat.completion",
            "created": 0,
            "model": "deepseek-chat",
            "choices": [
                {
                    "index": 0,
                    "message": {
                        "role": "assistant",
                        "content": "Why don’t skeletons fight each other?  \nBecause they don’t have the guts! 😄"
                    },
                    "logprobs": null,
                    "finish_reason": "stop"
                }
            ],
            "usage": {
                "prompt_tokens": 13,
                "completion_tokens": 32,
                "total_tokens": 45,
                "prompt_tokens_details": {
                    "cached_tokens": 0
                },
                "prompt_cache_hit_tokens": 0,
                "prompt_cache_miss_tokens": 13
            },
            "system_fingerprint": "fp_4b6881f2c5"
        }
        "#;
        let jd = &mut serde_json::Deserializer::from_str(data);
        let result: Result<CompletionResponse, _> = serde_path_to_error::deserialize(jd);

        match result {
            Ok(response) => match &response.choices.first().unwrap().message {
                Message::Assistant { content, .. } => assert_eq!(
                    content,
                    "Why don’t skeletons fight each other?  \nBecause they don’t have the guts! 😄"
                ),
                _ => panic!("Expected assistant message"),
            },
            Err(err) => {
                panic!("Deserialization error at {}: {}", err.path(), err);
            }
        }
    }

    #[test]
    fn test_serialize_deserialize_tool_call_message() {
        let tool_call_choice_json = r#"
            {
              "finish_reason": "tool_calls",
              "index": 0,
              "logprobs": null,
              "message": {
                "content": "",
                "role": "assistant",
                "tool_calls": [
                  {
                    "function": {
                      "arguments": "{\"x\":2,\"y\":5}",
                      "name": "subtract"
                    },
                    "id": "call_0_2b4a85ee-b04a-40ad-a16b-a405caf6e65b",
                    "index": 0,
                    "type": "function"
                  }
                ]
              }
            }
        "#;

        let choice: Choice = serde_json::from_str(tool_call_choice_json).unwrap();

        let expected_choice: Choice = Choice {
            finish_reason: "tool_calls".to_string(),
            index: 0,
            logprobs: None,
            message: Message::Assistant {
                content: "".to_string(),
                name: None,
                tool_calls: vec![ToolCall {
                    id: "call_0_2b4a85ee-b04a-40ad-a16b-a405caf6e65b".to_string(),
                    function: Function {
                        name: "subtract".to_string(),
                        arguments: serde_json::from_str(r#"{"x":2,"y":5}"#).unwrap(),
                    },
                    index: 0,
                    r#type: ToolType::Function,
                }],
                reasoning_content: None,
            },
        };

        assert_eq!(choice, expected_choice);
    }
    #[test]
    fn test_user_message_multiple_text_items_merged() {
        use crate::completion::message::{Message as RigMessage, UserContent};

        let rig_msg = RigMessage::User {
            content: OneOrMany::many(vec![
                UserContent::text("first part"),
                UserContent::text("second part"),
            ])
            .expect("content should not be empty"),
        };

        let messages: Vec<Message> = rig_msg.try_into().expect("conversion should succeed");

        let user_messages: Vec<&Message> = messages
            .iter()
            .filter(|m| matches!(m, Message::User { .. }))
            .collect();

        assert_eq!(
            user_messages.len(),
            1,
            "multiple text items should produce a single user message"
        );
        match &user_messages[0] {
            Message::User { content, .. } => {
                assert_eq!(content, "first part\nsecond part");
            }
            _ => unreachable!(),
        }
    }

    #[test]
    fn test_assistant_message_with_reasoning_and_tool_calls() {
        use crate::completion::message::{AssistantContent, Message as RigMessage};

        let rig_msg = RigMessage::Assistant {
            id: None,
            content: OneOrMany::many(vec![
                AssistantContent::reasoning("thinking about the problem"),
                AssistantContent::text("I'll call the tool"),
                AssistantContent::tool_call(
                    "call_1",
                    "subtract",
                    serde_json::json!({"x": 2, "y": 5}),
                ),
            ])
            .expect("content should not be empty"),
        };

        let messages: Vec<Message> = rig_msg.try_into().expect("conversion should succeed");

        assert_eq!(messages.len(), 1, "should produce exactly one message");
        match &messages[0] {
            Message::Assistant {
                content,
                tool_calls,
                reasoning_content,
                ..
            } => {
                assert_eq!(content, "I'll call the tool");
                assert_eq!(
                    reasoning_content.as_deref(),
                    Some("thinking about the problem")
                );
                assert_eq!(tool_calls.len(), 1);
                assert_eq!(tool_calls[0].function.name, "subtract");
            }
            _ => panic!("Expected assistant message"),
        }
    }

    #[test]
    fn test_assistant_message_without_reasoning() {
        use crate::completion::message::{AssistantContent, Message as RigMessage};

        let rig_msg = RigMessage::Assistant {
            id: None,
            content: OneOrMany::many(vec![
                AssistantContent::text("calling tool"),
                AssistantContent::tool_call("call_1", "add", serde_json::json!({"a": 1, "b": 2})),
            ])
            .expect("content should not be empty"),
        };

        let messages: Vec<Message> = rig_msg.try_into().expect("conversion should succeed");

        assert_eq!(messages.len(), 1);
        match &messages[0] {
            Message::Assistant {
                reasoning_content,
                tool_calls,
                ..
            } => {
                assert!(reasoning_content.is_none());
                assert_eq!(tool_calls.len(), 1);
            }
            _ => panic!("Expected assistant message"),
        }
    }

    #[test]
    fn test_client_initialization() {
        let _client =
            crate::providers::deepseek::Client::new("dummy-key").expect("Client::new() failed");
        let _client_from_builder = crate::providers::deepseek::Client::builder()
            .api_key("dummy-key")
            .build()
            .expect("Client::builder() failed");
    }

    #[test]
    fn test_deserialize_list_models_response() {
        let data = r#"{
            "object": "list",
            "data": [
                {
                    "id": "deepseek-v4-flash",
                    "object": "model",
                    "owned_by": "deepseek"
                },
                {
                    "id": "deepseek-v4-pro",
                    "object": "model",
                    "owned_by": "deepseek"
                }
            ]
        }"#;

        let response: ListModelsResponse = serde_json::from_str(data).unwrap();

        assert_eq!(response.data.len(), 2);
        assert_eq!(response.data[0].id, "deepseek-v4-flash");
        assert_eq!(response.data[0].owned_by, "deepseek");
    }

    #[tokio::test]
    async fn test_list_models_uses_models_endpoint() {
        let response_body = r#"{
            "object": "list",
            "data": [
                {
                    "id": "deepseek-v4-flash",
                    "object": "model",
                    "owned_by": "deepseek"
                },
                {
                    "id": "deepseek-v4-pro",
                    "object": "model",
                    "owned_by": "deepseek"
                }
            ]
        }"#;

        let http_client = RecordingHttpClient::new(response_body);
        let client = Client::builder()
            .api_key("dummy-key")
            .http_client(http_client.clone())
            .build()
            .expect("client should build");

        let models = client
            .list_models()
            .await
            .expect("list_models should succeed");

        assert_eq!(models.len(), 2);
        assert_eq!(models.data[0].id, "deepseek-v4-flash");
        assert_eq!(models.data[0].r#type, None);
        assert_eq!(models.data[0].owned_by.as_deref(), Some("deepseek"));
        let requests = http_client.requests();
        assert_eq!(requests.len(), 1);
        assert_eq!(requests[0].uri, "https://api.deepseek.com/models");
    }

    #[tokio::test]
    async fn test_list_models_preserves_api_error_context() {
        let http_client = RecordingHttpClient::with_error(
            http::StatusCode::UNAUTHORIZED,
            r#"{"error":{"message":"invalid api key"}}"#,
        );
        let client = Client::builder()
            .api_key("dummy-key")
            .http_client(http_client)
            .build()
            .expect("client should build");

        let error = client
            .list_models()
            .await
            .expect_err("list_models should fail");

        match error {
            ModelListingError::ApiError {
                status_code,
                message,
            } => {
                assert_eq!(status_code, 401);
                assert!(message.contains("provider=DeepSeek"));
                assert!(message.contains("path=/models"));
                assert!(message.contains("invalid api key"));
            }
            other => panic!("expected api error, got {other:?}"),
        }
    }
}