generation count update

still need to roll this out to everything, but this lands standardized
aspect ratio for several models

.env

@@ -1,6 +1,7 @@
 SERVER_ENVIRONMENT=development
 
 DATABASE_URL=mysql://storyteller:password@localhost/storyteller
+#DATABASE_URL=sqlite:/tmp/tasks.sqlite
 #DATABASE_URL=mysql://root:root@localhost/storyteller
 #DATABASE_URL=mysql://root@localhost/storyteller
 

.github/workflows/artcraft-macos-publish.yml

@@ -200,10 +200,10 @@ jobs:
 
           # Use npx to ensure we use the local nx version
           echo "Building with npx nx..."
-          npx nx run editor3d:build
+          npx nx run artcraft:build
 
           echo "Checking build output..."
-          ls -la apps/editor3d/dist/ | head -10
+          ls -la apps/artcraft/dist/ | head -10
 
           # Update Tauri config to skip the build (since we already built)
           cd ${{ github.workspace }}/crates/desktop/artcraft

.github/workflows/artcraft-windows-publish.yml

@@ -75,10 +75,10 @@ jobs:
 
           # Use npx to ensure we use the local nx version
           echo "Building with npx nx..."
-          npx nx run editor3d:build
+          npx nx run artcraft:build
 
           echo "Checking build output..."
-          dir apps/editor3d/dist/ | Select-Object -First 10
+          dir apps/artcraft/dist/ | Select-Object -First 10
 
           # Update Tauri config to skip the build (since we already built)
           cd ${{ github.workspace }}/crates/desktop/artcraft
@@ -104,4 +104,4 @@ jobs:
           releaseBody: "Windows build - See the assets to download this version and install."
           releaseDraft: true
           prerelease: false
-          args: ${{ matrix.args }}
+          args: ${{ matrix.args }}

.gitignore

@@ -45,6 +45,9 @@ crates/lib/sqlite_queries/target/
 # Javascript build directories
 node_modules/
 
+# Frontend build outputs
+frontend/apps/artcraft/bundle-report.html
+
 # Vim noise
 *.swo
 *.swp

.sqlx/query-13d435d8f681b73e4080748dcef0579e246e41049af67e1bd79ffc06aa6ba974.json

@@ -0,0 +1,12 @@
+{
+  "db_name": "SQLite",
+  "query": "\n    UPDATE tasks\n    SET is_dismissed_by_user = 1\n    WHERE id = ?\n  ",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Right": 1
+    },
+    "nullable": []
+  },
+  "hash": "13d435d8f681b73e4080748dcef0579e246e41049af67e1bd79ffc06aa6ba974"
+}

.sqlx/query-1e3758851fc8451670f86a9432009b73bdb81de4a08dcaa96a86afb5f9be7a68.json

@@ -0,0 +1,12 @@
+{
+  "db_name": "SQLite",
+  "query": "\n    UPDATE tasks\n    SET\n      task_status = ?,\n      on_complete_batch_token = ?,\n      on_complete_primary_media_file_token = ?,\n      on_complete_primary_media_file_class = ?,\n      on_complete_primary_media_file_cdn_url = ?,\n      on_complete_primary_media_file_thumbnail_url_template = ?\n    WHERE\n      provider = ?\n      AND provider_job_id = ?\n  ",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Right": 8
+    },
+    "nullable": []
+  },
+  "hash": "1e3758851fc8451670f86a9432009b73bdb81de4a08dcaa96a86afb5f9be7a68"
+}

.sqlx/query-38ff2df237b16f077865711dd81261de1b5fa05a9787679ce9ccd4ac576b1069.json

@@ -0,0 +1,68 @@
+{
+  "db_name": "SQLite",
+  "query": "\n    SELECT\n      id,\n      task_status,\n      task_type,\n      model_type,\n      provider,\n      provider_job_id,\n      frontend_caller,\n      frontend_subscriber_id,\n      frontend_subscriber_payload\n    FROM tasks\n    WHERE\n      provider = ?\n      AND provider_job_id = ?\n  ",
+  "describe": {
+    "columns": [
+      {
+        "name": "id",
+        "ordinal": 0,
+        "type_info": "Text"
+      },
+      {
+        "name": "task_status",
+        "ordinal": 1,
+        "type_info": "Text"
+      },
+      {
+        "name": "task_type",
+        "ordinal": 2,
+        "type_info": "Text"
+      },
+      {
+        "name": "model_type",
+        "ordinal": 3,
+        "type_info": "Text"
+      },
+      {
+        "name": "provider",
+        "ordinal": 4,
+        "type_info": "Text"
+      },
+      {
+        "name": "provider_job_id",
+        "ordinal": 5,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_caller",
+        "ordinal": 6,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_subscriber_id",
+        "ordinal": 7,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_subscriber_payload",
+        "ordinal": 8,
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Right": 2
+    },
+    "nullable": [
+      false,
+      false,
+      false,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true
+    ]
+  },
+  "hash": "38ff2df237b16f077865711dd81261de1b5fa05a9787679ce9ccd4ac576b1069"
+}

.sqlx/query-47f94c127abc20683b25c257aed3c2190b9bd29eb829d61392f08e10fd986767.json

@@ -0,0 +1,116 @@
+{
+  "db_name": "SQLite",
+  "query": "\n    SELECT\n      id,\n      task_status,\n      task_type,\n      model_type,\n      provider,\n      provider_job_id,\n      frontend_caller,\n      frontend_subscriber_id,\n      frontend_subscriber_payload,\n      on_complete_primary_media_file_token,\n      on_complete_primary_media_file_class,\n      on_complete_batch_token,\n      on_complete_primary_media_file_cdn_url,\n      on_complete_primary_media_file_thumbnail_url_template,\n      created_at as \"created_at: DateTime<Utc>\",\n      updated_at as \"updated_at: DateTime<Utc>\",\n      completed_at as \"completed_at: DateTime<Utc>\"\n    FROM tasks\n    WHERE is_dismissed_by_user == 0\n  ",
+  "describe": {
+    "columns": [
+      {
+        "name": "id",
+        "ordinal": 0,
+        "type_info": "Text"
+      },
+      {
+        "name": "task_status",
+        "ordinal": 1,
+        "type_info": "Text"
+      },
+      {
+        "name": "task_type",
+        "ordinal": 2,
+        "type_info": "Text"
+      },
+      {
+        "name": "model_type",
+        "ordinal": 3,
+        "type_info": "Text"
+      },
+      {
+        "name": "provider",
+        "ordinal": 4,
+        "type_info": "Text"
+      },
+      {
+        "name": "provider_job_id",
+        "ordinal": 5,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_caller",
+        "ordinal": 6,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_subscriber_id",
+        "ordinal": 7,
+        "type_info": "Text"
+      },
+      {
+        "name": "frontend_subscriber_payload",
+        "ordinal": 8,
+        "type_info": "Text"
+      },
+      {
+        "name": "on_complete_primary_media_file_token",
+        "ordinal": 9,
+        "type_info": "Text"
+      },
+      {
+        "name": "on_complete_primary_media_file_class",
+        "ordinal": 10,
+        "type_info": "Text"
+      },
+      {
+        "name": "on_complete_batch_token",
+        "ordinal": 11,
+        "type_info": "Text"
+      },
+      {
+        "name": "on_complete_primary_media_file_cdn_url",
+        "ordinal": 12,
+        "type_info": "Text"
+      },
+      {
+        "name": "on_complete_primary_media_file_thumbnail_url_template",
+        "ordinal": 13,
+        "type_info": "Text"
+      },
+      {
+        "name": "created_at: DateTime<Utc>",
+        "ordinal": 14,
+        "type_info": "Int64"
+      },
+      {
+        "name": "updated_at: DateTime<Utc>",
+        "ordinal": 15,
+        "type_info": "Int64"
+      },
+      {
+        "name": "completed_at: DateTime<Utc>",
+        "ordinal": 16,
+        "type_info": "Int64"
+      }
+    ],
+    "parameters": {
+      "Right": 0
+    },
+    "nullable": [
+      false,
+      false,
+      false,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      true,
+      false,
+      false,
+      true
+    ]
+  },
+  "hash": "47f94c127abc20683b25c257aed3c2190b9bd29eb829d61392f08e10fd986767"
+}

.sqlx/query-50a2449b87abed7df3302a74ecd0e8ab87940077a247636680cb09b2fe488321.json

@@ -0,0 +1,12 @@
+{
+  "db_name": "SQLite",
+  "query": "\n    UPDATE tasks\n    SET\n      task_status = ?,\n      on_complete_batch_token = ?,\n      on_complete_primary_media_file_token = ?,\n      on_complete_primary_media_file_class = ?,\n      on_complete_primary_media_file_cdn_url = ?,\n      on_complete_primary_media_file_thumbnail_url_template = ?\n    WHERE id = ?\n  ",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Right": 7
+    },
+    "nullable": []
+  },
+  "hash": "50a2449b87abed7df3302a74ecd0e8ab87940077a247636680cb09b2fe488321"
+}

Cargo.toml

@@ -6,17 +6,19 @@ resolver = "2"
 members = [
   # API clients for external services
   # These might be custom implementations or wrappers over other client libraries.
-  "crates/api_clients/artcraft_api_defs", # TODO: api_clients/{artcraft}/{api_defs,client}
+  "crates/api_clients/artcraft/artcraft_api_defs", # TODO: api_clients/{artcraft}/{api_defs,client}
+  "crates/api_clients/artcraft/storyteller_client",
   "crates/api_clients/cloud_storage",
   "crates/api_clients/fal_client",
   "crates/api_clients/google_drive_common",
   "crates/api_clients/google_sign_in",
+  "crates/api_clients/grok_client",
   "crates/api_clients/midjourney_client",
   "crates/api_clients/openai_sora_client",
   "crates/api_clients/openart_client",
   "crates/api_clients/runwayml_client",
-  "crates/api_clients/storyteller_client",
   "crates/api_clients/thumbnail_generator",
+  "crates/api_clients/world_labs_client",
 
   # CLI tools for operations, etc.
   "crates/cli/devenv/dev-database-seed",
@@ -43,6 +45,7 @@ members = [
   "crates/lib/caching/memory_store",
   "crates/lib/caching/redis_caching",
   "crates/lib/cloudflare_errors",
+  "crates/lib/cloudflare_mitigation",
   "crates/lib/concurrency",
   "crates/lib/cookie_store",
   "crates/lib/cookies",
@@ -65,6 +68,7 @@ members = [
   "crates/lib/files/media",
   "crates/lib/files/mimetypes",
   "crates/lib/files/videos",
+  "crates/lib/http_headers",
   "crates/lib/idempotency",
   "crates/lib/image_algorithms",
   "crates/lib/jobs_common",
@@ -84,6 +88,7 @@ members = [
   "crates/lib/testing",
   "crates/lib/tts_common",
   "crates/lib/url_config",
+  "crates/lib/url_utils",
   "crates/lib/user_input_common",
   "crates/lib/web_scrapers",
   #"crates/lib/api_clients/fakeyou_client",
@@ -181,10 +186,11 @@ workspace-build-acceleration = { path = "crates/workspace-build-acceleration" }
 # Internal libs
 # Setting paths here ensures we don't need to do so in every crate's Cargo.toml
 # It makes refactoring much easier.
-artcraft_api_defs = { path = "crates/api_clients/artcraft_api_defs" }
+artcraft_api_defs = { path = "crates/api_clients/artcraft/artcraft_api_defs" }
 build_metadata = { path = "crates/lib/build_metadata" }
 cloud_storage = { path = "crates/api_clients/cloud_storage" }
 cloudflare_errors = { path = "crates/lib/cloudflare_errors" }
+cloudflare_mitigation = { path = "crates/lib/cloudflare_mitigation" }
 component_traits = { path = "crates/service/plugins/component_traits" }
 cookie_store = { path = "crates/lib/cookie_store" }
 crockford = { path = "crates/lib/data/crockford" }
@@ -194,6 +200,8 @@ errors = { path = "crates/lib/errors" }
 fal_client = { path = "crates/api_clients/fal_client" }
 file_to_base64_url = { path = "crates/lib/files/file_to_base64_url" }
 filesys = { path = "crates/lib/files/filesys" }
+grok_client = { path = "crates/api_clients/grok_client" }
+http_headers = { path = "crates/lib/http_headers" }
 idempotency = { path = "crates/lib/idempotency" }
 images = { path = "crates/lib/files/images" }
 jwt_light = { path = "crates/lib/jwt_light" }
@@ -205,12 +213,14 @@ openai_sora_client = { path = "crates/api_clients/openai_sora_client" }
 primitives = { path = "crates/lib/data/primitives" }
 shared_service_components = { path = "crates/lib/shared_service_components" }
 sqlite_tasks = { path = "crates/schema/database/sqlite_tasks" }
-storyteller_client = { path = "crates/api_clients/storyteller_client" }
+storyteller_client = { path = "crates/api_clients/artcraft/storyteller_client" }
 subprocess_common = { path = "crates/lib/subprocess_common" }
 test_data = { path = "crates/lib/test_data" }
 testing = { path = "crates/lib/testing" }
 tokens = { path = "crates/schema/public/tokens" }
+url_utils = { path = "crates/lib/url_utils" }
 web_base64 = { path = "crates/lib/data/web_base64" }
+world_labs_client = { path = "crates/api_clients/world_labs_client" }
 
 # External libs
 # Centrally pinned versions so that we can uniformly upgrade/downgrade versions across the workspace
@@ -227,22 +237,26 @@ actix-web-actors = "4.3.1+deprecated"
 actix-web-lab = "0.20.2"
 anyhow = "1.0.98"
 base64 = "0.22.1"
+bstr = "1.12.1"
 bytes = "1.10.1"
 chrono = { version = "0.4.41", features = ["serde"] }
 cookie = "0.18.1"
+env = "1.0.1"
+env_logger = "0.11.8"
 futures = { version = "0.3.31" }
 hex-literal = "1.0.0"
 indexmap = "2.9.0"
 log = "0.4.27"
 once_cell = "1.21.3"
 rand = "0.9.1"
-regex = "1.11.1"
-reqwest = { version = "0.12.22", features = ["json", "cookies"] }
+regex = "1.12.2"
+reqwest = { version = "0.12.23", default-features = false, features = [ "charset", "http2", "system-proxy", "json", "cookies", "rustls-tls" ] }
 reqwest-websocket = {  version = "0.5.0" }
 serde = {  version = "1.0.219", features = ["derive"] }
 serde_derive = "1.0.219"
 serde_json = "1.0.140"
 serde_yaml = "0.9.34+deprecated"
+sha2 = "0.11.0-rc.2"
 sha3 = "0.10.8"
 tauri = { version = "2.8.3", features = ["unstable", "devtools"] } # NB: "devtools" prevents App Store distribution; "ubstable" is needed for window access
 tauri-plugin-dialog = "2.3.3"
@@ -255,3 +269,17 @@ thiserror = "2.0.12"
 tokio = { version = "1.45.1", features = ["macros"] }
 toml = "0.8.22"
 url = { version = "2.5.4", features = ["serde"] }
+uuid = "1.18.1"
+
+# NB: wreq gets past fingerprinting, but we have to deal with platform-specific build issues.
+#
+# Mac:
+#   brew install --formula cmake pkg-config llvm
+# Windows:
+#   choco install cmake -y
+#   choco install strawberryperl -y
+#   choco install pkgconfiglite -y
+#   choco install llvm -y
+#   choco install nasm -y
+wreq = { version = "6.0.0-rc.20", features = ["json", "multipart", "cookies", "ws", "stream"] }
+wreq-util = "3.0.0-rc.6"

_sql/actions_and_analytics/artcraft/artcraft_active_users.sql

@@ -0,0 +1,20 @@
+-- noinspection SqlDialectInspectionForFile
+-- noinspection SqlNoDataSourceInspectionForFile
+-- noinspection SqlResolveForFile
+
+select
+    u.display_name,
+    count(*) as sessions,
+    a.session_duration_seconds / 60 / 60 as hours,
+    a.measurement_count,
+    u.email_address,
+    a.os_platform,
+    a.os_version,
+    u.created_at,
+    u.maybe_source,
+    a.last_active_at
+from analytics_app_sessions as a
+left outer join users as u
+on a.user_token = u.token
+group by a.user_token
+order by hours desc;

_sql/artcraft_migrations/20250708000000_create_tasks_table.sql

@@ -9,6 +9,14 @@
 --   tasks_v1.sqlite - initial version
 --   tasks_v2.sqlite - added model_type (nullable)
 --   tasks_v3.sqlite - added frontend_caller, comments
+--   tasks_v4.sqlite - added is_dismissed_by_user
+--   tasks_v5.sqlite - added five fields:
+--      on_complete_batch_token,
+--      on_complete_primary_media_file_token,
+--      on_complete_primary_media_file_class,
+--      on_complete_primary_media_file_cdn_url,
+--      on_complete_primary_media_file_thumbnail_url_template,
+--
 
 CREATE TABLE tasks (
     -- Task auto-incrementing primary key.
@@ -50,6 +58,32 @@ CREATE TABLE tasks (
     -- will be re-emitted back to the frontend.
     frontend_subscriber_payload TEXT,
 
+    -- Whether the user has dismissed the task from view.
+    is_dismissed_by_user INTEGER NOT NULL DEFAULT 0,
+
+    -- OPTIONAL.
+    -- When the generation completes, the batch token (if any).
+    -- This will point to a collection of multiple files.
+    on_complete_batch_token TEXT,
+
+    -- OPTIONAL.
+    -- When the generation completes, the media token of the "primary" media file.
+    -- We'll probably always generate homogenous types, but just in case this is the first or primary item in a batch.
+    on_complete_primary_media_file_token TEXT,
+
+    -- OPTIONAL.
+    -- When the generation completes, the file type of the primary media file.
+    -- We'll probably always generate homogenous types, but just in case this is the first or primary item in a batch.
+    on_complete_primary_media_file_class TEXT,
+
+    -- OPTIONAL.
+    -- The URL to the media file.
+    on_complete_primary_media_file_cdn_url TEXT,
+
+    -- OPTIONAL.
+    -- The URL template for the thumbnail of the primary media file.
+    on_complete_primary_media_file_thumbnail_url_template TEXT,
+
     created_at INTEGER NOT NULL DEFAULT (unixepoch('now')),
     updated_at INTEGER NOT NULL DEFAULT (unixepoch('now')),
     completed_at INTEGER DEFAULT NULL

_sql/migrations_squashed/media_files.sql

@@ -152,6 +152,7 @@ CREATE TABLE media_files (
   --   * `vmd`
   --   * `pmx`
   --   * `csv`
+  --   * `spz` Gaussian splats
   media_type VARCHAR(16) NOT NULL,
 
   -- Broad class of media:

crates/api_clients/artcraft/artcraft_api_defs/src/common/responses/mod.rs

@@ -3,4 +3,5 @@ pub mod job_details;
 pub mod media_file_cover_image_details;
 pub mod media_links;
 pub mod simple_entity_stats;
+pub mod simple_generic_json_success;
 pub mod user_details_light;

crates/api_clients/artcraft/artcraft_api_defs/src/common/responses/simple_generic_json_success.rs

@@ -0,0 +1,7 @@
+use serde_derive::{Deserialize, Serialize};
+use utoipa::ToSchema;
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct SimpleGenericJsonSuccess {
+  pub success: bool,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/bg_removal/mod.rs

@@ -0,0 +1 @@
+pub mod remove_image_background;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/edit/mod.rs

@@ -1,6 +1,6 @@
 pub mod flux_pro_kontext_max_edit_image;
-pub mod gpt_image_1_edit_image;
 pub mod gemini_25_flash_edit_image;
+pub mod gpt_image_1_edit_image;
 pub mod qwen_edit_image;
 pub mod seededit_3_edit_image;
 

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/mod.rs

@@ -0,0 +1,5 @@
+pub mod bg_removal;
+pub mod edit;
+pub mod inpaint;
+pub mod multi_function;
+pub mod text;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/bytedance_seedream_v4_multi_function_image_gen.rs

@@ -0,0 +1,69 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const BYTEDANCE_SEEDREAM_V4_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/bytedance_seedream_4";
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct BytedanceSeedreamV4MultiFunctionImageGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  /// Image media tokens to include in the editing context.
+  /// If present, we're doing image editing (image-to-image / image-editing)
+  /// If absent, we're doing image generation (text-to-image)
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<BytedanceSeedreamV4MultiFunctionImageGenNumImages>,
+
+  pub max_images: Option<BytedanceSeedreamV4MultiFunctionImageGenMaxImages>,
+
+  pub image_size: Option<BytedanceSeedreamV4MultiFunctionImageGenImageSize>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4MultiFunctionImageGenNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4MultiFunctionImageGenMaxImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4MultiFunctionImageGenImageSize {
+  // Square
+  Square,
+  SquareHd,
+  // Tall
+  PortraitFourThree,
+  PortraitSixteenNine,
+  // Wide
+  LandscapeFourThree,
+  LandscapeSixteenNine,
+  // Auto
+  Auto,
+  Auto2k,
+  Auto4k,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct BytedanceSeedreamV4MultiFunctionImageGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/bytedance_seedream_v4p5_multi_function_image_gen.rs

@@ -0,0 +1,68 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const BYTEDANCE_SEEDREAM_V4P5_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/bytedance_seedream_4p5";
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct BytedanceSeedreamV4p5MultiFunctionImageGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  /// Image media tokens to include in the editing context.
+  /// If present, we're doing image editing (image-to-image / image-editing)
+  /// If absent, we're doing image generation (text-to-image)
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<BytedanceSeedreamV4p5MultiFunctionImageGenNumImages>,
+
+  pub max_images: Option<BytedanceSeedreamV4p5MultiFunctionImageGenMaxImages>,
+
+  pub image_size: Option<BytedanceSeedreamV4p5MultiFunctionImageGenImageSize>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4p5MultiFunctionImageGenNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4p5MultiFunctionImageGenMaxImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum BytedanceSeedreamV4p5MultiFunctionImageGenImageSize {
+  // Square
+  Square,
+  SquareHd,
+  // Tall
+  PortraitFourThree,
+  PortraitSixteenNine,
+  // Wide
+  LandscapeFourThree,
+  LandscapeSixteenNine,
+  // Auto
+  Auto2k,
+  Auto4k,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct BytedanceSeedreamV4p5MultiFunctionImageGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/gpt_image_1p5_multi_function_image_gen.rs

@@ -0,0 +1,95 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const GPT_IMAGE_1P5_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/gpt_image_1p5";
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct GptImage1p5MultiFunctionImageGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  /// Image media tokens to include in the editing context (**OR INPAINTING CONTEXT!**).
+  /// If present, we're doing image editing (image-to-image / image-editing / image-inpainting)
+  /// If absent, we're doing image generation (text-to-image)
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+  
+  /// Only for image editing, which turns this into an inpainting task.
+  /// Text to image should not set this.
+  pub mask_image_token: Option<MediaFileToken>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<GptImage1p5MultiFunctionImageGenNumImages>,
+  
+  pub image_size: Option<GptImage1p5MultiFunctionImageGenSize>,
+
+  pub background: Option<GptImage1p5MultiFunctionImageGenBackground>,
+
+  pub quality: Option<GptImage1p5MultiFunctionImageGenQuality>,
+
+  /// Only for image editing.
+  pub input_fidelity: Option<GptImage1p5MultiFunctionImageGenInputFidelity>,
+
+  pub output_format: Option<GptImage1p5MultiFunctionImageGenOutputFormat>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenSize {
+  /// 1024x1024
+  Square,
+  /// 1536x1024
+  Wide,
+  /// 1024x1536
+  Tall,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenBackground {
+  Auto,
+  Transparent,
+  Opaque,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenQuality {
+  Low,
+  Medium,
+  High,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenInputFidelity {
+  Low,
+  High,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GptImage1p5MultiFunctionImageGenOutputFormat {
+  Jpeg,
+  Png,
+  Webp,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct GptImage1p5MultiFunctionImageGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/mod.rs

@@ -0,0 +1,5 @@
+pub mod bytedance_seedream_v4_multi_function_image_gen;
+pub mod bytedance_seedream_v4p5_multi_function_image_gen;
+pub mod gpt_image_1p5_multi_function_image_gen;
+pub mod nano_banana_multi_function_image_gen;
+pub mod nano_banana_pro_multi_function_image_gen;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/nano_banana_multi_function_image_gen.rs

@@ -0,0 +1,62 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const NANO_BANANA_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/nano_banana";
+
+#[derive(Serialize, Deserialize, ToSchema, Debug)]
+pub struct NanoBananaMultiFunctionImageGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  /// Image media tokens to include in the editing context.
+  /// If present, we're doing image editing (image-to-image / image-editing)
+  /// If absent, we're doing image generation (text-to-image)
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<NanoBananaMultiFunctionImageGenNumImages>,
+
+  /// Aspect ratio of the images to generate. Default is "1:1"
+  pub aspect_ratio: Option<NanoBananaMultiFunctionImageGenAspectRatio>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum NanoBananaMultiFunctionImageGenNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum NanoBananaMultiFunctionImageGenAspectRatio {
+  // Auto (only for image editing)
+  Auto, // Default for image editing.
+  // Square
+  OneByOne, // Default for text-to-image
+  // Wide
+  FiveByFour,
+  FourByThree,
+  ThreeByTwo,
+  SixteenByNine,
+  TwentyOneByNine,
+  // Tall
+  FourByFive,
+  ThreeByFour,
+  TwoByThree,
+  NineBySixteen, // NB: No NineByTwentyOne ?
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct NanoBananaMultiFunctionImageGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/multi_function/nano_banana_pro_multi_function_image_gen.rs

@@ -0,0 +1,73 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const NANO_BANANA_PRO_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/nano_banana_pro";
+
+#[derive(Serialize, Deserialize, ToSchema, Debug)]
+pub struct NanoBananaProMultiFunctionImageGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  /// Image media tokens to include in the editing context.
+  /// If present, we're doing image editing (image-to-image / image-editing)
+  /// If absent, we're doing image generation (text-to-image)
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<NanoBananaProMultiFunctionImageGenNumImages>,
+
+  /// Resolution of the image to generate. Default is OneK (1K).
+  pub resolution: Option<NanoBananaProMultiFunctionImageGenImageResolution>,
+
+  /// Aspect ratio of the images to generate. Default is "1:1"
+  pub aspect_ratio: Option<NanoBananaProMultiFunctionImageGenAspectRatio>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum NanoBananaProMultiFunctionImageGenNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum NanoBananaProMultiFunctionImageGenImageResolution {
+  OneK,
+  TwoK,
+  FourK,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum NanoBananaProMultiFunctionImageGenAspectRatio {
+  // Auto (only for image editing)
+  Auto, // Default for image editing.
+  // Square
+  OneByOne, // Default for text-to-image
+  // Wide
+  FiveByFour,
+  FourByThree,
+  ThreeByTwo,
+  SixteenByNine,
+  TwentyOneByNine,
+  // Tall
+  FourByFive,
+  ThreeByFour,
+  TwoByThree,
+  NineBySixteen, // NB: No NineByTwentyOne ?
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct NanoBananaProMultiFunctionImageGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/text/gemini_25_flash_text_to_image.rs

@@ -0,0 +1,50 @@
+use serde_derive::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const GEMINI_25_FLASH_EDIT_IMAGE_PATH: &str = "/v1/generate/image/gemini_25_flash";
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Gemini25FlashTextToImageImageRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// Text prompt to generate the image from.
+  pub prompt: Option<String>,
+
+  // TODO: OneOf for image_urls or image_media_tokens
+  /// Image media tokens to include in the editing context.
+  pub image_media_tokens: Option<Vec<MediaFileToken>>,
+
+  /// Number of images to generate. Default is one.
+  pub num_images: Option<Gemini25FlashTextToImageImageNumImages>,
+
+  /// Quality of the image to generate. Default is High.
+  pub image_quality: Option<Gemini25FlashTextToImageImageImageQuality>,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Gemini25FlashTextToImageImageNumImages {
+  One, // Default
+  Two,
+  Three,
+  Four,
+}
+
+#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Gemini25FlashTextToImageImageImageQuality {
+  Auto,
+  Low,
+  Medium,
+  High, // Default
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Gemini25FlashTextToImageImageResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/text/generate_flux_pro_11_text_to_image.rs

@@ -19,6 +19,7 @@ pub struct GenerateFluxPro11TextToImageRequest {
   pub num_images: Option<GenerateFluxPro11TextToImageNumImages>,
 }
 
+// TODO(btw,2026-01-01): Fal.ai and Fal.rs client libraries seem to disagree about the options.
 #[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
 #[serde(rename_all = "snake_case")]
 pub enum GenerateFluxPro11TextToImageAspectRatio {

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/text/generate_flux_pro_11_ultra_text_to_image.rs

@@ -23,10 +23,14 @@ pub struct GenerateFluxPro11UltraTextToImageRequest {
 #[serde(rename_all = "snake_case")]
 pub enum GenerateFluxPro11UltraTextToImageAspectRatio {
   Square, // 1:1
+  LandscapeThreeByTwo, // 3:2
   LandscapeFourByThree, // 4:3
   LandscapeSixteenByNine, // 16:9
+  LandscapeTwentyOneByNine, // 21:9
+  PortraitTwoByThree, // 2:3
   PortraitThreeByFour, // 3:4
   PortraitNineBySixteen, // 9:16
+  PortraitNineByTwentyOne, // 9:21
   //Custom { width: u32, height: u32 }, // TODO
 }
 

crates/api_clients/artcraft/artcraft_api_defs/src/generate/image/text/mod.rs

@@ -1,8 +1,6 @@
-pub mod edit;
+pub mod gemini_25_flash_text_to_image;
 pub mod generate_flux_1_dev_text_to_image;
 pub mod generate_flux_1_schnell_text_to_image;
 pub mod generate_flux_pro_11_text_to_image;
 pub mod generate_flux_pro_11_ultra_text_to_image;
 pub mod generate_gpt_image_1_text_to_image;
-pub mod inpaint;
-pub mod remove_image_background;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/generate_seedance_1_0_lite_image_to_video.rs

@@ -21,11 +21,14 @@ pub struct GenerateSeedance10LiteImageToVideoRequest {
   /// Optional text prompt.
   pub prompt: Option<String>,
   
-  /// Optional: aspect ratio of the generated video.
+  /// Optional: resolution of the generated video.
   pub resolution: Option<GenerateSeedance10LiteResolution>,
 
-  /// Optional: aspect ratio of the generated video.
+  /// Optional: duration of the generated video.
   pub duration: Option<GenerateSeedance10LiteDuration>,
+
+  /// Optional: aspect ratio of the generated video.
+  pub aspect_ratio: Option<GenerateSeedance10LiteAspectRatio>,
 }
 
 #[derive(Serialize, Deserialize, ToSchema)]
@@ -42,6 +45,19 @@ pub enum GenerateSeedance10LiteDuration {
   TenSeconds,
 }
 
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum GenerateSeedance10LiteAspectRatio {
+  Auto,
+  TwentyOneByNine,
+  SixteenByNine,
+  FourByThree,
+  Square,
+  ThreeByFour,
+  NineBySixteen,
+}
+
 #[derive(Serialize, Deserialize, ToSchema)]
 pub struct GenerateSeedance10LiteImageToVideoResponse {
   pub success: bool,

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/mod.rs

@@ -5,4 +5,5 @@ pub mod generate_seedance_1_0_lite_image_to_video;
 pub mod generate_seedance_1_0_pro_image_to_video;
 pub mod generate_veo_2_image_to_video;
 pub mod generate_veo_3_fast_image_to_video;
-pub mod generate_veo_3_image_to_video;
+pub mod generate_veo_3_image_to_video;
+pub mod multi_function;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/kling_2_5_turbo_multi_function_video_gen.rs

@@ -0,0 +1,64 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const KLING_2P5_TURBO_PRO_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/kling_2p5_turbo_pro";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Kling2p5TurboProMultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for text-to-video.
+  pub negative_prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub start_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the ending frame (if desired)
+  pub end_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional
+  /// Duration of the video
+  pub duration: Option<Kling2p5TurboProMultiFunctionVideoGenDuration>,
+
+  /// Optional.
+  /// Only for text-to-video
+  pub aspect_ratio: Option<Kling2p5TurboProMultiFunctionVideoGenAspectRatio>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Kling2p5TurboProMultiFunctionVideoGenDuration {
+  FiveSeconds,
+  TenSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Kling2p5TurboProMultiFunctionVideoGenAspectRatio {
+  Square,
+  SixteenByNine,
+  NineBySixteen,
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Kling2p5TurboProMultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/kling_2_6_multi_function_video_gen.rs

@@ -0,0 +1,68 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const KLING_2P6_PRO_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/kling_2p6_pro";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Kling2p6ProMultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for text-to-video.
+  pub negative_prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub start_frame_image_media_token: Option<MediaFileToken>,
+
+  // /// Optional.
+  // /// Only for image-to-video
+  // /// Source of the ending frame (if desired)
+  // pub end_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional
+  /// Duration of the video
+  pub duration: Option<Kling2p6ProMultiFunctionVideoGenDuration>,
+
+  /// Optional
+  /// Whether to generate audio.
+  pub generate_audio: Option<bool>,
+
+  /// Optional.
+  /// Only for text-to-video
+  pub aspect_ratio: Option<Kling2p6ProMultiFunctionVideoGenAspectRatio>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Kling2p6ProMultiFunctionVideoGenDuration {
+  FiveSeconds,
+  TenSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Kling2p6ProMultiFunctionVideoGenAspectRatio {
+  Square,
+  SixteenByNine,
+  NineBySixteen,
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Kling2p6ProMultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/mod.rs

@@ -0,0 +1,6 @@
+pub mod kling_2_5_turbo_multi_function_video_gen;
+pub mod kling_2_6_multi_function_video_gen;
+pub mod sora_2_multi_function_video_gen;
+pub mod sora_2_pro_multi_function_video_gen;
+pub mod veo_3p1_fast_multi_function_video_gen;
+pub mod veo_3p1_multi_function_video_gen;

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/sora_2_multi_function_video_gen.rs

@@ -0,0 +1,66 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const SORA_2_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/sora_2";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Sora2MultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub image_media_token: Option<MediaFileToken>,
+
+  /// Optional.
+  pub resolution: Option<Sora2MultiFunctionVideoGenResolution>,
+
+  /// Optional
+  /// Duration of the video
+  /// (this is uniform against all modes)
+  pub duration: Option<Sora2MultiFunctionVideoGenDuration>,
+
+  /// Optional.
+  pub aspect_ratio: Option<Sora2MultiFunctionVideoGenAspectRatio>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2MultiFunctionVideoGenResolution {
+  Auto,
+  SevenTwentyP,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2MultiFunctionVideoGenDuration {
+  FourSeconds,
+  EightSeconds,
+  TwelveSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2MultiFunctionVideoGenAspectRatio {
+  Auto,
+  NineBySixteen,
+  SixteenByNine,
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Sora2MultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/sora_2_pro_multi_function_video_gen.rs

@@ -0,0 +1,67 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const SORA_2_PRO_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/sora_2_pro";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Sora2ProMultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub image_media_token: Option<MediaFileToken>,
+
+  /// Optional.
+  pub resolution: Option<Sora2ProMultiFunctionVideoGenResolution>,
+
+  /// Optional
+  /// Duration of the video
+  /// (this is uniform against all modes)
+  pub duration: Option<Sora2ProMultiFunctionVideoGenDuration>,
+
+  /// Optional.
+  pub aspect_ratio: Option<Sora2ProMultiFunctionVideoGenAspectRatio>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2ProMultiFunctionVideoGenResolution {
+  Auto,
+  SevenTwentyP,
+  TenEightyP,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2ProMultiFunctionVideoGenDuration {
+  FourSeconds,
+  EightSeconds,
+  TwelveSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Sora2ProMultiFunctionVideoGenAspectRatio {
+  Auto,
+  NineBySixteen,
+  SixteenByNine,
+}
+
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Sora2ProMultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/veo_3p1_fast_multi_function_video_gen.rs

@@ -0,0 +1,90 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const VEO_3P1_FAST_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/veo_3p1_fast";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Veo3p1FastMultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for text-to-video.
+  pub negative_prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub start_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the ending frame (if desired)
+  pub end_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional
+  /// Duration of the video
+  /// (this is uniform against all modes)
+  pub duration: Option<Veo3p1FastMultiFunctionVideoGenDuration>,
+
+  /// Optional.
+  pub aspect_ratio: Option<Veo3p1FastMultiFunctionVideoGenAspectRatio>,
+  
+  /// Optional.
+  pub resolution: Option<Veo3p1FastMultiFunctionVideoGenResolution>,
+
+  /// Optional
+  /// Whether to generate audio.
+  pub generate_audio: Option<bool>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub enhance_prompt: Option<bool>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub seed: Option<i64>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub auto_fix: Option<bool>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1FastMultiFunctionVideoGenDuration {
+  FourSeconds,
+  SixSeconds,
+  EightSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1FastMultiFunctionVideoGenAspectRatio {
+  Auto,
+  SixteenByNine,
+  NineBySixteen,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1FastMultiFunctionVideoGenResolution {
+  SevenTwentyP,
+  TenEightyP,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Veo3p1FastMultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/generate/video/multi_function/veo_3p1_multi_function_video_gen.rs

@@ -0,0 +1,90 @@
+use serde::{Deserialize, Serialize};
+use tokens::tokens::generic_inference_jobs::InferenceJobToken;
+use tokens::tokens::media_files::MediaFileToken;
+use utoipa::ToSchema;
+
+pub const VEO_3P1_MULTI_FUNCTION_VIDEO_VIDEO_PATH: &str = "/v1/generate/video/multi_function/veo_3p1";
+
+/// Both text-to-video and image-to-video in one request.
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Veo3p1MultiFunctionVideoGenRequest {
+  /// Idempotency token to prevent duplicate requests.
+  pub uuid_idempotency_token: String,
+
+  /// "Required".
+  /// Required downstream, but we'll coerce null to empty string.
+  /// Text prompt.
+  pub prompt: Option<String>,
+
+  /// Optional.
+  /// Only for text-to-video.
+  pub negative_prompt: Option<String>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the starting frame
+  /// If present, we're doing image-to-video
+  /// If absent, we're doing text-to-video
+  pub start_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional.
+  /// Only for image-to-video
+  /// Source of the ending frame (if desired)
+  pub end_frame_image_media_token: Option<MediaFileToken>,
+
+  /// Optional
+  /// Duration of the video
+  /// (this is uniform against all modes)
+  pub duration: Option<Veo3p1MultiFunctionVideoGenDuration>,
+
+  /// Optional.
+  pub aspect_ratio: Option<Veo3p1MultiFunctionVideoGenAspectRatio>,
+  
+  /// Optional.
+  pub resolution: Option<Veo3p1MultiFunctionVideoGenResolution>,
+
+  /// Optional
+  /// Whether to generate audio.
+  pub generate_audio: Option<bool>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub enhance_prompt: Option<bool>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub seed: Option<i64>,
+
+  /// Optional
+  /// Only for text-to-video
+  pub auto_fix: Option<bool>,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1MultiFunctionVideoGenDuration {
+  FourSeconds,
+  SixSeconds,
+  EightSeconds,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1MultiFunctionVideoGenAspectRatio {
+  Auto,
+  SixteenByNine,
+  NineBySixteen,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum Veo3p1MultiFunctionVideoGenResolution {
+  SevenTwentyP,
+  TenEightyP,
+}
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct Veo3p1MultiFunctionVideoGenResponse {
+  pub success: bool,
+  pub inference_job_token: InferenceJobToken,
+}

crates/api_clients/artcraft/artcraft_api_defs/src/lib.rs

@@ -6,4 +6,5 @@ pub mod jobs;
 pub mod media_file;
 pub mod prompts;
 pub mod stripe_artcraft;
-pub mod subscriptions;
+pub mod subscriptions;
+pub mod utils;

crates/api_clients/artcraft/artcraft_api_defs/src/media_file/delete_media_file.rs

@@ -0,0 +1,19 @@
+use serde_derive::{Deserialize, Serialize};
+use utoipa::ToSchema;
+use tokens::tokens::media_files::MediaFileToken;
+
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct DeleteMediaFileRequest {
+  pub set_delete: bool,
+  
+  /// NB: this is only to disambiguate when a user is both a mod and an author.
+  pub as_mod: Option<bool>,
+}
+
+/// For the URL PathInfo
+#[derive(Serialize, Deserialize, ToSchema)]
+pub struct DeleteMediaFilePathInfo {
+  pub token: MediaFileToken,
+}
+
+

crates/api_clients/artcraft/artcraft_api_defs/src/media_file/mod.rs

@@ -1 +1,2 @@
+pub mod delete_media_file;
 pub mod list_batch_generated_media_files;

crates/api_clients/artcraft/artcraft_api_defs/src/utils/media_links_to_thumbnail_template.rs

@@ -0,0 +1,12 @@
+use crate::common::responses::media_links::MediaLinks;
+
+/// Get the thumbnail template for either images or videos.
+pub fn media_links_to_thumbnail_template(links: &MediaLinks) -> Option<&str> {
+  // NB: We only populate video previews for video tasks.
+  if let Some(video) = links.maybe_video_previews.as_ref() {
+    return Some(&video.animated_thumbnail_template);
+  }
+
+  // Image case - use the image thumbnail template if it exists.
+  links.maybe_thumbnail_template.as_deref()
+}

crates/api_clients/artcraft/artcraft_api_defs/src/utils/mod.rs

@@ -0,0 +1 @@
+pub mod media_links_to_thumbnail_template;