add gemini flash to inpainting

This commit is contained in:
Brandon Thomas committed 2025-08-28 20:53:53 -04:00
1 parent dc696eb341
commit 841243cf41
6 files changed
+261 -4

No files matched your search

@@ -4,6 +4,7 @@ use crate::core::commands::enqueue::image_inpaint::errors::InternalImageInpaintE
use crate::core::commands::enqueue::image_inpaint::flux_dev_juggernaut_inpaint::handle_flux_dev_juggernaut_inpaint::handle_flux_dev_juggernaut_inpaint;
use crate::core::commands::enqueue::image_inpaint::flux_pro_1_inpaint::handle_flux_pro_1_inpaint::handle_flux_pro_1_inpaint;
use crate::core::commands::enqueue::image_inpaint::flux_pro_kontext_inpaint::handle_flux_pro_kontext_inpaint::handle_flux_pro_kontext_inpaint;
use crate::core::commands::enqueue::image_inpaint::gemini_25_flash_inpaint::handle_gemini_25_flash_inpaint::handle_gemini_25_flash_inpaint;
use crate::core::commands::enqueue::task_enqueue_success::TaskEnqueueSuccess;
use crate::core::commands::response::failure_response_wrapper::{CommandErrorResponseWrapper, CommandErrorStatus};
use crate::core::commands::response::shorthand::{Response, ResponseOrErrorType};
@@ -48,16 +49,26 @@ use tokens::tokens::media_files::MediaFileToken;
#[derive(Deserialize, Debug, Copy, Clone)]
#[serde(rename_all = "snake_case")]
pub enum ImageInpaintModel {
// Mask-based inpainting models
/*
* Mask-based inpainting models
*/
#[serde(rename = "flux_dev_juggernaut")]
FluxDevJuggernaut,
#[serde(rename = "flux_pro_1")]
FluxPro1,
// Non-inpainting, instructive editing models
// NB: We're supporting these to keep the app simple and not over-complicate the javascript
/*
* Non-inpainting, instructive editing models
* NB: We're supporting these to keep the app simple and not over-complicate the javascript
*/
#[serde(rename = "flux_pro_kontext_max")]
FluxProKontextMax,
#[serde(rename = "gemini_25_flash")]
Gemini25Flash,
}
#[derive(Deserialize, Debug)]
@@ -250,7 +261,7 @@ pub async fn handle_request(
).await?
}
Some(ImageInpaintModel::FluxProKontextMax) => {
// Redirect to contextual image edit handler.
// TODO: Redirect to contextual image edit handler.
handle_flux_pro_kontext_inpaint(
request,
app,
@@ -264,6 +275,21 @@ pub async fn handle_request(
sora_task_queue,
).await?
}
Some(ImageInpaintModel::Gemini25Flash) => {
// TODO: Redirect to contextual image edit handler.
handle_gemini_25_flash_inpaint(
request,
app,
app_data_root,
app_env_configs,
provider_priority_store,
storyteller_creds_manager,
fal_creds_manager,
fal_task_queue,
sora_creds_manager,
sora_task_queue,
).await?
}
};
let result = success_event
@@ -0,0 +1,60 @@
use crate::core::commands::enqueue::image_edit::enqueue_contextual_edit_image_command::EnqueueContextualEditImageCommand;
use crate::core::commands::enqueue::image_edit::errors::InternalContextualEditImageError;
use crate::core::commands::enqueue::image_inpaint::enqueue_image_inpaint_command::EnqueueInpaintImageCommand;
use crate::core::commands::enqueue::image_inpaint::errors::InternalImageInpaintError;
use crate::core::commands::enqueue::image_inpaint::flux_pro_kontext_inpaint::handle_flux_pro_kontext_inpaint_artcraft::handle_flux_pro_kontext_inpaint_artcraft;
use crate::core::commands::enqueue::image_inpaint::gemini_25_flash_inpaint::handle_gemini_25_flash_inpaint_artcraft::handle_gemini_25_flash_inpaint_artcraft;
use crate::core::commands::enqueue::task_enqueue_success::TaskEnqueueSuccess;
use crate::core::state::app_env_configs::app_env_configs::AppEnvConfigs;
use crate::core::state::data_dir::app_data_root::AppDataRoot;
use crate::core::state::provider_priority::{Provider, ProviderPriorityStore};
use crate::services::fal::state::fal_credential_manager::FalCredentialManager;
use crate::services::fal::state::fal_task_queue::FalTaskQueue;
use crate::services::sora::state::sora_credential_manager::SoraCredentialManager;
use crate::services::sora::state::sora_task_queue::SoraTaskQueue;
use crate::services::storyteller::state::storyteller_credential_manager::StorytellerCredentialManager;
use log::info;
use tauri::AppHandle;
pub async fn handle_gemini_25_flash_inpaint(
request: &EnqueueInpaintImageCommand,
app: &AppHandle,
app_data_root: &AppDataRoot,
app_env_configs: &AppEnvConfigs,
provider_priority_store: &ProviderPriorityStore,
storyteller_creds_manager: &StorytellerCredentialManager,
fal_creds_manager: &FalCredentialManager,
fal_task_queue: &FalTaskQueue,
sora_creds_manager: &SoraCredentialManager,
sora_task_queue: &SoraTaskQueue,
) -> Result<TaskEnqueueSuccess, InternalImageInpaintError> {
let priority = provider_priority_store.get_priority()?;
// TODO: Check if providers are available before proceeding.
info!("Providers by priority: {:?}", priority);
for provider in priority.iter() {
match provider {
Provider::Fal => {
// Fallthrough (for now)
}
Provider::Sora => {
// Fallthrough
}
Provider::Artcraft => {
info!("Dispatching gemini 25 flash inpaint via Artcraft...");
return handle_gemini_25_flash_inpaint_artcraft(
request,
app,
app_data_root,
app_env_configs,
storyteller_creds_manager
).await;
}
}
}
Err(InternalImageInpaintError::NoProviderAvailable)
}
@@ -0,0 +1,167 @@
use crate::core::commands::enqueue::image_bg_removal::enqueue_image_bg_removal_command::EnqueueImageBgRemovalCommand;
use crate::core::commands::enqueue::image_bg_removal::errors::InternalBgRemovalError;
use crate::core::commands::enqueue::image_edit::enqueue_contextual_edit_image_command::{EditImageQuality, EditImageSize, EnqueueContextualEditImageCommand};
use crate::core::commands::enqueue::image_edit::errors::InternalContextualEditImageError;
use crate::core::commands::enqueue::image_inpaint::enqueue_image_inpaint_command::EnqueueInpaintImageCommand;
use crate::core::commands::enqueue::image_inpaint::errors::InternalImageInpaintError;
use crate::core::commands::enqueue::task_enqueue_success::TaskEnqueueSuccess;
use crate::core::events::basic_sendable_event_trait::BasicSendableEvent;
use crate::core::events::generation_events::common::{GenerationAction, GenerationModel, GenerationServiceProvider};
use crate::core::events::generation_events::generation_enqueue_failure_event::GenerationEnqueueFailureEvent;
use crate::core::state::app_env_configs::app_env_configs::AppEnvConfigs;
use crate::core::state::data_dir::app_data_root::AppDataRoot;
use crate::core::state::provider_priority::ProviderPriorityStore;
use crate::core::utils::save_base64_image_to_temp_dir::save_base64_image_to_temp_dir;
use crate::services::fal::state::fal_credential_manager::FalCredentialManager;
use crate::services::fal::state::fal_task_queue::FalTaskQueue;
use crate::services::sora::state::sora_credential_manager::SoraCredentialManager;
use crate::services::sora::state::sora_task_queue::SoraTaskQueue;
use crate::services::storyteller::state::storyteller_credential_manager::StorytellerCredentialManager;
use anyhow::anyhow;
use artcraft_api_defs::generate::image::edit::flux_pro_kontext_max_edit_image::{FluxProKontextMaxEditImageNumImages, FluxProKontextMaxEditImageRequest};
use artcraft_api_defs::generate::image::edit::gemini_25_flash_edit_image::{Gemini25FlashEditImageNumImages, Gemini25FlashEditImageRequest};
use artcraft_api_defs::generate::image::edit::gpt_image_1_edit_image::{GptImage1EditImageImageQuality, GptImage1EditImageImageSize, GptImage1EditImageNumImages, GptImage1EditImageRequest};
use artcraft_api_defs::generate::image::generate_flux_1_dev_text_to_image::GenerateFlux1DevTextToImageRequest;
use artcraft_api_defs::generate::image::generate_flux_1_schnell_text_to_image::GenerateFlux1SchnellTextToImageRequest;
use artcraft_api_defs::generate::image::generate_flux_pro_11_text_to_image::GenerateFluxPro11TextToImageRequest;
use artcraft_api_defs::generate::image::generate_flux_pro_11_ultra_text_to_image::GenerateFluxPro11UltraTextToImageRequest;
use artcraft_api_defs::generate::image::inpaint::flux_pro_1_inpaint_image::{FluxPro1InpaintImageNumImages, FluxPro1InpaintImageRequest};
use enums::common::generation_provider::GenerationProvider;
use enums::tauri::tasks::task_type::TaskType;
use fal_client::requests::queue::image_gen::enqueue_flux_pro_11_ultra_text_to_image::{enqueue_flux_pro_11_ultra_text_to_image, FluxPro11UltraTextToImageArgs};
use idempotency::uuid::generate_random_uuid;
use images::encoding::image_bytes_to_png_bytes::image_bytes_to_png_bytes;
use images::encoding::image_bytes_to_png_bytes_with_dimensions::image_bytes_to_png_bytes_with_dimensions;
use images::mask_images::normalize_image_bytes_to_flux_mask::normalize_image_bytes_to_flux_mask;
use log::{error, info};
use mimetypes::mimetype_info::mimetype_info::MimetypeInfo;
use storyteller_client::credentials::storyteller_credential_set::StorytellerCredentialSet;
use storyteller_client::generate::image::edit::flux_pro_kontext_max_edit_image::flux_pro_kontext_max_edit_image;
use storyteller_client::generate::image::edit::gemini_25_flash_edit_image::gemini_25_flash_edit_image;
use storyteller_client::generate::image::edit::gpt_image_1_edit_image::gpt_image_1_edit_image;
use storyteller_client::generate::image::generate_flux_1_dev_text_to_image::generate_flux_1_dev_text_to_image;
use storyteller_client::generate::image::generate_flux_1_schnell_text_to_image::generate_flux_1_schnell_text_to_image;
use storyteller_client::generate::image::generate_flux_pro_11_text_to_image::generate_flux_pro_11_text_to_image;
use storyteller_client::generate::image::generate_flux_pro_11_ultra_text_to_image::generate_flux_pro_11_ultra_text_to_image;
use storyteller_client::generate::image::inpaint::flux_pro_1_inpaint_image::flux_pro_1_inpaint_image;
use storyteller_client::media_files::get_media_file::get_media_file;
use storyteller_client::media_files::upload_image_media_file_from_bytes::{upload_image_media_file_from_bytes, ImageType, UploadImageBytesArgs};
use storyteller_client::media_files::upload_image_media_file_from_file::upload_image_media_file_from_file;
use tauri::AppHandle;
use tokens::tokens::media_files::MediaFileToken;
pub async fn handle_gemini_25_flash_inpaint_artcraft(
request: &EnqueueInpaintImageCommand,
app: &AppHandle,
app_data_root: &AppDataRoot,
app_env_configs: &AppEnvConfigs,
storyteller_creds_manager: &StorytellerCredentialManager,
) -> Result<TaskEnqueueSuccess, InternalImageInpaintError> {
let creds = match storyteller_creds_manager.get_credentials()? {
Some(creds) => creds,
None => {
return Err(InternalImageInpaintError::NeedsStorytellerCredentials);
},
};
let image_media_token = match &request.image_media_token {
Some(token) => token.clone(),
None => {
return Err(InternalImageInpaintError::NoSourceImageSpecified);
},
};
info!("Calling Artcraft gemini 2.5 flash inpaint ...");
let uuid_idempotency_token = generate_random_uuid();
let num_images = match request.image_count {
None => None,
Some(1) => Some(Gemini25FlashEditImageNumImages::One),
Some(2) => Some(Gemini25FlashEditImageNumImages::Two),
Some(3) => Some(Gemini25FlashEditImageNumImages::Three),
Some(4) => Some(Gemini25FlashEditImageNumImages::Four),
Some(other) => {
return Err(InternalImageInpaintError::InvalidNumberOfRequestedImages {
min: 1,
max: 4,
requested: other,
});
},
};
let request = Gemini25FlashEditImageRequest {
uuid_idempotency_token,
prompt: Some(request.prompt.clone()),
image_media_tokens: Some(vec![image_media_token]),
num_images,
image_quality: None,
};
let result = gemini_25_flash_edit_image(
&app_env_configs.storyteller_host,
Some(&creds),
request,
).await;
let job_id = match result {
Ok(enqueued) => {
// TODO(bt,2025-07-05): Enqueue job token?
info!("Successfully enqueued Artcraft gemini 2.5 flash inpaint. Job token: {}",
enqueued.inference_job_token);
enqueued.inference_job_token
}
Err(err) => {
error!("Failed to use Artcraft gemini 2.5 flash inpaint: {:?}", err);
return Err(InternalImageInpaintError::StorytellerError(err));
}
};
Ok(TaskEnqueueSuccess {
provider: GenerationProvider::Artcraft,
model: Some(GenerationModel::Gemini25Flash),
provider_job_id: Some(job_id.to_string()),
task_type: TaskType::ImageInpaintEdit,
})
}
// async fn get_mask(
// request: &EnqueueInpaintImageCommand,
// app_env_configs: &AppEnvConfigs,
// storyteller_creds: &StorytellerCredentialSet,
// ) -> Result<MediaFileToken, InternalImageInpaintError> {
//
// if request.mask_image_media_token.is_some() && request.mask_image_raw_bytes.is_some() {
// return Err(InternalImageInpaintError::MaskMediaTokenAndBytesSupplied);
// }
//
// if let Some(token) = request.mask_image_media_token.as_ref() {
// return Ok(token.clone());
// };
//
// let image_bytes = request.mask_image_raw_bytes.as_ref()
// .ok_or(InternalImageInpaintError::NoMaskImageSpecified)?;
//
// let image_bytes = normalize_image_bytes_to_flux_mask(image_bytes)
// .map_err(|err| {
// error!("Failed to convert image bytes to png: {:?}", err);
// InternalImageInpaintError::CouldNotEncodeMask
// })?;
//
// info!("Uploading image media file from bytes...");
//
// let result = upload_image_media_file_from_bytes(UploadImageBytesArgs {
// api_host: &app_env_configs.storyteller_host,
// maybe_creds: Some(&storyteller_creds),
// image_bytes: image_bytes.0,
// image_type: ImageType::Png,
// is_intermediate_system_file: true,
// }).await
// .map_err(|err| {
// error!("Failed to upload image media file: {:?}", err);
// InternalImageInpaintError::StorytellerError(err)
// })?;
//
// Ok(result.media_file_token)
// }
@@ -0,0 +1,2 @@
pub (super) mod handle_gemini_25_flash_inpaint;
pub (super) mod handle_gemini_25_flash_inpaint_artcraft;
@@ -1,5 +1,6 @@
mod flux_dev_juggernaut_inpaint;
mod flux_pro_1_inpaint;
mod flux_pro_kontext_inpaint;
mod gemini_25_flash_inpaint;
pub mod enqueue_image_inpaint_command;
pub mod errors;
@@ -74,6 +74,7 @@ export const IMAGE_MODELS: ImageModel[] = [
selectorBadges: ["35 sec."],
maxGenerationCount: 4,
defaultGenerationCount: 1,
canEditImages: true,
tags: [ModelTag.InstructiveEdit],
}),
new ImageModel({