construct/providers/

anthropic.rs

use crate::providers::traits::{
    ChatMessage, ChatRequest as ProviderChatRequest, ChatResponse as ProviderChatResponse,
    Provider, ProviderCapabilities, StreamChunk, StreamError, StreamEvent, StreamOptions,
    StreamResult, TokenUsage, ToolCall as ProviderToolCall,
};
use crate::tools::ToolSpec;
use async_trait::async_trait;
use base64::Engine as _;
use futures_util::stream::{self, StreamExt};
use reqwest::Client;
use serde::{Deserialize, Serialize};

pub struct AnthropicProvider {
    credential: Option<String>,
    base_url: String,
    max_tokens: u32,
}

const DEFAULT_ANTHROPIC_MAX_TOKENS: u32 = 4096;

#[derive(Debug, Serialize)]
struct ChatRequest {
    model: String,
    max_tokens: u32,
    #[serde(skip_serializing_if = "Option::is_none")]
    system: Option<String>,
    messages: Vec<Message>,
    temperature: f64,
}

#[derive(Debug, Serialize)]
struct Message {
    role: String,
    content: String,
}

#[derive(Debug, Deserialize)]
struct ChatResponse {
    content: Vec<ContentBlock>,
}

#[derive(Debug, Deserialize)]
struct ContentBlock {
    #[serde(rename = "type")]
    kind: String,
    #[serde(default)]
    text: Option<String>,
}

#[derive(Debug, Serialize)]
struct NativeChatRequest<'a> {
    model: String,
    max_tokens: u32,
    #[serde(skip_serializing_if = "Option::is_none")]
    system: Option<SystemPrompt>,
    messages: Vec<NativeMessage>,
    temperature: f64,
    #[serde(skip_serializing_if = "Option::is_none")]
    tools: Option<Vec<NativeToolSpec<'a>>>,
    #[serde(skip_serializing_if = "Option::is_none")]
    tool_choice: Option<serde_json::Value>,
    #[serde(skip_serializing_if = "Option::is_none")]
    stream: Option<bool>,
}

#[derive(Debug, Serialize)]
struct NativeMessage {
    role: String,
    content: Vec<NativeContentOut>,
}

#[derive(Debug, Serialize)]
struct ImageSource {
    #[serde(rename = "type")]
    source_type: String,
    media_type: String,
    data: String,
}

#[derive(Debug, Serialize)]
#[serde(tag = "type")]
enum NativeContentOut {
    #[serde(rename = "text")]
    Text {
        text: String,
        #[serde(skip_serializing_if = "Option::is_none")]
        cache_control: Option<CacheControl>,
    },
    #[serde(rename = "image")]
    Image { source: ImageSource },
    #[serde(rename = "tool_use")]
    ToolUse {
        id: String,
        name: String,
        input: serde_json::Value,
        #[serde(skip_serializing_if = "Option::is_none")]
        cache_control: Option<CacheControl>,
    },
    #[serde(rename = "tool_result")]
    ToolResult {
        tool_use_id: String,
        content: String,
        #[serde(skip_serializing_if = "Option::is_none")]
        cache_control: Option<CacheControl>,
    },
}

#[derive(Debug, Serialize)]
struct NativeToolSpec<'a> {
    name: &'a str,
    description: &'a str,
    input_schema: &'a serde_json::Value,
    #[serde(skip_serializing_if = "Option::is_none")]
    cache_control: Option<CacheControl>,
}

#[derive(Debug, Clone, Serialize)]
struct CacheControl {
    #[serde(rename = "type")]
    cache_type: String,
}

impl CacheControl {
    fn ephemeral() -> Self {
        Self {
            cache_type: "ephemeral".to_string(),
        }
    }
}

#[derive(Debug, Serialize)]
#[serde(untagged)]
enum SystemPrompt {
    String(String),
    Blocks(Vec<SystemBlock>),
}

#[derive(Debug, Serialize)]
struct SystemBlock {
    #[serde(rename = "type")]
    block_type: String,
    text: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    cache_control: Option<CacheControl>,
}

#[derive(Debug, Deserialize)]
struct NativeChatResponse {
    #[serde(default)]
    content: Vec<NativeContentIn>,
    #[serde(default)]
    usage: Option<AnthropicUsage>,
}

#[derive(Debug, Deserialize)]
struct AnthropicUsage {
    #[serde(default)]
    input_tokens: Option<u64>,
    #[serde(default)]
    output_tokens: Option<u64>,
    #[serde(default)]
    cache_creation_input_tokens: Option<u64>,
    #[serde(default)]
    cache_read_input_tokens: Option<u64>,
}

#[derive(Debug, Deserialize)]
struct NativeContentIn {
    #[serde(rename = "type")]
    kind: String,
    #[serde(default)]
    text: Option<String>,
    #[serde(default)]
    id: Option<String>,
    #[serde(default)]
    name: Option<String>,
    #[serde(default)]
    input: Option<serde_json::Value>,
}

impl AnthropicProvider {
    pub fn new(credential: Option<&str>) -> Self {
        Self::with_base_url(credential, None)
    }

    pub fn with_base_url(credential: Option<&str>, base_url: Option<&str>) -> Self {
        let base_url = base_url
            .map(|u| u.trim_end_matches('/'))
            .unwrap_or("https://api.anthropic.com")
            .to_string();
        Self {
            credential: credential
                .map(str::trim)
                .filter(|k| !k.is_empty())
                .map(ToString::to_string),
            base_url,
            max_tokens: DEFAULT_ANTHROPIC_MAX_TOKENS,
        }
    }

    /// Override the maximum output tokens for API requests.
    pub fn with_max_tokens(mut self, max_tokens: u32) -> Self {
        self.max_tokens = max_tokens;
        self
    }

    fn is_setup_token(token: &str) -> bool {
        token.starts_with("sk-ant-oat01-")
    }

    fn apply_auth(
        &self,
        request: reqwest::RequestBuilder,
        credential: &str,
    ) -> reqwest::RequestBuilder {
        if Self::is_setup_token(credential) {
            request
                .header("Authorization", format!("Bearer {credential}"))
                .header(
                    "anthropic-beta",
                    "claude-code-20250219,oauth-2025-04-20,interleaved-thinking-2025-05-14",
                )
                .header("anthropic-dangerous-direct-browser-access", "true")
        } else {
            request.header("x-api-key", credential)
        }
    }

    /// For OAuth tokens, Anthropic requires the system prompt to start with the
    /// Claude Code identity prefix. This prepends it to any existing system prompt.
    fn apply_oauth_system_prompt(system: Option<SystemPrompt>) -> Option<SystemPrompt> {
        let prefix = SystemBlock {
            block_type: "text".to_string(),
            text: "You are Claude Code, Anthropic's official CLI for Claude.".to_string(),
            cache_control: Some(CacheControl::ephemeral()),
        };
        match system {
            Some(SystemPrompt::Blocks(mut blocks)) => {
                blocks.insert(0, prefix);
                Some(SystemPrompt::Blocks(blocks))
            }
            Some(SystemPrompt::String(s)) => Some(SystemPrompt::Blocks(vec![
                prefix,
                SystemBlock {
                    block_type: "text".to_string(),
                    text: s,
                    cache_control: Some(CacheControl::ephemeral()),
                },
            ])),
            None => Some(SystemPrompt::Blocks(vec![prefix])),
        }
    }

    /// Cache system prompts larger than ~1024 tokens (3KB of text)
    fn should_cache_system(text: &str) -> bool {
        text.len() > 3072
    }

    /// Cache conversations with more than 1 non-system message (i.e. after first exchange)
    fn should_cache_conversation(messages: &[ChatMessage]) -> bool {
        messages.iter().filter(|m| m.role != "system").count() > 1
    }

    /// Apply cache control to the last message content block
    fn apply_cache_to_last_message(messages: &mut [NativeMessage]) {
        if let Some(last_msg) = messages.last_mut() {
            if let Some(last_content) = last_msg.content.last_mut() {
                match last_content {
                    NativeContentOut::Text { cache_control, .. }
                    | NativeContentOut::ToolResult { cache_control, .. } => {
                        *cache_control = Some(CacheControl::ephemeral());
                    }
                    NativeContentOut::ToolUse { .. } | NativeContentOut::Image { .. } => {}
                }
            }
        }
    }

    fn convert_tools<'a>(tools: Option<&'a [ToolSpec]>) -> Option<Vec<NativeToolSpec<'a>>> {
        let items = tools?;
        if items.is_empty() {
            return None;
        }
        let mut native_tools: Vec<NativeToolSpec<'a>> = items
            .iter()
            .map(|tool| NativeToolSpec {
                name: &tool.name,
                description: &tool.description,
                input_schema: &tool.parameters,
                cache_control: None,
            })
            .collect();

        // Cache the last tool definition (caches all tools)
        if let Some(last_tool) = native_tools.last_mut() {
            last_tool.cache_control = Some(CacheControl::ephemeral());
        }

        Some(native_tools)
    }

    fn parse_assistant_tool_call_message(content: &str) -> Option<Vec<NativeContentOut>> {
        let value = serde_json::from_str::<serde_json::Value>(content).ok()?;
        let tool_calls = value
            .get("tool_calls")
            .and_then(|v| serde_json::from_value::<Vec<ProviderToolCall>>(v.clone()).ok())?;

        let mut blocks = Vec::new();
        if let Some(text) = value
            .get("content")
            .and_then(serde_json::Value::as_str)
            .map(str::trim)
            .filter(|t| !t.is_empty())
        {
            blocks.push(NativeContentOut::Text {
                text: text.to_string(),
                cache_control: None,
            });
        }
        for call in tool_calls {
            let input = serde_json::from_str::<serde_json::Value>(&call.arguments)
                .unwrap_or_else(|_| serde_json::Value::Object(serde_json::Map::new()));
            blocks.push(NativeContentOut::ToolUse {
                id: call.id,
                name: call.name,
                input,
                cache_control: None,
            });
        }
        Some(blocks)
    }

    fn parse_tool_result_message(content: &str) -> Option<NativeMessage> {
        let value = serde_json::from_str::<serde_json::Value>(content).ok()?;
        let tool_use_id = value
            .get("tool_call_id")
            .and_then(serde_json::Value::as_str)?
            .to_string();
        let result = value
            .get("content")
            .and_then(serde_json::Value::as_str)
            .unwrap_or("")
            .to_string();
        Some(NativeMessage {
            role: "user".to_string(),
            content: vec![NativeContentOut::ToolResult {
                tool_use_id,
                content: result,
                cache_control: None,
            }],
        })
    }

    fn convert_messages(messages: &[ChatMessage]) -> (Option<SystemPrompt>, Vec<NativeMessage>) {
        let mut system_text = None;
        let mut native_messages = Vec::new();

        for msg in messages {
            match msg.role.as_str() {
                "system" => {
                    if system_text.is_none() {
                        system_text = Some(msg.content.clone());
                    }
                }
                "assistant" => {
                    if let Some(blocks) = Self::parse_assistant_tool_call_message(&msg.content) {
                        native_messages.push(NativeMessage {
                            role: "assistant".to_string(),
                            content: blocks,
                        });
                    } else if !msg.content.trim().is_empty() {
                        native_messages.push(NativeMessage {
                            role: "assistant".to_string(),
                            content: vec![NativeContentOut::Text {
                                text: msg.content.clone(),
                                cache_control: None,
                            }],
                        });
                    }
                }
                "tool" => {
                    let tool_msg = if let Some(tr) = Self::parse_tool_result_message(&msg.content) {
                        tr
                    } else if !msg.content.trim().is_empty() {
                        NativeMessage {
                            role: "user".to_string(),
                            content: vec![NativeContentOut::Text {
                                text: msg.content.clone(),
                                cache_control: None,
                            }],
                        }
                    } else {
                        continue;
                    };
                    // Tool results map to role "user"; merge consecutive ones
                    // into a single message so Anthropic doesn't reject the
                    // request for having adjacent same-role messages.
                    if native_messages
                        .last()
                        .is_some_and(|m| m.role == tool_msg.role)
                    {
                        native_messages
                            .last_mut()
                            .unwrap()
                            .content
                            .extend(tool_msg.content);
                    } else {
                        native_messages.push(tool_msg);
                    }
                }
                _ => {
                    // Parse image markers from user message content
                    let (text, image_refs) = crate::multimodal::parse_image_markers(&msg.content);
                    let mut content_blocks: Vec<NativeContentOut> = Vec::new();

                    // Add image content blocks for each image reference
                    for img_ref in &image_refs {
                        let (media_type, data) = if img_ref.starts_with("data:") {
                            // Data URI format: data:image/jpeg;base64,/9j/4AAQ...
                            if let Some(comma) = img_ref.find(',') {
                                let header = &img_ref[5..comma];
                                let mime =
                                    header.split(';').next().unwrap_or("image/jpeg").to_string();
                                let b64 = img_ref[comma + 1..].trim().to_string();
                                (mime, b64)
                            } else {
                                continue;
                            }
                        } else if std::path::Path::new(img_ref.trim()).exists() {
                            // Local file path
                            match std::fs::read(img_ref.trim()) {
                                Ok(bytes) => {
                                    let b64 =
                                        base64::engine::general_purpose::STANDARD.encode(&bytes);
                                    let ext = std::path::Path::new(img_ref.trim())
                                        .extension()
                                        .and_then(|e| e.to_str())
                                        .unwrap_or("jpg");
                                    let mime = match ext {
                                        "png" => "image/png",
                                        "gif" => "image/gif",
                                        "webp" => "image/webp",
                                        _ => "image/jpeg",
                                    }
                                    .to_string();
                                    (mime, b64)
                                }
                                Err(_) => continue,
                            }
                        } else {
                            continue;
                        };

                        content_blocks.push(NativeContentOut::Image {
                            source: ImageSource {
                                source_type: "base64".to_string(),
                                media_type,
                                data,
                            },
                        });
                    }

                    // Add text content block (skip empty text when images are present)
                    if text.is_empty() && !image_refs.is_empty() {
                        content_blocks.push(NativeContentOut::Text {
                            text: "[image]".to_string(),
                            cache_control: None,
                        });
                    } else if !text.trim().is_empty() {
                        content_blocks.push(NativeContentOut::Text {
                            text,
                            cache_control: None,
                        });
                    }

                    // Merge into previous user message if present (e.g.
                    // when a user message immediately follows tool results
                    // which are also role "user" in Anthropic's format).
                    if native_messages.last().is_some_and(|m| m.role == "user") {
                        native_messages
                            .last_mut()
                            .unwrap()
                            .content
                            .extend(content_blocks);
                    } else {
                        native_messages.push(NativeMessage {
                            role: "user".to_string(),
                            content: content_blocks,
                        });
                    }
                }
            }
        }

        // Always use Blocks format with cache_control for system prompts
        let system_prompt = system_text.map(|text| {
            SystemPrompt::Blocks(vec![SystemBlock {
                block_type: "text".to_string(),
                text,
                cache_control: Some(CacheControl::ephemeral()),
            }])
        });

        (system_prompt, native_messages)
    }

    fn parse_text_response(response: ChatResponse) -> anyhow::Result<String> {
        response
            .content
            .into_iter()
            .find(|c| c.kind == "text")
            .and_then(|c| c.text)
            .ok_or_else(|| anyhow::anyhow!("No response from Anthropic"))
    }

    fn parse_native_response(response: NativeChatResponse) -> ProviderChatResponse {
        let mut text_parts = Vec::new();
        let mut tool_calls = Vec::new();

        let usage = response.usage.map(|u| TokenUsage {
            input_tokens: u.input_tokens,
            output_tokens: u.output_tokens,
            cached_input_tokens: u.cache_read_input_tokens,
        });

        for block in response.content {
            match block.kind.as_str() {
                "text" => {
                    if let Some(text) = block.text.map(|t| t.trim().to_string()) {
                        if !text.is_empty() {
                            text_parts.push(text);
                        }
                    }
                }
                "tool_use" => {
                    let name = block.name.unwrap_or_default();
                    if name.is_empty() {
                        continue;
                    }
                    let arguments = block
                        .input
                        .unwrap_or_else(|| serde_json::Value::Object(serde_json::Map::new()));
                    tool_calls.push(ProviderToolCall {
                        id: block.id.unwrap_or_else(|| uuid::Uuid::new_v4().to_string()),
                        name,
                        arguments: arguments.to_string(),
                    });
                }
                _ => {}
            }
        }

        ProviderChatResponse {
            text: if text_parts.is_empty() {
                None
            } else {
                Some(text_parts.join("\n"))
            },
            tool_calls,
            usage,
            reasoning_content: None,
        }
    }

    fn http_client(&self) -> Client {
        crate::config::build_runtime_proxy_client_with_timeouts("provider.anthropic", 120, 10)
    }

    /// Build a streaming request body from a `NativeChatRequest`.
    fn build_streaming_request(request: &NativeChatRequest<'_>) -> serde_json::Value {
        let mut body =
            serde_json::to_value(request).expect("NativeChatRequest should serialize to JSON");
        body["stream"] = serde_json::Value::Bool(true);
        body
    }

    /// Parse Anthropic SSE lines from `response` and send `StreamEvent`s to `tx`.
    async fn parse_anthropic_sse(
        response: reqwest::Response,
        tx: &tokio::sync::mpsc::Sender<StreamResult<StreamEvent>>,
    ) {
        use tokio::io::AsyncBufReadExt;
        use tokio_util::io::StreamReader;

        let byte_stream = response
            .bytes_stream()
            .map(|result| result.map_err(std::io::Error::other));
        let reader = StreamReader::new(byte_stream);
        let mut lines = reader.lines();

        let mut tool_id: Option<String> = None;
        let mut tool_name: Option<String> = None;
        let mut tool_input_json = String::new();

        while let Ok(Some(line)) = lines.next_line().await {
            let line = line.trim().to_string();
            if !line.starts_with("data: ") {
                continue;
            }
            let json_str = &line["data: ".len()..];

            let event: serde_json::Value = match serde_json::from_str(json_str) {
                Ok(v) => v,
                Err(_) => continue,
            };

            let event_type = event
                .get("type")
                .and_then(|t| t.as_str())
                .unwrap_or_default();

            match event_type {
                "message_start" => {
                    let model = event
                        .get("message")
                        .and_then(|m| m.get("model"))
                        .and_then(|m| m.as_str())
                        .unwrap_or("unknown");
                    let usage_obj = event.get("message").and_then(|m| m.get("usage"));
                    let input_tokens = usage_obj
                        .and_then(|u| u.get("input_tokens"))
                        .and_then(|t| t.as_u64())
                        .unwrap_or(0);
                    let cache_read = usage_obj
                        .and_then(|u| u.get("cache_read_input_tokens"))
                        .and_then(|t| t.as_u64());
                    tracing::debug!(
                        model = %model,
                        input_tokens = input_tokens,
                        "Anthropic stream: message_start"
                    );
                    let _ = tx
                        .send(Ok(StreamEvent::Usage(
                            crate::providers::traits::TokenUsage {
                                input_tokens: Some(input_tokens),
                                output_tokens: None,
                                cached_input_tokens: cache_read,
                            },
                        )))
                        .await;
                }
                "content_block_start" => {
                    if let Some(block) = event.get("content_block") {
                        let block_type = block
                            .get("type")
                            .and_then(|t| t.as_str())
                            .unwrap_or_default();
                        if block_type == "tool_use" {
                            if let Some(id) = tool_id.take() {
                                let name = tool_name.take().unwrap_or_default();
                                let input = std::mem::take(&mut tool_input_json);
                                let _ = tx
                                    .send(Ok(StreamEvent::ToolCall(ProviderToolCall {
                                        id,
                                        name,
                                        arguments: input,
                                    })))
                                    .await;
                            }
                            tool_id = block
                                .get("id")
                                .and_then(|v| v.as_str())
                                .map(ToString::to_string);
                            tool_name = block
                                .get("name")
                                .and_then(|v| v.as_str())
                                .map(ToString::to_string);
                            tool_input_json.clear();
                        }
                    }
                }
                "content_block_delta" => {
                    if let Some(delta) = event.get("delta") {
                        let delta_type = delta
                            .get("type")
                            .and_then(|t| t.as_str())
                            .unwrap_or_default();
                        match delta_type {
                            "text_delta" => {
                                if let Some(text) = delta.get("text").and_then(|t| t.as_str()) {
                                    if !text.is_empty()
                                        && tx
                                            .send(Ok(StreamEvent::TextDelta(StreamChunk::delta(
                                                text.to_string(),
                                            ))))
                                            .await
                                            .is_err()
                                    {
                                        return;
                                    }
                                }
                            }
                            "input_json_delta" => {
                                if let Some(json) =
                                    delta.get("partial_json").and_then(|j| j.as_str())
                                {
                                    tool_input_json.push_str(json);
                                }
                            }
                            _ => {}
                        }
                    }
                }
                "content_block_stop" => {
                    if let Some(id) = tool_id.take() {
                        let name = tool_name.take().unwrap_or_default();
                        let input = std::mem::take(&mut tool_input_json);
                        let _ = tx
                            .send(Ok(StreamEvent::ToolCall(ProviderToolCall {
                                id,
                                name,
                                arguments: input,
                            })))
                            .await;
                    }
                }
                "message_delta" => {
                    let stop_reason = event
                        .get("delta")
                        .and_then(|d| d.get("stop_reason"))
                        .and_then(|s| s.as_str())
                        .unwrap_or("none");
                    let output_tokens = event
                        .get("usage")
                        .and_then(|u| u.get("output_tokens"))
                        .and_then(|t| t.as_u64())
                        .unwrap_or(0);
                    if stop_reason == "max_tokens" {
                        tracing::warn!(
                            output_tokens = output_tokens,
                            "Anthropic response truncated: hit max_tokens limit. Increase provider_max_tokens in config."
                        );
                    } else {
                        tracing::debug!(
                            stop_reason = %stop_reason,
                            output_tokens = output_tokens,
                            "Anthropic stream: message_delta"
                        );
                    }
                    if output_tokens > 0 {
                        let _ = tx
                            .send(Ok(StreamEvent::Usage(
                                crate::providers::traits::TokenUsage {
                                    input_tokens: None,
                                    output_tokens: Some(output_tokens),
                                    cached_input_tokens: None,
                                },
                            )))
                            .await;
                    }
                }
                "message_stop" => {
                    tracing::debug!("Anthropic stream: message_stop");
                    let _ = tx.send(Ok(StreamEvent::Final)).await;
                    return;
                }
                "error" => {
                    let msg = event
                        .get("error")
                        .and_then(|e| e.get("message"))
                        .and_then(|m| m.as_str())
                        .unwrap_or("unknown streaming error");
                    let _ = tx.send(Err(StreamError::Provider(msg.to_string()))).await;
                    return;
                }
                _ => {}
            }
        }

        let _ = tx.send(Ok(StreamEvent::Final)).await;
    }
}

#[async_trait]
impl Provider for AnthropicProvider {
    async fn chat_with_system(
        &self,
        system_prompt: Option<&str>,
        message: &str,
        model: &str,
        temperature: f64,
    ) -> anyhow::Result<String> {
        let credential = self.credential.as_ref().ok_or_else(|| {
            anyhow::anyhow!(
                "Anthropic credentials not set. Set ANTHROPIC_API_KEY or ANTHROPIC_OAUTH_TOKEN (setup-token)."
            )
        })?;

        let system = system_prompt.map(|s| SystemPrompt::String(s.to_string()));
        let system = if Self::is_setup_token(credential) {
            Self::apply_oauth_system_prompt(system)
        } else {
            system
        };

        tracing::debug!(max_tokens = self.max_tokens, model = %model, "Anthropic API request");
        let request = NativeChatRequest {
            model: model.to_string(),
            max_tokens: self.max_tokens,
            system,
            messages: vec![NativeMessage {
                role: "user".to_string(),
                content: vec![NativeContentOut::Text {
                    text: message.to_string(),
                    cache_control: None,
                }],
            }],
            temperature,
            tools: None,
            tool_choice: None,
            stream: None,
        };

        let mut request = self
            .http_client()
            .post(format!("{}/v1/messages", self.base_url))
            .header("anthropic-version", "2023-06-01")
            .header("content-type", "application/json")
            .json(&request);

        request = self.apply_auth(request, credential);

        let response = request.send().await?;

        if !response.status().is_success() {
            return Err(super::api_error("Anthropic", response).await);
        }

        let chat_response: NativeChatResponse = response.json().await?;
        let parsed = Self::parse_native_response(chat_response);
        parsed
            .text
            .ok_or_else(|| anyhow::anyhow!("No response from Anthropic"))
    }

    async fn chat(
        &self,
        request: ProviderChatRequest<'_>,
        model: &str,
        temperature: f64,
    ) -> anyhow::Result<ProviderChatResponse> {
        let credential = self.credential.as_ref().ok_or_else(|| {
            anyhow::anyhow!(
                "Anthropic credentials not set. Set ANTHROPIC_API_KEY or ANTHROPIC_OAUTH_TOKEN (setup-token)."
            )
        })?;

        let (system_prompt, mut messages) = Self::convert_messages(request.messages);

        // Auto-cache last message if conversation is long
        if Self::should_cache_conversation(request.messages) {
            Self::apply_cache_to_last_message(&mut messages);
        }

        // Check for tool_choice override from the agent loop (e.g. "any"
        // to force tool use for hardware requests).
        let tool_choice_override = crate::agent::loop_::TOOL_CHOICE_OVERRIDE
            .try_with(Clone::clone)
            .ok()
            .flatten();
        let native_tools = Self::convert_tools(request.tools);
        let tool_choice = if native_tools.is_some() {
            tool_choice_override.map(|tc| serde_json::json!({ "type": tc }))
        } else {
            None
        };

        // For OAuth tokens, prepend Claude Code identity to system prompt
        let system_prompt = if Self::is_setup_token(credential) {
            Self::apply_oauth_system_prompt(system_prompt)
        } else {
            system_prompt
        };
        tracing::debug!(max_tokens = self.max_tokens, model = %model, "Anthropic streaming API request");
        let native_request = NativeChatRequest {
            model: model.to_string(),
            max_tokens: self.max_tokens,
            system: system_prompt,
            messages,
            temperature,
            tools: native_tools,
            tool_choice,
            stream: None,
        };

        let req = self
            .http_client()
            .post(format!("{}/v1/messages", self.base_url))
            .header("anthropic-version", "2023-06-01")
            .header("content-type", "application/json")
            .json(&native_request);

        let response = self.apply_auth(req, credential).send().await?;
        if !response.status().is_success() {
            return Err(super::api_error("Anthropic", response).await);
        }

        let native_response: NativeChatResponse = response.json().await?;
        Ok(Self::parse_native_response(native_response))
    }

    fn capabilities(&self) -> ProviderCapabilities {
        ProviderCapabilities {
            native_tool_calling: true,
            vision: true,
            prompt_caching: true,
        }
    }

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

    async fn chat_with_tools(
        &self,
        messages: &[ChatMessage],
        tools: &[serde_json::Value],
        model: &str,
        temperature: f64,
    ) -> anyhow::Result<ProviderChatResponse> {
        // Convert OpenAI-format tool JSON to ToolSpec so we can reuse the
        // existing `chat()` method which handles full message history,
        // system prompt extraction, caching, and Anthropic native formatting.
        let tool_specs: Vec<ToolSpec> = tools
            .iter()
            .filter_map(|t| {
                let func = t.get("function").or_else(|| {
                    tracing::warn!("Skipping malformed tool definition (missing 'function' key)");
                    None
                })?;
                let name = func.get("name").and_then(|n| n.as_str()).or_else(|| {
                    tracing::warn!("Skipping tool with missing or non-string 'name'");
                    None
                })?;
                Some(ToolSpec {
                    name: name.to_string(),
                    description: func
                        .get("description")
                        .and_then(|d| d.as_str())
                        .unwrap_or("")
                        .to_string(),
                    parameters: func
                        .get("parameters")
                        .cloned()
                        .unwrap_or(serde_json::json!({"type": "object"})),
                })
            })
            .collect();

        let request = ProviderChatRequest {
            messages,
            tools: if tool_specs.is_empty() {
                None
            } else {
                Some(&tool_specs)
            },
        };
        self.chat(request, model, temperature).await
    }

    async fn warmup(&self) -> anyhow::Result<()> {
        if let Some(credential) = self.credential.as_ref() {
            let mut request = self
                .http_client()
                .post(format!("{}/v1/messages", self.base_url))
                .header("anthropic-version", "2023-06-01");
            request = self.apply_auth(request, credential);
            // Send a minimal request; the goal is TLS + HTTP/2 setup, not a valid response.
            // Anthropic has no lightweight GET endpoint, so we accept any non-network error.
            let _ = request.send().await?;
        }
        Ok(())
    }

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

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

    fn stream_chat(
        &self,
        request: ProviderChatRequest<'_>,
        model: &str,
        temperature: f64,
        options: StreamOptions,
    ) -> stream::BoxStream<'static, StreamResult<StreamEvent>> {
        if !options.enabled {
            return stream::once(async { Ok(StreamEvent::Final) }).boxed();
        }

        let credential = match self.credential.as_ref() {
            Some(c) => c.clone(),
            None => {
                return stream::once(async {
                    Err(StreamError::Provider(
                        "Anthropic credentials not set".to_string(),
                    ))
                })
                .boxed();
            }
        };

        let (system_prompt, mut messages) = Self::convert_messages(request.messages);
        if Self::should_cache_conversation(request.messages) {
            Self::apply_cache_to_last_message(&mut messages);
        }

        let tool_choice_override = crate::agent::loop_::TOOL_CHOICE_OVERRIDE
            .try_with(Clone::clone)
            .ok()
            .flatten();
        let native_tools = Self::convert_tools(request.tools);
        let tool_choice = if native_tools.is_some() {
            tool_choice_override.map(|tc| serde_json::json!({ "type": tc }))
        } else {
            None
        };

        let system_prompt = if Self::is_setup_token(&credential) {
            Self::apply_oauth_system_prompt(system_prompt)
        } else {
            system_prompt
        };

        tracing::debug!(max_tokens = self.max_tokens, model = %model, "Anthropic stream_chat request");
        let native_request = NativeChatRequest {
            model: model.to_string(),
            max_tokens: self.max_tokens,
            system: system_prompt,
            messages,
            temperature,
            tools: native_tools,
            tool_choice,
            stream: Some(true),
        };

        let body = Self::build_streaming_request(&native_request);
        let client = self.http_client();
        let url = format!("{}/v1/messages", self.base_url);
        let is_oauth = Self::is_setup_token(&credential);

        let (tx, rx) = tokio::sync::mpsc::channel::<StreamResult<StreamEvent>>(64);

        tokio::spawn(async move {
            let mut req = client
                .post(&url)
                .header("anthropic-version", "2023-06-01")
                .header("content-type", "application/json")
                .json(&body);

            if is_oauth {
                req = req
                    .header("Authorization", format!("Bearer {credential}"))
                    .header(
                        "anthropic-beta",
                        "claude-code-20250219,oauth-2025-04-20,interleaved-thinking-2025-05-14",
                    )
                    .header("anthropic-dangerous-direct-browser-access", "true");
            } else {
                req = req.header("x-api-key", &credential);
            }

            let response = match req.send().await {
                Ok(r) => r,
                Err(e) => {
                    let _ = tx.send(Err(StreamError::Http(e))).await;
                    return;
                }
            };

            if !response.status().is_success() {
                let status = response.status();
                let error = response
                    .text()
                    .await
                    .unwrap_or_else(|_| format!("HTTP error: {status}"));
                let _ = tx
                    .send(Err(StreamError::Provider(format!("{status}: {error}"))))
                    .await;
                return;
            }

            Self::parse_anthropic_sse(response, &tx).await;
        });

        stream::unfold(rx, |mut rx| async move {
            rx.recv().await.map(|event| (event, rx))
        })
        .boxed()
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::auth::anthropic_token::{AnthropicAuthKind, detect_auth_kind};

    #[test]
    fn creates_with_key() {
        let p = AnthropicProvider::new(Some("anthropic-test-credential"));
        assert!(p.credential.is_some());
        assert_eq!(p.credential.as_deref(), Some("anthropic-test-credential"));
        assert_eq!(p.base_url, "https://api.anthropic.com");
    }

    #[test]
    fn creates_without_key() {
        let p = AnthropicProvider::new(None);
        assert!(p.credential.is_none());
        assert_eq!(p.base_url, "https://api.anthropic.com");
    }

    #[test]
    fn creates_with_empty_key() {
        let p = AnthropicProvider::new(Some(""));
        assert!(p.credential.is_none());
    }

    #[test]
    fn creates_with_whitespace_key() {
        let p = AnthropicProvider::new(Some("  anthropic-test-credential  "));
        assert!(p.credential.is_some());
        assert_eq!(p.credential.as_deref(), Some("anthropic-test-credential"));
    }

    #[test]
    fn creates_with_custom_base_url() {
        let p = AnthropicProvider::with_base_url(
            Some("anthropic-credential"),
            Some("https://api.example.com"),
        );
        assert_eq!(p.base_url, "https://api.example.com");
        assert_eq!(p.credential.as_deref(), Some("anthropic-credential"));
    }

    #[test]
    fn custom_base_url_trims_trailing_slash() {
        let p = AnthropicProvider::with_base_url(None, Some("https://api.example.com/"));
        assert_eq!(p.base_url, "https://api.example.com");
    }

    #[test]
    fn default_base_url_when_none_provided() {
        let p = AnthropicProvider::with_base_url(None, None);
        assert_eq!(p.base_url, "https://api.anthropic.com");
    }

    #[tokio::test]
    async fn chat_fails_without_key() {
        let p = AnthropicProvider::new(None);
        let result = p
            .chat_with_system(None, "hello", "claude-3-opus", 0.7)
            .await;
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(
            err.contains("credentials not set"),
            "Expected key error, got: {err}"
        );
    }

    #[test]
    fn setup_token_detection_works() {
        assert!(AnthropicProvider::is_setup_token("sk-ant-oat01-abcdef"));
        assert!(!AnthropicProvider::is_setup_token("sk-ant-api-key"));
    }

    #[test]
    fn apply_auth_uses_bearer_and_beta_for_setup_tokens() {
        let provider = AnthropicProvider::new(None);
        let request = provider
            .apply_auth(
                provider
                    .http_client()
                    .get("https://api.anthropic.com/v1/models"),
                "sk-ant-oat01-test-token",
            )
            .build()
            .expect("request should build");

        assert_eq!(
            request
                .headers()
                .get("authorization")
                .and_then(|v| v.to_str().ok()),
            Some("Bearer sk-ant-oat01-test-token")
        );
        assert_eq!(
            request
                .headers()
                .get("anthropic-beta")
                .and_then(|v| v.to_str().ok()),
            Some("claude-code-20250219,oauth-2025-04-20,interleaved-thinking-2025-05-14")
        );
        assert_eq!(
            request
                .headers()
                .get("anthropic-dangerous-direct-browser-access")
                .and_then(|v| v.to_str().ok()),
            Some("true")
        );
        assert!(request.headers().get("x-api-key").is_none());
    }

    #[test]
    fn apply_auth_uses_x_api_key_for_regular_tokens() {
        let provider = AnthropicProvider::new(None);
        let request = provider
            .apply_auth(
                provider
                    .http_client()
                    .get("https://api.anthropic.com/v1/models"),
                "sk-ant-api-key",
            )
            .build()
            .expect("request should build");

        assert_eq!(
            request
                .headers()
                .get("x-api-key")
                .and_then(|v| v.to_str().ok()),
            Some("sk-ant-api-key")
        );
        assert!(request.headers().get("authorization").is_none());
        assert!(request.headers().get("anthropic-beta").is_none());
    }

    #[tokio::test]
    async fn chat_with_system_fails_without_key() {
        let p = AnthropicProvider::new(None);
        let result = p
            .chat_with_system(Some("You are Construct"), "hello", "claude-3-opus", 0.7)
            .await;
        assert!(result.is_err());
    }

    #[test]
    fn chat_request_serializes_without_system() {
        let req = ChatRequest {
            model: "claude-3-opus".to_string(),
            max_tokens: 4096,
            system: None,
            messages: vec![Message {
                role: "user".to_string(),
                content: "hello".to_string(),
            }],
            temperature: 0.7,
        };
        let json = serde_json::to_string(&req).unwrap();
        assert!(
            !json.contains("system"),
            "system field should be skipped when None"
        );
        assert!(json.contains("claude-3-opus"));
        assert!(json.contains("hello"));
    }

    #[test]
    fn chat_request_serializes_with_system() {
        let req = ChatRequest {
            model: "claude-3-opus".to_string(),
            max_tokens: 4096,
            system: Some("You are Construct".to_string()),
            messages: vec![Message {
                role: "user".to_string(),
                content: "hello".to_string(),
            }],
            temperature: 0.7,
        };
        let json = serde_json::to_string(&req).unwrap();
        assert!(json.contains("\"system\":\"You are Construct\""));
    }

    #[test]
    fn chat_response_deserializes() {
        let json = r#"{"content":[{"type":"text","text":"Hello there!"}]}"#;
        let resp: ChatResponse = serde_json::from_str(json).unwrap();
        assert_eq!(resp.content.len(), 1);
        assert_eq!(resp.content[0].kind, "text");
        assert_eq!(resp.content[0].text.as_deref(), Some("Hello there!"));
    }

    #[test]
    fn chat_response_empty_content() {
        let json = r#"{"content":[]}"#;
        let resp: ChatResponse = serde_json::from_str(json).unwrap();
        assert!(resp.content.is_empty());
    }

    #[test]
    fn chat_response_multiple_blocks() {
        let json =
            r#"{"content":[{"type":"text","text":"First"},{"type":"text","text":"Second"}]}"#;
        let resp: ChatResponse = serde_json::from_str(json).unwrap();
        assert_eq!(resp.content.len(), 2);
        assert_eq!(resp.content[0].text.as_deref(), Some("First"));
        assert_eq!(resp.content[1].text.as_deref(), Some("Second"));
    }

    #[test]
    fn temperature_range_serializes() {
        for temp in [0.0, 0.5, 1.0, 2.0] {
            let req = ChatRequest {
                model: "claude-3-opus".to_string(),
                max_tokens: 4096,
                system: None,
                messages: vec![],
                temperature: temp,
            };
            let json = serde_json::to_string(&req).unwrap();
            assert!(json.contains(&format!("{temp}")));
        }
    }

    #[test]
    fn detects_auth_from_jwt_shape() {
        let kind = detect_auth_kind("a.b.c", None);
        assert_eq!(kind, AnthropicAuthKind::Authorization);
    }

    #[test]
    fn cache_control_serializes_correctly() {
        let cache = CacheControl::ephemeral();
        let json = serde_json::to_string(&cache).unwrap();
        assert_eq!(json, r#"{"type":"ephemeral"}"#);
    }

    #[test]
    fn system_prompt_string_variant_serializes() {
        let prompt = SystemPrompt::String("You are a helpful assistant".to_string());
        let json = serde_json::to_string(&prompt).unwrap();
        assert_eq!(json, r#""You are a helpful assistant""#);
    }

    #[test]
    fn system_prompt_blocks_variant_serializes() {
        let prompt = SystemPrompt::Blocks(vec![SystemBlock {
            block_type: "text".to_string(),
            text: "You are a helpful assistant".to_string(),
            cache_control: Some(CacheControl::ephemeral()),
        }]);
        let json = serde_json::to_string(&prompt).unwrap();
        assert!(json.contains(r#""type":"text""#));
        assert!(json.contains("You are a helpful assistant"));
        assert!(json.contains(r#""type":"ephemeral""#));
    }

    #[test]
    fn system_prompt_blocks_without_cache_control() {
        let prompt = SystemPrompt::Blocks(vec![SystemBlock {
            block_type: "text".to_string(),
            text: "Short prompt".to_string(),
            cache_control: None,
        }]);
        let json = serde_json::to_string(&prompt).unwrap();
        assert!(json.contains("Short prompt"));
        assert!(!json.contains("cache_control"));
    }

    #[test]
    fn native_content_text_without_cache_control() {
        let content = NativeContentOut::Text {
            text: "Hello".to_string(),
            cache_control: None,
        };
        let json = serde_json::to_string(&content).unwrap();
        assert!(json.contains(r#""type":"text""#));
        assert!(json.contains("Hello"));
        assert!(!json.contains("cache_control"));
    }

    #[test]
    fn native_content_text_with_cache_control() {
        let content = NativeContentOut::Text {
            text: "Hello".to_string(),
            cache_control: Some(CacheControl::ephemeral()),
        };
        let json = serde_json::to_string(&content).unwrap();
        assert!(json.contains(r#""type":"text""#));
        assert!(json.contains("Hello"));
        assert!(json.contains(r#""cache_control":{"type":"ephemeral"}"#));
    }

    #[test]
    fn native_content_tool_use_without_cache_control() {
        let content = NativeContentOut::ToolUse {
            id: "tool_123".to_string(),
            name: "get_weather".to_string(),
            input: serde_json::json!({"location": "San Francisco"}),
            cache_control: None,
        };
        let json = serde_json::to_string(&content).unwrap();
        assert!(json.contains(r#""type":"tool_use""#));
        assert!(json.contains("tool_123"));
        assert!(json.contains("get_weather"));
        assert!(!json.contains("cache_control"));
    }

    #[test]
    fn native_content_tool_result_with_cache_control() {
        let content = NativeContentOut::ToolResult {
            tool_use_id: "tool_123".to_string(),
            content: "Result data".to_string(),
            cache_control: Some(CacheControl::ephemeral()),
        };
        let json = serde_json::to_string(&content).unwrap();
        assert!(json.contains(r#""type":"tool_result""#));
        assert!(json.contains("tool_123"));
        assert!(json.contains("Result data"));
        assert!(json.contains(r#""cache_control":{"type":"ephemeral"}"#));
    }

    #[test]
    fn native_tool_spec_without_cache_control() {
        let schema = serde_json::json!({"type": "object"});
        let tool = NativeToolSpec {
            name: "get_weather",
            description: "Get weather info",
            input_schema: &schema,
            cache_control: None,
        };
        let json = serde_json::to_string(&tool).unwrap();
        assert!(json.contains("get_weather"));
        assert!(!json.contains("cache_control"));
    }

    #[test]
    fn native_tool_spec_with_cache_control() {
        let schema = serde_json::json!({"type": "object"});
        let tool = NativeToolSpec {
            name: "get_weather",
            description: "Get weather info",
            input_schema: &schema,
            cache_control: Some(CacheControl::ephemeral()),
        };
        let json = serde_json::to_string(&tool).unwrap();
        assert!(json.contains("get_weather"));
        assert!(json.contains(r#""cache_control":{"type":"ephemeral"}"#));
    }

    #[test]
    fn should_cache_system_small_prompt() {
        let small_prompt = "You are a helpful assistant.";
        assert!(!AnthropicProvider::should_cache_system(small_prompt));
    }

    #[test]
    fn should_cache_system_large_prompt() {
        let large_prompt = "a".repeat(3073); // Just over 3072 bytes
        assert!(AnthropicProvider::should_cache_system(&large_prompt));
    }

    #[test]
    fn should_cache_system_boundary() {
        let boundary_prompt = "a".repeat(3072); // Exactly 3072 bytes
        assert!(!AnthropicProvider::should_cache_system(&boundary_prompt));

        let over_boundary = "a".repeat(3073);
        assert!(AnthropicProvider::should_cache_system(&over_boundary));
    }

    #[test]
    fn should_cache_conversation_short() {
        let messages = vec![
            ChatMessage {
                role: "system".to_string(),
                content: "System prompt".to_string(),
            },
            ChatMessage {
                role: "user".to_string(),
                content: "Hello".to_string(),
            },
        ];
        // Only 1 non-system message — should not cache
        assert!(!AnthropicProvider::should_cache_conversation(&messages));
    }

    #[test]
    fn should_cache_conversation_long() {
        let mut messages = vec![ChatMessage {
            role: "system".to_string(),
            content: "System prompt".to_string(),
        }];
        // Add 3 non-system messages
        for i in 0..3 {
            messages.push(ChatMessage {
                role: if i % 2 == 0 { "user" } else { "assistant" }.to_string(),
                content: format!("Message {i}"),
            });
        }
        assert!(AnthropicProvider::should_cache_conversation(&messages));
    }

    #[test]
    fn should_cache_conversation_boundary() {
        let messages = vec![ChatMessage {
            role: "user".to_string(),
            content: "Hello".to_string(),
        }];
        // Exactly 1 non-system message — should not cache
        assert!(!AnthropicProvider::should_cache_conversation(&messages));

        // Add one more to cross boundary (>1)
        let messages = vec![
            ChatMessage {
                role: "user".to_string(),
                content: "Hello".to_string(),
            },
            ChatMessage {
                role: "assistant".to_string(),
                content: "Hi".to_string(),
            },
        ];
        assert!(AnthropicProvider::should_cache_conversation(&messages));
    }

    #[test]
    fn apply_cache_to_last_message_text() {
        let mut messages = vec![NativeMessage {
            role: "user".to_string(),
            content: vec![NativeContentOut::Text {
                text: "Hello".to_string(),
                cache_control: None,
            }],
        }];

        AnthropicProvider::apply_cache_to_last_message(&mut messages);

        match &messages[0].content[0] {
            NativeContentOut::Text { cache_control, .. } => {
                assert!(cache_control.is_some());
            }
            _ => panic!("Expected Text variant"),
        }
    }

    #[test]
    fn apply_cache_to_last_message_tool_result() {
        let mut messages = vec![NativeMessage {
            role: "user".to_string(),
            content: vec![NativeContentOut::ToolResult {
                tool_use_id: "tool_123".to_string(),
                content: "Result".to_string(),
                cache_control: None,
            }],
        }];

        AnthropicProvider::apply_cache_to_last_message(&mut messages);

        match &messages[0].content[0] {
            NativeContentOut::ToolResult { cache_control, .. } => {
                assert!(cache_control.is_some());
            }
            _ => panic!("Expected ToolResult variant"),
        }
    }

    #[test]
    fn apply_cache_to_last_message_does_not_affect_tool_use() {
        let mut messages = vec![NativeMessage {
            role: "assistant".to_string(),
            content: vec![NativeContentOut::ToolUse {
                id: "tool_123".to_string(),
                name: "get_weather".to_string(),
                input: serde_json::json!({}),
                cache_control: None,
            }],
        }];

        AnthropicProvider::apply_cache_to_last_message(&mut messages);

        // ToolUse should not be affected
        match &messages[0].content[0] {
            NativeContentOut::ToolUse { cache_control, .. } => {
                assert!(cache_control.is_none());
            }
            _ => panic!("Expected ToolUse variant"),
        }
    }

    #[test]
    fn apply_cache_empty_messages() {
        let mut messages = vec![];
        AnthropicProvider::apply_cache_to_last_message(&mut messages);
        // Should not panic
        assert!(messages.is_empty());
    }

    #[test]
    fn convert_tools_adds_cache_to_last_tool() {
        let tools = vec![
            ToolSpec {
                name: "tool1".to_string(),
                description: "First tool".to_string(),
                parameters: serde_json::json!({"type": "object"}),
            },
            ToolSpec {
                name: "tool2".to_string(),
                description: "Second tool".to_string(),
                parameters: serde_json::json!({"type": "object"}),
            },
        ];

        let native_tools = AnthropicProvider::convert_tools(Some(&tools)).unwrap();

        assert_eq!(native_tools.len(), 2);
        assert!(native_tools[0].cache_control.is_none());
        assert!(native_tools[1].cache_control.is_some());
    }

    #[test]
    fn convert_tools_single_tool_gets_cache() {
        let tools = vec![ToolSpec {
            name: "tool1".to_string(),
            description: "Only tool".to_string(),
            parameters: serde_json::json!({"type": "object"}),
        }];

        let native_tools = AnthropicProvider::convert_tools(Some(&tools)).unwrap();

        assert_eq!(native_tools.len(), 1);
        assert!(native_tools[0].cache_control.is_some());
    }

    #[test]
    fn convert_messages_small_system_prompt_uses_blocks_with_cache() {
        let messages = vec![ChatMessage {
            role: "system".to_string(),
            content: "Short system prompt".to_string(),
        }];

        let (system_prompt, _) = AnthropicProvider::convert_messages(&messages);

        match system_prompt.unwrap() {
            SystemPrompt::Blocks(blocks) => {
                assert_eq!(blocks.len(), 1);
                assert_eq!(blocks[0].text, "Short system prompt");
                assert!(
                    blocks[0].cache_control.is_some(),
                    "Small system prompts should have cache_control"
                );
            }
            SystemPrompt::String(_) => {
                panic!("Expected Blocks variant with cache_control for small prompt")
            }
        }
    }

    #[test]
    fn convert_messages_large_system_prompt() {
        let large_content = "a".repeat(3073);
        let messages = vec![ChatMessage {
            role: "system".to_string(),
            content: large_content.clone(),
        }];

        let (system_prompt, _) = AnthropicProvider::convert_messages(&messages);

        match system_prompt.unwrap() {
            SystemPrompt::Blocks(blocks) => {
                assert_eq!(blocks.len(), 1);
                assert_eq!(blocks[0].text, large_content);
                assert!(blocks[0].cache_control.is_some());
            }
            SystemPrompt::String(_) => panic!("Expected Blocks variant for large prompt"),
        }
    }

    #[test]
    fn native_chat_request_with_blocks_system() {
        // System prompts now always use Blocks format with cache_control
        let req = NativeChatRequest {
            model: "claude-3-opus".to_string(),
            max_tokens: 4096,
            system: Some(SystemPrompt::Blocks(vec![SystemBlock {
                block_type: "text".to_string(),
                text: "System".to_string(),
                cache_control: Some(CacheControl::ephemeral()),
            }])),
            messages: vec![NativeMessage {
                role: "user".to_string(),
                content: vec![NativeContentOut::Text {
                    text: "Hello".to_string(),
                    cache_control: None,
                }],
            }],
            temperature: 0.7,
            tools: None,
            tool_choice: None,
            stream: None,
        };

        let json = serde_json::to_string(&req).unwrap();
        assert!(json.contains("System"));
        assert!(
            json.contains(r#""cache_control":{"type":"ephemeral"}"#),
            "System prompt should include cache_control"
        );
    }

    #[tokio::test]
    async fn warmup_without_key_is_noop() {
        let provider = AnthropicProvider::new(None);
        let result = provider.warmup().await;
        assert!(result.is_ok());
    }

    #[test]
    fn convert_messages_preserves_multi_turn_history() {
        let messages = vec![
            ChatMessage {
                role: "system".to_string(),
                content: "You are helpful.".to_string(),
            },
            ChatMessage {
                role: "user".to_string(),
                content: "gen a 2 sum in golang".to_string(),
            },
            ChatMessage {
                role: "assistant".to_string(),
                content: "```go\nfunc twoSum(nums []int) {}\n```".to_string(),
            },
            ChatMessage {
                role: "user".to_string(),
                content: "what's meaning of make here?".to_string(),
            },
        ];

        let (system, native_msgs) = AnthropicProvider::convert_messages(&messages);

        // System prompt extracted
        assert!(system.is_some());
        // All 3 non-system messages preserved in order
        assert_eq!(native_msgs.len(), 3);
        assert_eq!(native_msgs[0].role, "user");
        assert_eq!(native_msgs[1].role, "assistant");
        assert_eq!(native_msgs[2].role, "user");
    }

    /// Integration test: spin up a mock Anthropic API server, call chat_with_tools
    /// with a multi-turn conversation + tools, and verify the request body contains
    /// ALL conversation turns and native tool definitions.
    #[tokio::test]
    async fn chat_with_tools_sends_full_history_and_native_tools() {
        use axum::{Json, Router, routing::post};
        use std::sync::{Arc, Mutex};
        use tokio::net::TcpListener;

        // Captured request body for assertion
        let captured: Arc<Mutex<Option<serde_json::Value>>> = Arc::new(Mutex::new(None));
        let captured_clone = captured.clone();

        let app = Router::new().route(
            "/v1/messages",
            post(move |Json(body): Json<serde_json::Value>| {
                let cap = captured_clone.clone();
                async move {
                    *cap.lock().unwrap() = Some(body);
                    // Return a minimal valid Anthropic response
                    Json(serde_json::json!({
                        "id": "msg_test",
                        "type": "message",
                        "role": "assistant",
                        "content": [{"type": "text", "text": "The make function creates a map."}],
                        "model": "claude-opus-4-6",
                        "stop_reason": "end_turn",
                        "usage": {"input_tokens": 100, "output_tokens": 20}
                    }))
                }
            }),
        );

        let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
        let addr = listener.local_addr().unwrap();
        let server_handle = tokio::spawn(async move {
            axum::serve(listener, app).await.unwrap();
        });

        // Create provider pointing at mock server
        let provider = AnthropicProvider {
            credential: Some("test-key".to_string()),
            base_url: format!("http://{addr}"),
            max_tokens: DEFAULT_ANTHROPIC_MAX_TOKENS,
        };

        // Multi-turn conversation: system → user (Go code) → assistant (code response) → user (follow-up)
        let messages = vec![
            ChatMessage::system("You are a helpful assistant."),
            ChatMessage::user("gen a 2 sum in golang"),
            ChatMessage::assistant(
                "```go\nfunc twoSum(nums []int, target int) []int {\n    m := make(map[int]int)\n    for i, n := range nums {\n        if j, ok := m[target-n]; ok {\n            return []int{j, i}\n        }\n        m[n] = i\n    }\n    return nil\n}\n```",
            ),
            ChatMessage::user("what's meaning of make here?"),
        ];

        let tools = vec![serde_json::json!({
            "type": "function",
            "function": {
                "name": "shell",
                "description": "Run a shell command",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "command": {"type": "string"}
                    },
                    "required": ["command"]
                }
            }
        })];

        let result = provider
            .chat_with_tools(&messages, &tools, "claude-opus-4-6", 0.7)
            .await;
        assert!(result.is_ok(), "chat_with_tools failed: {:?}", result.err());

        let body = captured
            .lock()
            .unwrap()
            .take()
            .expect("No request captured");

        // Verify system prompt extracted to top-level field
        let system = &body["system"];
        assert!(
            system.to_string().contains("helpful assistant"),
            "System prompt missing: {system}"
        );

        // Verify ALL conversation turns present in messages array
        let msgs = body["messages"].as_array().expect("messages not an array");
        assert_eq!(
            msgs.len(),
            3,
            "Expected 3 messages (2 user + 1 assistant), got {}",
            msgs.len()
        );

        // Turn 1: user with Go request
        assert_eq!(msgs[0]["role"], "user");
        let turn1_text = msgs[0]["content"].to_string();
        assert!(
            turn1_text.contains("2 sum"),
            "Turn 1 missing Go request: {turn1_text}"
        );

        // Turn 2: assistant with Go code
        assert_eq!(msgs[1]["role"], "assistant");
        let turn2_text = msgs[1]["content"].to_string();
        assert!(
            turn2_text.contains("make(map[int]int)"),
            "Turn 2 missing Go code: {turn2_text}"
        );

        // Turn 3: user follow-up
        assert_eq!(msgs[2]["role"], "user");
        let turn3_text = msgs[2]["content"].to_string();
        assert!(
            turn3_text.contains("meaning of make"),
            "Turn 3 missing follow-up: {turn3_text}"
        );

        // Verify native tools are present
        let api_tools = body["tools"].as_array().expect("tools not an array");
        assert_eq!(api_tools.len(), 1);
        assert_eq!(api_tools[0]["name"], "shell");
        assert!(
            api_tools[0]["input_schema"].is_object(),
            "Missing input_schema"
        );

        server_handle.abort();
    }

    #[test]
    fn native_response_parses_usage() {
        let json = r#"{
            "content": [{"type": "text", "text": "Hello"}],
            "usage": {"input_tokens": 300, "output_tokens": 75}
        }"#;
        let resp: NativeChatResponse = serde_json::from_str(json).unwrap();
        let result = AnthropicProvider::parse_native_response(resp);
        let usage = result.usage.unwrap();
        assert_eq!(usage.input_tokens, Some(300));
        assert_eq!(usage.output_tokens, Some(75));
    }

    #[test]
    fn native_response_parses_without_usage() {
        let json = r#"{"content": [{"type": "text", "text": "Hello"}]}"#;
        let resp: NativeChatResponse = serde_json::from_str(json).unwrap();
        let result = AnthropicProvider::parse_native_response(resp);
        assert!(result.usage.is_none());
    }

    #[test]
    fn capabilities_returns_vision_and_native_tools() {
        let provider = AnthropicProvider::new(Some("test-key"));
        let caps = provider.capabilities();
        assert!(
            caps.native_tool_calling,
            "Anthropic should support native tool calling"
        );
        assert!(caps.vision, "Anthropic should support vision");
    }

    #[test]
    fn convert_messages_with_image_marker_data_uri() {
        let messages = vec![ChatMessage {
            role: "user".to_string(),
            content: "Check this image: [IMAGE:data:image/jpeg;base64,/9j/4AAQ] What do you see?"
                .to_string(),
        }];

        let (_, native_msgs) = AnthropicProvider::convert_messages(&messages);

        assert_eq!(native_msgs.len(), 1);
        assert_eq!(native_msgs[0].role, "user");
        // Should have 2 content blocks: image + text
        assert_eq!(native_msgs[0].content.len(), 2);

        // First block should be image
        match &native_msgs[0].content[0] {
            NativeContentOut::Image { source } => {
                assert_eq!(source.source_type, "base64");
                assert_eq!(source.media_type, "image/jpeg");
                assert_eq!(source.data, "/9j/4AAQ");
            }
            _ => panic!("Expected Image content block"),
        }

        // Second block should be text (parse_image_markers may leave extra spaces)
        match &native_msgs[0].content[1] {
            NativeContentOut::Text { text, .. } => {
                // The text may have extra spaces where the marker was removed
                assert!(
                    text.contains("Check this image:") && text.contains("What do you see?"),
                    "Expected text to contain 'Check this image:' and 'What do you see?', got: {}",
                    text
                );
            }
            _ => panic!("Expected Text content block"),
        }
    }

    #[test]
    fn convert_messages_with_only_image_marker() {
        let messages = vec![ChatMessage {
            role: "user".to_string(),
            content: "[IMAGE:data:image/png;base64,iVBORw0KGgo]".to_string(),
        }];

        let (_, native_msgs) = AnthropicProvider::convert_messages(&messages);

        assert_eq!(native_msgs.len(), 1);
        assert_eq!(native_msgs[0].content.len(), 2);

        // First block should be image
        match &native_msgs[0].content[0] {
            NativeContentOut::Image { source } => {
                assert_eq!(source.media_type, "image/png");
            }
            _ => panic!("Expected Image content block"),
        }

        // Second block should be placeholder text
        match &native_msgs[0].content[1] {
            NativeContentOut::Text { text, .. } => {
                assert_eq!(text, "[image]");
            }
            _ => panic!("Expected Text content block with [image] placeholder"),
        }
    }

    #[test]
    fn convert_messages_without_image_marker() {
        let messages = vec![ChatMessage {
            role: "user".to_string(),
            content: "Hello, how are you?".to_string(),
        }];

        let (_, native_msgs) = AnthropicProvider::convert_messages(&messages);

        assert_eq!(native_msgs.len(), 1);
        assert_eq!(native_msgs[0].content.len(), 1);

        match &native_msgs[0].content[0] {
            NativeContentOut::Text { text, .. } => {
                assert_eq!(text, "Hello, how are you?");
            }
            _ => panic!("Expected Text content block"),
        }
    }

    #[test]
    fn image_content_serializes_correctly() {
        let content = NativeContentOut::Image {
            source: ImageSource {
                source_type: "base64".to_string(),
                media_type: "image/jpeg".to_string(),
                data: "testdata".to_string(),
            },
        };
        let json = serde_json::to_string(&content).unwrap();
        // The outer "type" is the enum tag, inner "type" (source_type) is renamed
        assert!(json.contains(r#""type":"image""#), "JSON: {}", json);
        assert!(json.contains(r#""type":"base64""#), "JSON: {}", json); // source_type is serialized as "type"
        assert!(
            json.contains(r#""media_type":"image/jpeg""#),
            "JSON: {}",
            json
        );
        assert!(json.contains(r#""data":"testdata""#), "JSON: {}", json);
    }

    #[test]
    fn convert_messages_merges_consecutive_tool_results() {
        // Simulate a multi-tool-call turn: assistant with two tool_use blocks
        // followed by two separate tool result messages.
        let messages = vec![
            ChatMessage {
                role: "system".to_string(),
                content: "You are helpful.".to_string(),
            },
            ChatMessage {
                role: "user".to_string(),
                content: "Do two things.".to_string(),
            },
            ChatMessage {
                role: "assistant".to_string(),
                content: serde_json::json!({
                    "content": "",
                    "tool_calls": [
                        {"id": "call_1", "name": "shell", "arguments": "{\"command\":\"ls\"}"},
                        {"id": "call_2", "name": "shell", "arguments": "{\"command\":\"pwd\"}"}
                    ]
                })
                .to_string(),
            },
            ChatMessage {
                role: "tool".to_string(),
                content: serde_json::json!({
                    "tool_call_id": "call_1",
                    "content": "file1.txt\nfile2.txt"
                })
                .to_string(),
            },
            ChatMessage {
                role: "tool".to_string(),
                content: serde_json::json!({
                    "tool_call_id": "call_2",
                    "content": "/home/user"
                })
                .to_string(),
            },
        ];

        let (system, native_msgs) = AnthropicProvider::convert_messages(&messages);

        assert!(system.is_some());
        // Should be: user, assistant, user (merged tool results)
        // NOT: user, assistant, user, user (which Anthropic rejects)
        assert_eq!(
            native_msgs.len(),
            3,
            "Expected 3 messages (user, assistant, merged tool results), got {}.\nRoles: {:?}",
            native_msgs.len(),
            native_msgs.iter().map(|m| &m.role).collect::<Vec<_>>()
        );
        assert_eq!(native_msgs[0].role, "user");
        assert_eq!(native_msgs[1].role, "assistant");
        assert_eq!(native_msgs[2].role, "user");
        // The merged user message should contain both tool results
        assert_eq!(
            native_msgs[2].content.len(),
            2,
            "Expected 2 tool_result blocks in merged message"
        );
    }

    #[test]
    fn convert_messages_no_adjacent_same_role() {
        // Verify that convert_messages never produces adjacent messages with the
        // same role, regardless of input ordering.
        let messages = vec![
            ChatMessage {
                role: "user".to_string(),
                content: "Hello".to_string(),
            },
            ChatMessage {
                role: "assistant".to_string(),
                content: serde_json::json!({
                    "content": "I'll run a command",
                    "tool_calls": [
                        {"id": "tc1", "name": "shell", "arguments": "{\"command\":\"echo hi\"}"}
                    ]
                })
                .to_string(),
            },
            ChatMessage {
                role: "tool".to_string(),
                content: serde_json::json!({
                    "tool_call_id": "tc1",
                    "content": "hi"
                })
                .to_string(),
            },
            ChatMessage {
                role: "user".to_string(),
                content: "Thanks!".to_string(),
            },
        ];

        let (_system, native_msgs) = AnthropicProvider::convert_messages(&messages);

        for window in native_msgs.windows(2) {
            assert_ne!(
                window[0].role, window[1].role,
                "Adjacent messages must not share the same role: found two '{}' messages in a row",
                window[0].role
            );
        }
    }
}