kcode-openai-api 0.2.2

use std::fmt;

use base64::{Engine as _, engine::general_purpose::STANDARD};
use serde_json::{Value, json};

use crate::{DEFAULT_TRANSCRIPTION_PROMPT, Error, GPT_5_6, Result};

pub(crate) const MAX_AUDIO_BYTES: usize = 25 * 1024 * 1024;
const MAX_AUDIO_FILENAME_CHARACTERS: usize = 120;
const MAX_IMAGE_ANALYSIS_BYTES: usize = 20 * 1024 * 1024;
const MAX_IMAGE_ANALYSIS_PROMPT_CHARACTERS: usize = 32_000;
const MAX_IMAGE_PROMPT_CHARACTERS: usize = 32_000;
const MIN_IMAGE_PIXELS: u64 = 655_360;
const MAX_IMAGE_PIXELS: u64 = 8_294_400;
const MAX_IMAGE_EDGE: u16 = 3_840;

/// One in-memory audio file accepted by `gpt-4o-transcribe`.
#[derive(Clone, Eq, PartialEq)]
pub struct AudioInput {
    file_name: String,
    mime_type: String,
    data: Vec<u8>,
}

impl AudioInput {
    /// Constructs and validates an audio input.
    pub fn new(
        file_name: impl Into<String>,
        mime_type: impl Into<String>,
        data: Vec<u8>,
    ) -> Result<Self> {
        let value = Self {
            file_name: file_name.into(),
            mime_type: mime_type.into().to_ascii_lowercase(),
            data,
        };
        value.validate()?;
        Ok(value)
    }

    /// Returns the validated upload filename.
    pub fn file_name(&self) -> &str {
        &self.file_name
    }

    /// Returns the validated audio MIME type.
    pub fn mime_type(&self) -> &str {
        &self.mime_type
    }

    /// Returns the raw audio bytes.
    pub fn data(&self) -> &[u8] {
        &self.data
    }

    /// Returns the raw audio byte count.
    pub fn len(&self) -> usize {
        self.data.len()
    }

    /// Returns whether the audio input contains no bytes.
    pub fn is_empty(&self) -> bool {
        self.data.is_empty()
    }

    pub(crate) fn into_parts(self) -> (String, String, Vec<u8>) {
        (self.file_name, self.mime_type, self.data)
    }

    fn validate(&self) -> Result<()> {
        if self.file_name.is_empty()
            || self.file_name.chars().count() > MAX_AUDIO_FILENAME_CHARACTERS
            || !self.file_name.chars().all(|character| {
                character.is_ascii_alphanumeric() || matches!(character, '.' | '-' | '_')
            })
            || matches!(self.file_name.as_str(), "." | "..")
        {
            return Err(Error::InvalidInput(format!(
                "audio filename must contain 1 through {MAX_AUDIO_FILENAME_CHARACTERS} ASCII letters, digits, dots, hyphens, or underscores"
            )));
        }
        if !matches!(
            self.mime_type.as_str(),
            "audio/flac"
                | "audio/x-flac"
                | "audio/m4a"
                | "audio/mp3"
                | "audio/mp4"
                | "audio/mpeg"
                | "audio/mpga"
                | "audio/ogg"
                | "audio/opus"
                | "audio/wav"
                | "audio/x-wav"
                | "audio/webm"
                | "application/ogg"
                | "video/mp4"
                | "video/webm"
        ) {
            return Err(Error::InvalidInput(
                "audio MIME type must describe a supported FLAC, MP3, MP4, M4A, OGG, WAV, or WebM recording".into(),
            ));
        }
        let extension = self
            .file_name
            .rsplit_once('.')
            .map(|(_, extension)| extension.to_ascii_lowercase());
        if !matches!(
            extension.as_deref(),
            Some("flac" | "mp3" | "mp4" | "mpeg" | "mpga" | "m4a" | "ogg" | "wav" | "webm")
        ) {
            return Err(Error::InvalidInput(
                "audio filename must use a supported flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, or webm extension".into(),
            ));
        }
        if self.data.is_empty() || self.data.len() > MAX_AUDIO_BYTES {
            return Err(Error::InvalidInput(format!(
                "audio must contain between 1 and {MAX_AUDIO_BYTES} bytes"
            )));
        }
        Ok(())
    }
}

impl fmt::Debug for AudioInput {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("AudioInput")
            .field("file_name", &self.file_name)
            .field("mime_type", &self.mime_type)
            .field("bytes", &self.data.len())
            .finish()
    }
}

/// One audio transcription request.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct TranscriptionRequest {
    /// Audio file to transcribe.
    pub audio: AudioInput,
    /// Optional text that guides transcription style and context.
    pub prompt: Option<String>,
    /// Optional ISO-639-1 input language code, such as `en`.
    pub language: Option<String>,
}

impl TranscriptionRequest {
    /// Constructs a request with Kennedy's current faithful-transcription prompt.
    pub fn new(audio: AudioInput) -> Self {
        Self {
            audio,
            prompt: Some(DEFAULT_TRANSCRIPTION_PROMPT.into()),
            language: None,
        }
    }

    pub(crate) fn validate(&self) -> Result<()> {
        self.audio.validate()?;
        if let Some(prompt) = &self.prompt
            && prompt.trim().is_empty()
        {
            return Err(Error::InvalidInput(
                "transcription prompt must not be blank when supplied".into(),
            ));
        }
        if let Some(language) = &self.language
            && (language.len() != 2 || !language.bytes().all(|value| value.is_ascii_lowercase()))
        {
            return Err(Error::InvalidInput(
                "transcription language must be a two-letter lowercase ISO-639-1 code".into(),
            ));
        }
        Ok(())
    }
}

/// Modality detail for token-billed audio transcription input.
#[derive(Clone, Debug, Default, Eq, PartialEq)]
pub struct TranscriptionTokenDetails {
    /// Audio tokens billed for the request, when reported.
    pub audio_tokens: Option<u64>,
    /// Prompt text tokens billed for the request, when reported.
    pub text_tokens: Option<u64>,
}

/// Token-billed transcription usage.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct TranscriptionTokenUsage {
    /// Input tokens billed for the request.
    pub input_tokens: u64,
    /// Output tokens generated by the request.
    pub output_tokens: u64,
    /// Total input and output tokens.
    pub total_tokens: u64,
    /// Optional input modality breakdown.
    pub input_details: Option<TranscriptionTokenDetails>,
}

/// Usage returned for a transcription request.
#[derive(Clone, Debug, PartialEq)]
pub enum TranscriptionUsage {
    /// Usage billed in tokens.
    Tokens(TranscriptionTokenUsage),
    /// Usage billed from audio duration in seconds.
    DurationSeconds(f64),
}

/// Normalized `gpt-4o-transcribe` result.
#[derive(Clone, Debug, PartialEq)]
pub struct Transcription {
    /// Complete non-empty transcript text.
    pub text: String,
    /// Provider usage, when returned.
    pub usage: Option<TranscriptionUsage>,
    /// OpenAI request identifier, when returned.
    pub request_id: Option<String>,
}

/// Stored image media accepted for image analysis.
#[derive(Clone, Copy, Debug, Eq, Hash, PartialEq)]
pub enum ImageMediaType {
    /// PNG image data.
    Png,
    /// JPEG image data.
    Jpeg,
    /// WebP image data.
    WebP,
    /// GIF image data. OpenAI accepts non-animated GIF input.
    Gif,
}

impl ImageMediaType {
    /// Returns the MIME type sent to OpenAI.
    pub const fn mime_type(self) -> &'static str {
        match self {
            Self::Png => "image/png",
            Self::Jpeg => "image/jpeg",
            Self::WebP => "image/webp",
            Self::Gif => "image/gif",
        }
    }
}

/// One caller-owned in-memory image accepted for image analysis.
#[derive(Clone, Eq, PartialEq)]
pub struct ImageInput {
    media_type: ImageMediaType,
    data: Vec<u8>,
}

impl ImageInput {
    /// Constructs and validates an image input.
    pub fn new(media_type: ImageMediaType, data: Vec<u8>) -> Result<Self> {
        let value = Self { media_type, data };
        value.validate()?;
        Ok(value)
    }

    /// Returns the declared stored-image media type.
    pub const fn media_type(&self) -> ImageMediaType {
        self.media_type
    }

    /// Returns the raw image bytes.
    pub fn data(&self) -> &[u8] {
        &self.data
    }

    /// Returns the raw image byte count.
    pub fn len(&self) -> usize {
        self.data.len()
    }

    /// Returns whether the image contains no bytes.
    pub fn is_empty(&self) -> bool {
        self.data.is_empty()
    }

    fn validate(&self) -> Result<()> {
        if self.data.is_empty() || self.data.len() > MAX_IMAGE_ANALYSIS_BYTES {
            return Err(Error::InvalidInput(format!(
                "analysis image must contain between 1 and {MAX_IMAGE_ANALYSIS_BYTES} bytes"
            )));
        }
        Ok(())
    }
}

impl fmt::Debug for ImageInput {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("ImageInput")
            .field("media_type", &self.media_type)
            .field("bytes", &self.data.len())
            .finish()
    }
}

/// OpenAI image-understanding detail level.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum ImageDetail {
    /// Let the fixed model select the detail level.
    #[default]
    Auto,
    /// Use low-detail image understanding.
    Low,
    /// Use high-detail image understanding.
    High,
    /// Use original-resolution image understanding where supported.
    Original,
}

impl ImageDetail {
    pub(crate) const fn as_str(self) -> &'static str {
        match self {
            Self::Auto => "auto",
            Self::Low => "low",
            Self::High => "high",
            Self::Original => "original",
        }
    }
}

/// One prompt-driven image-analysis request.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageAnalysisRequest {
    /// Caller-owned image bytes and declared media type.
    pub image: ImageInput,
    /// Exact prompt paired with the image.
    pub prompt: String,
    /// Provider image-understanding detail level.
    pub detail: ImageDetail,
}

impl ImageAnalysisRequest {
    /// Constructs a request using automatic image detail.
    pub fn new(image: ImageInput, prompt: impl Into<String>) -> Self {
        Self {
            image,
            prompt: prompt.into(),
            detail: ImageDetail::Auto,
        }
    }

    pub(crate) fn validate(&self) -> Result<()> {
        self.image.validate()?;
        if self.prompt.trim().is_empty()
            || self.prompt.chars().count() > MAX_IMAGE_ANALYSIS_PROMPT_CHARACTERS
        {
            return Err(Error::InvalidInput(format!(
                "image-analysis prompt must contain 1 through {MAX_IMAGE_ANALYSIS_PROMPT_CHARACTERS} characters"
            )));
        }
        Ok(())
    }

    pub(crate) fn payload(&self) -> Value {
        let image_url = format!(
            "data:{};base64,{}",
            self.image.media_type.mime_type(),
            STANDARD.encode(&self.image.data)
        );
        json!({
            "model": GPT_5_6,
            "store": false,
            "input": [{
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": self.prompt
                    },
                    {
                        "type": "input_image",
                        "image_url": image_url,
                        "detail": self.detail.as_str()
                    }
                ]
            }]
        })
    }
}

/// Completion state of a non-streaming image-analysis response.
#[derive(Clone, Debug, Eq, PartialEq)]
pub enum ImageAnalysisStatus {
    /// OpenAI completed the response.
    Completed,
    /// OpenAI returned a valid partial response.
    Incomplete {
        /// Provider reason for incompleteness, when returned.
        reason: Option<String>,
    },
}

/// Documented Responses API token usage retained for image analysis.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageAnalysisUsage {
    /// Total input tokens.
    pub input_tokens: u64,
    /// Total output tokens.
    pub output_tokens: u64,
    /// Total input and output tokens.
    pub total_tokens: u64,
    /// Cached input tokens, when reported.
    pub cached_input_tokens: Option<u64>,
    /// Cache-write input tokens, when reported.
    pub cache_write_input_tokens: Option<u64>,
    /// Reasoning output tokens, when reported.
    pub reasoning_output_tokens: Option<u64>,
}

/// Normalized prompt-driven image-analysis result.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageAnalysis {
    /// Ordered non-empty assistant text.
    pub text: String,
    /// OpenAI response identifier.
    pub response_id: String,
    /// Model identifier returned by OpenAI.
    pub model: String,
    /// Completion or partial-result status.
    pub status: ImageAnalysisStatus,
    /// Provider token usage, when returned.
    pub usage: Option<ImageAnalysisUsage>,
    /// OpenAI request identifier, when returned.
    pub request_id: Option<String>,
}

/// Generated image dimensions.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum ImageSize {
    /// Let GPT Image choose the dimensions.
    #[default]
    Auto,
    /// Explicit width and height in pixels.
    Dimensions {
        /// Width in pixels.
        width: u16,
        /// Height in pixels.
        height: u16,
    },
}

impl ImageSize {
    /// Constructs explicit dimensions after validating GPT Image 2 constraints.
    pub fn dimensions(width: u16, height: u16) -> Result<Self> {
        let value = Self::Dimensions { width, height };
        value.validate()?;
        Ok(value)
    }

    pub(crate) fn as_api_value(self) -> String {
        match self {
            Self::Auto => "auto".into(),
            Self::Dimensions { width, height } => format!("{width}x{height}"),
        }
    }

    pub(crate) fn validate(self) -> Result<()> {
        let Self::Dimensions { width, height } = self else {
            return Ok(());
        };
        let pixels = u64::from(width).saturating_mul(u64::from(height));
        let short = width.min(height);
        let long = width.max(height);
        if width % 16 != 0
            || height % 16 != 0
            || width > MAX_IMAGE_EDGE
            || height > MAX_IMAGE_EDGE
            || short == 0
            || u32::from(long) > u32::from(short).saturating_mul(3)
            || !(MIN_IMAGE_PIXELS..=MAX_IMAGE_PIXELS).contains(&pixels)
        {
            return Err(Error::InvalidInput(format!(
                "GPT Image 2 dimensions must be multiples of 16, no edge may exceed {MAX_IMAGE_EDGE}, the aspect ratio must be at most 3:1, and total pixels must be between {MIN_IMAGE_PIXELS} and {MAX_IMAGE_PIXELS}"
            )));
        }
        Ok(())
    }
}

/// GPT Image rendering quality.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum ImageQuality {
    /// Let GPT Image select quality.
    #[default]
    Auto,
    /// Fast draft quality.
    Low,
    /// Balanced quality.
    Medium,
    /// Highest supported quality.
    High,
}

impl ImageQuality {
    pub(crate) const fn as_str(self) -> &'static str {
        match self {
            Self::Auto => "auto",
            Self::Low => "low",
            Self::Medium => "medium",
            Self::High => "high",
        }
    }

    pub(crate) fn parse(value: &str) -> Option<Self> {
        match value {
            "auto" => Some(Self::Auto),
            "low" => Some(Self::Low),
            "medium" => Some(Self::Medium),
            "high" => Some(Self::High),
            _ => None,
        }
    }
}

/// GPT Image output file format.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum ImageFormat {
    /// PNG output.
    #[default]
    Png,
    /// JPEG output.
    Jpeg,
    /// WebP output.
    WebP,
}

impl ImageFormat {
    /// Returns the output MIME type.
    pub const fn mime_type(self) -> &'static str {
        match self {
            Self::Png => "image/png",
            Self::Jpeg => "image/jpeg",
            Self::WebP => "image/webp",
        }
    }

    pub(crate) const fn as_str(self) -> &'static str {
        match self {
            Self::Png => "png",
            Self::Jpeg => "jpeg",
            Self::WebP => "webp",
        }
    }

    pub(crate) fn parse(value: &str) -> Option<Self> {
        match value {
            "png" => Some(Self::Png),
            "jpeg" => Some(Self::Jpeg),
            "webp" => Some(Self::WebP),
            _ => None,
        }
    }
}

/// Background selection supported by GPT Image 2.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum ImageBackground {
    /// Let GPT Image select the background.
    #[default]
    Auto,
    /// Require an opaque background.
    Opaque,
}

impl ImageBackground {
    pub(crate) const fn as_str(self) -> &'static str {
        match self {
            Self::Auto => "auto",
            Self::Opaque => "opaque",
        }
    }
}

/// Provider content-moderation level.
#[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
pub enum Moderation {
    /// Standard provider moderation.
    #[default]
    Auto,
    /// Less restrictive provider moderation.
    Low,
}

impl Moderation {
    pub(crate) const fn as_str(self) -> &'static str {
        match self {
            Self::Auto => "auto",
            Self::Low => "low",
        }
    }
}

/// One GPT Image 2 text-to-image request.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageGenerationRequest {
    /// Text description of the desired image.
    pub prompt: String,
    /// Output dimensions.
    pub size: ImageSize,
    /// Rendering quality.
    pub quality: ImageQuality,
    /// Output file format.
    pub output_format: ImageFormat,
    /// JPEG or WebP compression from 0 through 100, when explicitly requested.
    pub output_compression: Option<u8>,
    /// Background behavior. GPT Image 2 does not support transparency.
    pub background: ImageBackground,
    /// Provider moderation level.
    pub moderation: Moderation,
    /// Optional stable end-user identifier supplied to OpenAI abuse monitoring.
    pub user: Option<String>,
}

/// One GPT Image 2 request that modifies or combines reference images.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageEditRequest {
    /// One or more source or reference images.
    pub images: Vec<ImageInput>,
    /// Text description of the requested result.
    pub prompt: String,
    /// Output dimensions.
    pub size: ImageSize,
    /// Rendering quality.
    pub quality: ImageQuality,
    /// Output file format.
    pub output_format: ImageFormat,
    /// JPEG or WebP compression from 0 through 100, when explicitly requested.
    pub output_compression: Option<u8>,
    /// Background behavior.
    pub background: ImageBackground,
    /// Provider moderation level.
    pub moderation: Moderation,
    /// Optional stable end-user identifier supplied to abuse monitoring.
    pub user: Option<String>,
}

impl ImageEditRequest {
    /// Constructs a single-reference edit with automatic size and quality.
    pub fn new(image: ImageInput, prompt: impl Into<String>) -> Self {
        Self {
            images: vec![image],
            prompt: prompt.into(),
            size: ImageSize::Auto,
            quality: ImageQuality::Auto,
            output_format: ImageFormat::Png,
            output_compression: None,
            background: ImageBackground::Auto,
            moderation: Moderation::Auto,
            user: None,
        }
    }

    pub(crate) fn validate(&self) -> Result<()> {
        if self.images.is_empty() || self.images.len() > 16 {
            return Err(Error::InvalidInput(
                "image edit requires between 1 and 16 reference images".into(),
            ));
        }
        for image in &self.images {
            image.validate()?;
        }
        let generation = ImageGenerationRequest {
            prompt: self.prompt.clone(),
            size: self.size,
            quality: self.quality,
            output_format: self.output_format,
            output_compression: self.output_compression,
            background: self.background,
            moderation: self.moderation,
            user: self.user.clone(),
        };
        generation.validate()
    }
}

impl ImageGenerationRequest {
    /// Constructs a single-image request with automatic size and quality and PNG output.
    pub fn new(prompt: impl Into<String>) -> Self {
        Self {
            prompt: prompt.into(),
            size: ImageSize::Auto,
            quality: ImageQuality::Auto,
            output_format: ImageFormat::Png,
            output_compression: None,
            background: ImageBackground::Auto,
            moderation: Moderation::Auto,
            user: None,
        }
    }

    pub(crate) fn validate(&self) -> Result<()> {
        if self.prompt.trim().is_empty()
            || self.prompt.chars().count() > MAX_IMAGE_PROMPT_CHARACTERS
        {
            return Err(Error::InvalidInput(format!(
                "image prompt must contain 1 through {MAX_IMAGE_PROMPT_CHARACTERS} characters"
            )));
        }
        self.size.validate()?;
        if self.output_compression.is_some_and(|value| value > 100) {
            return Err(Error::InvalidInput(
                "output compression must be between 0 and 100".into(),
            ));
        }
        if self.output_compression.is_some() && self.output_format == ImageFormat::Png {
            return Err(Error::InvalidInput(
                "output compression is supported only for JPEG and WebP images".into(),
            ));
        }
        if let Some(user) = &self.user
            && (user.trim().is_empty()
                || user.chars().count() > 512
                || user.chars().any(char::is_control))
        {
            return Err(Error::InvalidInput(
                "image user identifier must contain 1 through 512 non-control characters when supplied".into(),
            ));
        }
        Ok(())
    }

    pub(crate) fn payload(&self) -> Value {
        let mut payload = json!({
            "model": crate::GPT_IMAGE_2,
            "prompt": self.prompt,
            "n": 1,
            "size": self.size.as_api_value(),
            "quality": self.quality.as_str(),
            "output_format": self.output_format.as_str(),
            "background": self.background.as_str(),
            "moderation": self.moderation.as_str(),
            "stream": false
        });
        let object = payload.as_object_mut().expect("image payload is an object");
        if let Some(compression) = self.output_compression {
            object.insert("output_compression".into(), json!(compression));
        }
        if let Some(user) = &self.user {
            object.insert("user".into(), json!(user));
        }
        payload
    }
}

/// One generated image held in memory.
#[derive(Clone, Eq, PartialEq)]
pub struct GeneratedImage {
    /// Decoded image bytes.
    pub data: Vec<u8>,
    /// Returned image file format.
    pub format: ImageFormat,
}

impl fmt::Debug for GeneratedImage {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("GeneratedImage")
            .field("format", &self.format)
            .field("bytes", &self.data.len())
            .finish()
    }
}

/// Text and image token detail for GPT Image usage.
#[derive(Clone, Debug, Default, Eq, PartialEq)]
pub struct ImageTokenDetails {
    /// Text tokens reported for the modality side.
    pub text_tokens: u64,
    /// Image tokens reported for the modality side.
    pub image_tokens: u64,
}

/// Token usage returned by GPT Image.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct ImageUsage {
    /// Total input tokens.
    pub input_tokens: u64,
    /// Total output tokens.
    pub output_tokens: u64,
    /// Total input and output tokens.
    pub total_tokens: u64,
    /// Input text and image token breakdown.
    pub input_details: ImageTokenDetails,
    /// Output text and image token breakdown, when returned.
    pub output_details: Option<ImageTokenDetails>,
}

/// Normalized single-image GPT Image 2 result.
#[derive(Clone, Debug, PartialEq)]
pub struct ImageGeneration {
    /// Provider creation time as Unix seconds.
    pub created: u64,
    /// Decoded generated image.
    pub image: GeneratedImage,
    /// Actual provider-selected size, when returned.
    pub size: Option<String>,
    /// Actual provider-selected quality, when returned.
    pub quality: Option<ImageQuality>,
    /// Provider token usage, when returned.
    pub usage: Option<ImageUsage>,
    /// OpenAI request identifier, when returned.
    pub request_id: Option<String>,
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn transcription_accepts_long_prompt_unchanged_and_rejects_blank_prompt() {
        let audio = AudioInput::new("note.webm", "audio/webm", vec![1]).unwrap();
        let mut request = TranscriptionRequest::new(audio);
        let prompt = "x".repeat(4 * 1024 * 1024 + 1);
        request.prompt = Some(prompt.clone());
        request.validate().unwrap();
        assert_eq!(request.prompt.as_deref(), Some(prompt.as_str()));

        request.prompt = Some(" \n\t".into());
        assert!(request.validate().is_err());
    }

    #[test]
    fn audio_debug_omits_bytes_and_rejects_unsafe_names() {
        let audio = AudioInput::new("note.webm", "audio/webm", vec![7, 8, 9]).unwrap();
        let debug = format!("{audio:?}");
        assert!(debug.contains("bytes: 3"));
        assert!(!debug.contains("7, 8, 9"));
        assert!(AudioInput::new("../note.webm", "audio/webm", vec![1]).is_err());
    }

    #[test]
    fn image_analysis_input_is_bounded_and_redacted() {
        let image = ImageInput::new(ImageMediaType::Png, vec![1, 2, 3]).unwrap();
        let debug = format!("{image:?}");
        assert!(debug.contains("Png"));
        assert!(debug.contains("bytes: 3"));
        assert!(!debug.contains("1, 2, 3"));
        assert!(ImageInput::new(ImageMediaType::Png, Vec::new()).is_err());
    }

    #[test]
    fn image_analysis_payload_preserves_prompt_and_has_no_output_cap() {
        let image = ImageInput::new(ImageMediaType::Jpeg, vec![1, 2, 3]).unwrap();
        let mut request = ImageAnalysisRequest::new(image, "  Explain this image.  ");
        request.detail = ImageDetail::High;
        request.validate().unwrap();

        let payload = request.payload();
        assert_eq!(payload["model"], "gpt-5.6");
        assert_eq!(payload["store"], false);
        assert_eq!(
            payload["input"][0]["content"][0]["text"],
            "  Explain this image.  "
        );
        assert_eq!(
            payload["input"][0]["content"][1]["image_url"],
            "data:image/jpeg;base64,AQID"
        );
        assert_eq!(payload["input"][0]["content"][1]["detail"], "high");
        assert!(payload.get("max_output_tokens").is_none());
        assert!(payload.get("tools").is_none());
    }

    #[test]
    fn image_dimensions_enforce_current_gpt_image_2_constraints() {
        assert_eq!(
            ImageSize::dimensions(2048, 2048).unwrap(),
            ImageSize::Dimensions {
                width: 2048,
                height: 2048
            }
        );
        assert!(ImageSize::dimensions(1000, 1000).is_err());
        assert!(ImageSize::dimensions(3840, 3840).is_err());
        assert!(ImageSize::dimensions(3072, 1024).is_ok());
        assert!(ImageSize::dimensions(3088, 1024).is_err());
    }

    #[test]
    fn png_rejects_compression_but_jpeg_accepts_it() {
        let mut request = ImageGenerationRequest::new("draw a lighthouse");
        request.output_compression = Some(80);
        assert!(request.validate().is_err());
        request.output_format = ImageFormat::Jpeg;
        assert!(request.validate().is_ok());
        request.output_compression = Some(101);
        assert!(request.validate().is_err());
    }

    #[test]
    fn image_payload_is_single_shot_gpt_image_2() {
        let request = ImageGenerationRequest::new("draw a lighthouse");
        let payload = request.payload();
        assert_eq!(payload["model"], "gpt-image-2");
        assert_eq!(payload["n"], 1);
        assert_eq!(payload["stream"], false);
        assert_eq!(payload["background"], "auto");
        assert_eq!(payload["output_format"], "png");
    }

    #[test]
    fn image_edits_require_a_bounded_ordered_reference_set() {
        let first = ImageInput::new(ImageMediaType::Png, vec![1]).unwrap();
        let mut request = ImageEditRequest::new(first, "make the sky darker");
        request
            .images
            .push(ImageInput::new(ImageMediaType::Jpeg, vec![2]).unwrap());
        assert!(request.validate().is_ok());
        assert_eq!(request.images[0].media_type(), ImageMediaType::Png);
        assert_eq!(request.images[1].media_type(), ImageMediaType::Jpeg);

        request.images.clear();
        assert!(request.validate().is_err());
    }
}