first copied version of artcraftx

This commit is contained in:
Hanashi 2026-07-30 21:40:27 -04:00
parent e1c1609dd7
commit 23da2ae5e3
5422 changed files with 498102 additions and 0 deletions

19
.gitignore vendored
View File

@ -1 +1,20 @@
*~
external/
# Secrets are BAD!
*secret*sh
.env
.env-secrets
*-secrets.env
secrets/
# Rust build output
/target
# Editors / OS
.DS_Store
.idea/
.vscode/
# Python virtual envs (named by convention)
venv/

View File

@ -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"
}

View File

@ -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"
}

View File

@ -0,0 +1,128 @@
{
"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 on_failure_type,\n on_failure_message,\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": "on_failure_type",
"ordinal": 14,
"type_info": "Text"
},
{
"name": "on_failure_message",
"ordinal": 15,
"type_info": "Text"
},
{
"name": "created_at: DateTime<Utc>",
"ordinal": 16,
"type_info": "Int64"
},
{
"name": "updated_at: DateTime<Utc>",
"ordinal": 17,
"type_info": "Int64"
},
{
"name": "completed_at: DateTime<Utc>",
"ordinal": 18,
"type_info": "Int64"
}
],
"parameters": {
"Right": 0
},
"nullable": [
false,
false,
false,
true,
true,
true,
true,
true,
true,
true,
true,
true,
true,
true,
true,
true,
false,
false,
true
]
},
"hash": "2436eb7bf4b0e23a899311a67a77f71d89eb35461e2113e43cd9d221bfa91c9a"
}

View File

@ -0,0 +1,12 @@
{
"db_name": "SQLite",
"query": "\n UPDATE tasks\n SET task_status = \"cancelled_by_user\"\n WHERE task_status NOT IN (\n \"complete_success\",\n \"complete_failure\",\n \"dead\",\n \"cancelled_by_user\",\n \"cancelled_by_provider\",\n \"cancelled_by_us\"\n )\n ",
"describe": {
"columns": [],
"parameters": {
"Right": 0
},
"nullable": []
},
"hash": "30272128d83543d9fbfca930b7cdd4af30adfb4d90f2de044463a111c6340495"
}

View File

@ -0,0 +1,86 @@
{
"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 queue_status_url,\n queue_response_url,\n prompt_token,\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": "queue_status_url",
"ordinal": 7,
"type_info": "Text"
},
{
"name": "queue_response_url",
"ordinal": 8,
"type_info": "Text"
},
{
"name": "prompt_token",
"ordinal": 9,
"type_info": "Text"
},
{
"name": "frontend_subscriber_id",
"ordinal": 10,
"type_info": "Text"
},
{
"name": "frontend_subscriber_payload",
"ordinal": 11,
"type_info": "Text"
}
],
"parameters": {
"Right": 2
},
"nullable": [
false,
false,
false,
true,
true,
true,
true,
true,
true,
true,
true,
true
]
},
"hash": "35677894f7b02f32a1bf6396fa6d10caba69bb0f3f03ee7ffbfa1437689d42df"
}

View File

@ -0,0 +1,12 @@
{
"db_name": "SQLite",
"query": "\n INSERT INTO tasks (\n id,\n task_status,\n task_type,\n model_type,\n provider,\n provider_job_id,\n queue_status_url,\n queue_response_url,\n prompt_token,\n frontend_caller,\n frontend_subscriber_id,\n frontend_subscriber_payload\n )\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)\n ",
"describe": {
"columns": [],
"parameters": {
"Right": 12
},
"nullable": []
},
"hash": "4fc9674ae39abaadc36054c200290ec95d0e2c7765336f0fa97805a2265f5c4c"
}

View File

@ -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"
}

View File

@ -0,0 +1,12 @@
{
"db_name": "SQLite",
"query": "\n UPDATE tasks\n SET task_status = ?,\n on_failure_type = ?,\n on_failure_message = ?\n WHERE id = ?\n ",
"describe": {
"columns": [],
"parameters": {
"Right": 4
},
"nullable": []
},
"hash": "9b99ccd0baf9a4b74d8903f63e3a3dcf4c09b75cad978f0532388cbd743b7516"
}

View File

@ -0,0 +1,12 @@
{
"db_name": "SQLite",
"query": "\n UPDATE tasks\n SET task_status = ?\n WHERE id = ?\n ",
"describe": {
"columns": [],
"parameters": {
"Right": 2
},
"nullable": []
},
"hash": "dc0367e9e43c3d7dda3623e77337ab4529eac2192475be7d08589886703ba1cb"
}

View File

@ -0,0 +1,12 @@
{
"db_name": "SQLite",
"query": "\n UPDATE tasks\n SET is_dismissed_by_user = 1\n WHERE is_dismissed_by_user != 1\n ",
"describe": {
"columns": [],
"parameters": {
"Right": 0
},
"nullable": []
},
"hash": "f83018ec5493982659a56576296914f937bb4e75db3a522eb4bda486d5caa402"
}

74
CLAUDE.md Normal file
View File

@ -0,0 +1,74 @@
# CLAUDE.md
This file provides guidance for Claude Code when working with the ArtCraftX monorepo.
## Project Overview
ArtCraftX is a pared-down desktop application for generating AI image and video. It is written in
Rust (Tauri) and TypeScript (React) and contains desktop and frontend components.
## Project Structure
```
artcraftx/
├── crates/ # Rust workspace
│ ├── api_clients/ # HTTP clients for calling internal and 3rd party services
│ ├── desktop/ # Desktop (Tauri) apps
│ │ └── artcraftx/ # (Important) ArtCraftX, the desktop app
│ ├── lib/ # Various utility libraries
│ ├── schema/ # Data definition layer
│ │ ├── database/
│ │ │ └── sqlite_tasks/ # Queries for the desktop app's "tasks" SQLite database
│ │ └── public/ # Token identifier and enum variant definitions
│ │ ├── enums/ # Enum variants stored as strings
│ │ └── tokens/ # Identifiers with Stripe-like ID prefixes, eg. "user_{entropy}"
│ ├── testing/ # Test data and test utilities
│ └── vendor/ # Vendored third-party crates (eg. tauri-plugin-http)
├── frontend/ # Nx typescript monorepo for the Tauri desktop app frontend
│ ├── apps/
│ │ └── artcraft/ # ArtCraftX the Tauri app's frontend. Used with the `artcraftx` Rust crate.
│ └── libs/ # Support libraries, reusable React components, etc.
├── script/
│ └── artcraftx/ # Launch and build scripts for the desktop app
└── Cargo.toml # Rust monorepo workspace
```
## Code Style
- Rust with no minimum supported version
- A mix of wreq and reqwest for Rust HTTP clients
- SQLx for SQLite (the desktop app's local task database in `sqlite_tasks`)
- Never use `println!` or `eprintln!` outside of tests; use `log` crate macros instead
- When two crates export the same type name, alias imports with a suffix: `use foo::Bar as BarFoo;`
- Prefer `use` imports over inline fully-qualified paths; only qualify inline for true one-offs or std collisions
- TypeScript with Nx, React, Vite, Zustand, and Three.js
- Use two spaces for indentation
### File Layout
Organize for top-to-bottom reading. Important things first, details later.
- **Constants** at the top (after imports)
- **Structs/enums** next; outer structs above inner sub-structs
- **API types** in order: Request, Response, Error
- **In impl blocks**: constructors first, then public methods, then private helpers
- Private helpers go *below* the methods that call them
- Among helpers: meatier logic above leaf-level formatters
- **In test modules**: constants first, then test cases (grouped into sub-modules when 2+), then helper functions last
## Markdown
- **Tables must be space-padded so columns align in plain text.** Markdown
tables are read raw (terminals, diffs, editors) at least as often as
rendered, and condensed tables are unreadable there. Pad every cell to its
column width:
```markdown
| Model | Configuration | Credits | Speed | Score |
|--------------|---------------|--------------------|---------|
| Meshy 6 | text or image | 104 | 80.0 | +24 |
| Rodin 2.5 | text or image | 13 | 10.0 | +3 |
```
Not: `| Model | Configuration | Credits | Speed | Score |` packed tight
with varying widths per row.

9483
Cargo.lock generated Normal file

File diff suppressed because it is too large Load Diff

171
Cargo.toml Normal file
View File

@ -0,0 +1,171 @@
[workspace]
resolver = "2"
# This block defines the workspace members.
# When we add a library or binary crate, it needs an entry here.
members = [
# API clients for external services
# These might be custom implementations or wrappers over other client libraries.
"crates/api_clients/artcraft/artcraft_api_defs",
"crates/api_clients/artcraft/artcraft_client",
"crates/api_clients/artcraft/artcraft_router",
"crates/api_clients/fal_client",
"crates/api_clients/gmicloud_client",
"crates/api_clients/grok_api_client",
"crates/api_clients/grok_consumer_client",
"crates/api_clients/midjourney_client",
"crates/api_clients/openai_sora_client",
"crates/api_clients/seedance2pro_client",
"crates/api_clients/worldlabs_api_client",
"crates/api_clients/worldlabs_consumer_client",
# Desktop apps
"crates/desktop/artcraftx",
# Shared libraries
"crates/lib/build_metadata",
"crates/lib/caching/memory_store",
"crates/lib/cloudflare_errors",
"crates/lib/cloudflare_mitigation",
"crates/lib/cookie_store",
"crates/lib/data/primitives",
"crates/lib/data/web_base64",
"crates/lib/errors",
"crates/lib/files/file_to_base64_url",
"crates/lib/files/filesys",
"crates/lib/files/images",
"crates/lib/files/mimetypes",
"crates/lib/http_headers",
"crates/lib/jwt_light",
"crates/lib/url_utils",
"crates/lib/uuid_utils",
# Schema (database): local task storage
"crates/schema/database/sqlite_tasks",
# Schema (public): common strongly-typed tokens, ids, payloads, etc.
"crates/schema/public/enums",
"crates/schema/public/tokens",
# Testing code
"crates/testing/test_data",
"crates/testing/test_utils",
# Vendored code
"crates/vendor/tauri-plugin-http",
# cargo-hakari workspace hack (compilation speedup hack)
# NB: Currently disabled (empty crate); kept so member crates can depend on it.
"crates/workspace-build-acceleration",
]
[workspace.package]
# Common inheritable settings for monorepo packages
publish = false
# NB: Profiles must live at the workspace root; profiles in member crates are ignored.
[profile.dev]
incremental = true # Compile your binary in smaller steps.
[profile.release]
codegen-units = 1 # Allows LLVM to perform better optimization.
lto = true # Enables link-time-optimizations.
opt-level = 3 # `s` prioritizes small binary size. Use `3` if you prefer speed.
panic = "abort" # Higher performance by disabling panic handlers.
strip = true # Ensures debug symbols are removed.
[workspace.dependencies]
# Hakari workspace hack (faster compile times)
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/artcraft_api_defs" }
artcraft_client = { path = "crates/api_clients/artcraft/artcraft_client" }
artcraft_router = { path = "crates/api_clients/artcraft/artcraft_router" }
build_metadata = { path = "crates/lib/build_metadata" }
cloudflare_errors = { path = "crates/lib/cloudflare_errors" }
cloudflare_mitigation = { path = "crates/lib/cloudflare_mitigation" }
cookie_store = { path = "crates/lib/cookie_store" }
enums = { path = "crates/schema/public/enums" }
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" }
gmicloud_client = { path = "crates/api_clients/gmicloud_client" }
grok_api_client = { path = "crates/api_clients/grok_api_client" }
grok_consumer_client = { path = "crates/api_clients/grok_consumer_client" }
http_headers = { path = "crates/lib/http_headers" }
images = { path = "crates/lib/files/images" }
jwt_light = { path = "crates/lib/jwt_light" }
memory_store = { path = "crates/lib/caching/memory_store" }
midjourney_client = { path = "crates/api_clients/midjourney_client" }
mimetypes = { path = "crates/lib/files/mimetypes" }
openai_sora_client = { path = "crates/api_clients/openai_sora_client" }
primitives = { path = "crates/lib/data/primitives" }
seedance2pro_client = { path = "crates/api_clients/seedance2pro_client" }
sqlite_tasks = { path = "crates/schema/database/sqlite_tasks" }
test_data = { path = "crates/testing/test_data" }
test_utils = { path = "crates/testing/test_utils" }
tokens = { path = "crates/schema/public/tokens" }
url_utils = { path = "crates/lib/url_utils" }
uuid_utils = { path = "crates/lib/uuid_utils" }
web_base64 = { path = "crates/lib/data/web_base64" }
worldlabs_api_client = { path = "crates/api_clients/worldlabs_api_client" }
worldlabs_consumer_client = { path = "crates/api_clients/worldlabs_consumer_client" }
# External libs
# Centrally pinned versions so that we can uniformly upgrade/downgrade versions across the workspace
# If a crate needs to use an older/newer version, the dependency should be managed in that crate's own Cargo.toml
anyhow = "1.0.102"
assertor = "0.0.3"
base64 = "0.22.1"
bstr = "1.12.1"
byteorder = "1.5.0"
bytes = "1.11.1"
chrono = { version = "0.4.44", features = ["serde"] }
cookie = "0.18.1"
env = "1.0.1"
env_logger = "0.11.9"
futures = { version = "0.3.32" }
indexmap = "2.13.0"
log = "0.4.29"
once_cell = "1.21.4"
rand = "0.9.1"
regex = "1.12.3"
reqwest = { version = "0.12.23", default-features = false, features = [ "charset", "http2", "system-proxy", "json", "cookies", "rustls-tls", "gzip" ] }
serde = { version = "1.0.228", features = ["derive"] }
serde_derive = "1.0.228"
serde_json = "1.0.149"
sha2 = "0.11.0-rc.5"
sha3 = "0.10.8"
strum = { version = "0.28.0", features = ["derive"] }
tauri = { version = "2.11.2", features = ["unstable", "devtools", "protocol-asset"] } # NB: "devtools" prevents App Store distribution; "unstable" is needed for window access
tauri-plugin-dialog = "2.7.1"
tauri-plugin-fs = "2.5.1"
tauri-plugin-log = "2.8.0"
tauri-plugin-upload = "2.4.0"
tempdir = "0.3.7"
tempfile = "3.27.0"
thiserror = "2.0.18"
tokio = { version = "1.50.0", features = ["macros"] }
tokio-util = "0.7.12"
toml = "1.0.7+spec-1.1.0"
url = { version = "2.5.8", features = ["serde"] }
utoipa = { version = "5.4.0", features = ["chrono", "actix_extras", "url"] }
uuid = "1.22.0"
walkdir = "2.4.0"
# 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-util = "3.0.0-rc.10"
wreq = { version = "6.0.0-rc.28", features = ["json", "query", "form", "multipart", "cookies", "ws", "stream", "gzip", "brotli", "zstd", "deflate"] }

70
Cranky.toml Normal file
View File

@ -0,0 +1,70 @@
# Persist clippy check configurations:
# https://github.com/ericseppanen/cargo-cranky
# Clippy doesn't allow this (yet):
# https://github.com/rust-lang/rust-clippy/issues/6625
deny = [
# Rust
"unsafe_code", # We may remove this if needed
# Clippy - "serious" lints
"clippy::borrow_interior_mutable_const", # Seems dangerous to allow this. Eg. `const F = Lazy::new()`;
# Clippy - bad math
#"clippy::cast_possible_truncation", # Seems dangerous, eg f64 to u64.
#"clippy::cast_possible_wrap",
# Clippy - pointer safety / readability / correctness
"clippy::as_ptr_cast_mut",
"clippy::borrow_as_ptr",
"clippy::borrow_deref_ref",
# Clippy async
"clippy::await_holding_lock",
# Clippy - performance
"clippy::cmp_owned",
"clippy::useless_vec",
"clippy::format_in_format_args",
#"clippy::redundant_clone",
#"clippy::unnecessary_to_owned",
#"clippy::vec_init_then_push",
# Clippy - "ergonomic" / "usage" lints
"clippy::clone_on_copy", # Don't use clone() on Copy types
"clippy::option_option", # Lint: Don't ever Option<Option<T>>! Enums instead.
"clippy::zero_sized_map_values", # Lint: use a HashSet instead of HashMap<T, ()>.
# Clippy - "aesthetic" / "style" lints
"clippy::clone_double_ref", # Lint: feels safer/hygienic to disallow this, but I'm not strongly opposed.
"clippy::double_comparisons", # Lint: readability
"clippy::empty_structs_with_brackets", # Lint: {} feels less aesthetic than ;
"clippy::implicit_clone", # Lint: Recommendation here seems more idiomatic.
"clippy::needless_late_init", # Don't declare variables without initializing them
"clippy::needless_return", # Lint: not the worst offender, but let's try to keep unnecessary returns out.
"clippy::redundant_field_names", # Lint: Duplicate names is a burden when reading: you wonder why they're dupes.
"clippy::redundant_static_lifetimes", # Lint: feels more modern to elide static lifetimes
# Rust
"non_fmt_panics", # panic!(format!(...)) will be unsupported soon
"unused_imports", # These are easy to automatically cull
"unused_mut", # Be strict about mutability
"unused_parens", # Unused parens are ugly
"unused_variables", # Don't leave cruft
# Possible:
# "clippy::ptr_arg",
# "clippy::too_many_lines" (note: this is configurable)
]
warn = [
]
allow = [
# Rust
"deprecated", # Other systems can handle deprecated code instead of the linter
"missing_docs", # Not everything needs documentation. Too much documentation is bad. Literate code.
# Clippy
"clippy::if_same_then_else", # There are cases this is clearer! Branches may have notes / considerations.
"clippy::single_char_pattern", # This is like punishing trailing commas. Chars are dumb.
]

View File

@ -0,0 +1,127 @@
-- noinspection SqlDialectInspectionForFile
--
-- NB: This version needs to be bumped even if just adding or changing comments!
-- Failure to do so will cause Windows to segfault at start. (This is horrible!)
--
-- NB: `app_state_dir.rs` contains the file version number.
--
-- Migration history:
-- 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,
-- tasks_v6.sqlite - added two fields:
-- on_failure_type
-- on_failure_message
-- tasks_v7.sqlite - added prompt token + two fields (for fal):
-- prompt_token
-- queue_status_url
-- queue_response_url
CREATE TABLE tasks (
-- Task auto-incrementing primary key.
-- We don't use this.
id TEXT NOT NULL PRIMARY KEY,
-- TaskStatus enum
-- This is how the job system manages job states.
-- e.g. pending, started, complete_success, dead, etc.
task_status TEXT NOT NULL,
-- TaskType enum
-- e.g. image_generation, video_generation, etc.
task_type TEXT NOT NULL,
-- TaskModelType enum
-- e.g. flux_1_dev, veo_2, etc.
model_type TEXT,
-- GenerationProvider enum
-- Together with `provider_job_id`, this creates a foreign key lookup to the job.
-- e.g. 'artcraft', 'fal', 'sora', etc.
provider TEXT,
-- The primary key for the job in the provider's system.
provider_job_id TEXT,
-- OPTIONAL.
-- For third party jobs, we need to record the prompt details upfront.
-- We create a prompt record and store the token to associate once the task completes.
prompt_token TEXT,
-- OPTIONAL.
-- Tell the job system which caller initiated the task.
-- e.g. 'canvas'
frontend_caller TEXT,
-- OPTIONAL.
-- An arbitrary opaque identifier that frontend can set.
frontend_subscriber_id TEXT,
-- OPTIONAL.
-- An opaque JSON payload set by the frontend subscriber that
-- 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.
-- For some third party systems (eg. FAL), the URL to check the status of the
-- job in the queue.
queue_status_url TEXT,
-- OPTIONAL.
-- For some third party systems (eg. FAL), the URL to check the
-- results (when the job completes)
queue_response_url TEXT,
-- 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,
-- OPTIONAL.
-- An enum-like string describing the failure reason (even for non-terminal failures).
-- The enum is `TaskFailureType`, which is a close analogue of the backend API's `FrontendFailureCategory`.
on_failure_type TEXT,
-- OPTIONAL.
-- A human-readable message describing the failure reason (even for non-terminal failures).
on_failure_message TEXT,
created_at INTEGER NOT NULL DEFAULT (unixepoch('now')),
updated_at INTEGER NOT NULL DEFAULT (unixepoch('now')),
completed_at INTEGER DEFAULT NULL
);
-- Unique Indices
CREATE UNIQUE INDEX idx_tasks_on_provider_job_id ON tasks(provider_job_id);
-- Indices
CREATE INDEX idx_tasks_on_status ON tasks(task_status);
CREATE INDEX idx_tasks_on_provider ON tasks(provider);

56
crates/CLAUDE.md Normal file
View File

@ -0,0 +1,56 @@
# Rust Workspace
All Rust code lives under `crates/`. The workspace is defined in `Cargo.toml` at the repo root.
## Code Style
- Two spaces for indentation (not four)
- No minimum supported Rust version; use latest stable features freely
- Never use `println!` or `eprintln!` outside of tests; use `log` crate macros instead
- Use `anyhow::Result` / `AnyhowResult` for fallible functions in application code
- Fields that are optional use the `maybe_` prefix: `maybe_creator_user_token`
- When two crates export the same type name, alias with a suffix: `use foo::Bar as BarFoo;`
- Prefer `use` imports over inline fully-qualified paths; only qualify inline for true one-offs or std prelude collisions (`Result`, `Option`, `Error`)
## File Layout
Organize for top-to-bottom reading. Important things first, details later.
1. **Constants** at the top (after imports)
2. **Primary public type** (the main struct/enum the file defines)
3. **Supporting types** below the primary type
4. **API types** in order: Request, Response, Error
5. **`impl` blocks** after their type definitions
- Constructors first, then public methods, then private helpers
- Callers above callees — follow the call chain downward
## Test Layout
Within `#[cfg(test)] mod tests { ... }`:
1. **Imports**
2. **Constants** (test IDs, URLs, fixture data) — small context that tests reference
3. **Test functions or sub-modules** grouped by category
4. **Helper functions** (builders, pipeline runners, assertions)
Constants provide small context up front. Test cases come next so engineers see *what* is
tested before *how* the helpers work.
Group related tests into sub-modules (`mod pricing_tests { ... }`) when a category has 2+
tests. Don't wrap a single test in its own module. Keep nesting to two levels max. Shared
helpers go in the parent `mod tests` so sub-modules can `use super::*`.
## Building
- Most crates: `cargo check -p {crate_name}` or `cargo test -p {crate_name}`
- Crates using SQLx (eg. `sqlite_tasks`): `SQLX_OFFLINE=true cargo check -p {crate_name}`
## Key Crates
- `artcraftx` (desktop) — Tauri desktop app
- `artcraft_router` — provider routing for image/video generation (Artcraft, Fal, Seedance2Pro)
- `artcraft_api_defs` — shared HTTP API type definitions
- `seedance2pro_client` — HTTP client for the Kinovi/Seedance2Pro video generation service
- `enums` — enum variants stored as strings
- `tokens` — identifiers with Stripe-like prefixes (e.g. `user_`, `mf_`)
- `sqlite_tasks` — SQLx queries for the desktop app's local task database

View File

@ -0,0 +1,22 @@
# artcraft_api_defs
Shared HTTP API type definitions used by both the server (`storyteller_web`) and clients.
## Structure
Organized by feature: `users/`, `media_file/`, `jobs/`, `moderation/`, etc.
## Conventions
- Request types: `#[derive(Deserialize, ToSchema)]`
- Response types: `#[derive(Serialize, ToSchema)]`
- Path params: `#[derive(Deserialize, ToSchema)]` with `PathInfo` suffix
- Query params: `#[derive(Deserialize, IntoParams)]`
## Orphan Rule
Types that need `ResponseError` impl (actix-web) CANNOT live here because of Rust's orphan rule. Keep error response types in `storyteller_web` and only put data types here.
## Dependencies
Only `serde`, `serde_derive`, `utoipa`, `chrono`, `url`, and internal schema crates (`enums`, `tokens`). No actix-web dependency.

View File

@ -0,0 +1,38 @@
[package]
name = "artcraft_api_defs"
edition = "2024"
version = "0.0.1"
authors = [
"Brandon Thomas <bt@brand.io>",
"Brandon Thomas <echelon@gmail.com>",
]
publish = false
[package.metadata.cargo-udeps.ignore]
normal = ["workspace-build-acceleration"]
[lib]
name = "artcraft_api_defs"
path = "src/lib.rs"
[dependencies]
# Hackari workspace (faster compile times)
workspace-build-acceleration.workspace = true
# Internal
enums.workspace = true
tokens.workspace = true
# External
chrono.workspace = true
url.workspace = true
# Serialization
serde.workspace = true
serde_derive.workspace = true
serde_json.workspace = true
utoipa.workspace = true
[dev-dependencies]
anyhow.workspace = true

View File

@ -0,0 +1,117 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::app_session::AppSessionToken;
use utoipa::ToSchema;
/// This is used in Artcraft versions 0.0.1 (several) through 0.3.0
/// Clients to this version do not send generation count data (even if the endpoint supports it).
pub const LOG_ACTIVE_USER_V1_PATH: &str = "/v1/analytics/active_user";
/// This is used in Artcraft versions 0.4.0 and later.
pub const LOG_ACTIVE_USER_V2_PATH: &str = "/v1/analytics/active_user_v2";
#[derive(Serialize, Deserialize, ToSchema, Debug, Default)]
pub struct LogAppActiveUserRequest {
/// Clientside-generated session token.
/// If present, it'll be validated against an expected format.
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_app_session_token: Option<AppSessionToken>,
/// An override for the application platform/OS (windows, mac, linux).
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_os_platform: Option<String>,
/// An override for the version of the OS (e.g. 10.15.7, 11, 22.04).
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_os_version: Option<String>,
/// An override for the application name.
/// If set together with `maybe_app_version`, the two will be
/// concatenated as `{maybe_app_name}/{maybe_app_version}`.
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_app_name: Option<String>,
/// An override for the application version.
/// If set together with `maybe_app_name`, the two will be
/// concatenated as `{maybe_app_name}/{maybe_app_version}`.
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_app_version: Option<String>,
/// How long the user has been active in the app, in seconds.
#[serde(skip_serializing_if = "Option::is_none")]
pub maybe_session_duration_seconds: Option<u64>,
// ========== GENERATION COUNTS ==========
/// Number of items generated across all types.
#[serde(skip_serializing_if = "Option::is_none")]
pub total_generation_count: Option<u16>,
/// Number of images generated.
#[serde(skip_serializing_if = "Option::is_none")]
pub image_generation_count: Option<u16>,
/// Number of videos generated.
#[serde(skip_serializing_if = "Option::is_none")]
pub video_generation_count: Option<u16>,
/// Number of objects generated (meshes, etc.).
#[serde(skip_serializing_if = "Option::is_none")]
pub object_generation_count: Option<u16>,
/// Number of images generated with text to image.
#[serde(skip_serializing_if = "Option::is_none")]
pub text_to_image_count: Option<u16>,
/// Number of images generated with image to image (image edits).
#[serde(skip_serializing_if = "Option::is_none")]
pub image_to_image_count: Option<u16>,
/// Number of videos generated with text to video.
#[serde(skip_serializing_if = "Option::is_none")]
pub text_to_video_count: Option<u16>,
/// Number of videos generated with image to video.
#[serde(skip_serializing_if = "Option::is_none")]
pub image_to_video_count: Option<u16>,
/// Number of objects generated with text to object.
#[serde(skip_serializing_if = "Option::is_none")]
pub text_to_object_count: Option<u16>,
/// Number of objects generated with image to object.
#[serde(skip_serializing_if = "Option::is_none")]
pub image_to_object_count: Option<u16>,
/// Number of prompts from the image page.
#[serde(skip_serializing_if = "Option::is_none")]
pub image_page_prompt_count: Option<u16>,
/// Number of prompts from the video page.
#[serde(skip_serializing_if = "Option::is_none")]
pub video_page_prompt_count: Option<u16>,
/// Number of prompts from the edit page.
#[serde(skip_serializing_if = "Option::is_none")]
pub edit_page_prompt_count: Option<u16>,
/// Number of prompts from the stage page.
#[serde(skip_serializing_if = "Option::is_none")]
pub stage_page_prompt_count: Option<u16>,
/// Number of prompts from the object page.
#[serde(skip_serializing_if = "Option::is_none")]
pub object_page_prompt_count: Option<u16>,
/// Number of prompts from other pages.
#[serde(skip_serializing_if = "Option::is_none")]
pub other_page_prompt_count: Option<u16>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct LogAppActiveUserResponse {
pub success: bool,
/// How long to wait until the next analytics event, in milliseconds.
/// The client should honor this and is free to add jitter.
pub wait_for_retry_millis: u64,
}

View File

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

View File

@ -0,0 +1,42 @@
use chrono::{DateTime, Utc};
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use tokens::tokens::api_keys::ApiKeyToken;
use tokens::tokens::users::UserToken;
/// Canonical wire shape for an API key, used by the list rows and the
/// single-key GET.
///
/// NB: the full secret `api_key` is NEVER included here — only
/// `truncated_api_key` (its first 20 characters), enough to disambiguate keys
/// in a UI. The full value is returned exactly once, at creation time.
#[derive(Clone, Debug, Serialize, Deserialize, ToSchema)]
pub struct ApiKeyInfo {
pub token: ApiKeyToken,
/// First 20 characters of the API key. The full secret is never returned
/// after creation.
pub truncated_api_key: String,
pub name: String,
pub maybe_description: Option<String>,
pub owner_user_token: UserToken,
pub ip_address_creation: String,
pub ip_address_update: String,
pub created_at: DateTime<Utc>,
pub updated_at: DateTime<Utc>,
/// Soft-delete timestamp. `None` for live keys.
pub maybe_deleted_at: Option<DateTime<Utc>>,
}
/// Path info for endpoints addressed by the API key's `token` (NOT the secret
/// `api_key` value): `GET` / `DELETE` / `PUT` `/v1/api_keys/{api_key_token}`.
#[derive(Deserialize, ToSchema)]
pub struct ApiKeyPathInfo {
pub api_key_token: ApiKeyToken,
}

View File

@ -0,0 +1,28 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use tokens::tokens::api_keys::ApiKeyToken;
// ── POST /v1/api_keys/create ──
#[derive(Deserialize, ToSchema)]
pub struct CreateApiKeyRequest {
/// Display name (title) for the key.
pub name: String,
/// Optional description.
pub maybe_description: Option<String>,
}
#[derive(Serialize, ToSchema)]
pub struct CreateApiKeySuccessResponse {
pub success: bool,
/// The key's `token` — used for all subsequent management
/// (get / update / delete). Safe to store and display.
pub api_key_token: ApiKeyToken,
/// The full secret API key value. This is the ONLY time it is ever
/// returned — the caller must store it now, as it cannot be retrieved again.
pub api_key: String,
}

View File

@ -0,0 +1,9 @@
use serde_derive::Serialize;
use utoipa::ToSchema;
// ── DELETE /v1/api_keys/{api_key_token} ──
#[derive(Serialize, ToSchema)]
pub struct DeleteApiKeySuccessResponse {
pub success: bool,
}

View File

@ -0,0 +1,12 @@
use serde_derive::Serialize;
use utoipa::ToSchema;
use crate::api_keys::common::ApiKeyInfo;
// ── GET /v1/api_keys/{api_key_token} ──
#[derive(Serialize, ToSchema)]
pub struct GetApiKeySuccessResponse {
pub success: bool,
pub api_key: ApiKeyInfo,
}

View File

@ -0,0 +1,24 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::{IntoParams, ToSchema};
use crate::api_keys::common::ApiKeyInfo;
// ── GET /v1/api_keys/list ──
#[derive(Deserialize, ToSchema, IntoParams)]
pub struct ListApiKeysQueryParams {
/// Page size. Defaults server-side if omitted.
pub limit: Option<u32>,
/// Number of rows to skip. Defaults to 0.
pub offset: Option<u32>,
}
#[derive(Serialize, ToSchema)]
pub struct ListApiKeysSuccessResponse {
pub success: bool,
/// The user's API keys (including soft-deleted ones), newest first. Each
/// entry's key value is truncated — the full secret is never listed.
pub api_keys: Vec<ApiKeyInfo>,
}

View File

@ -0,0 +1,6 @@
pub mod common;
pub mod create_api_key;
pub mod delete_api_key;
pub mod get_api_key;
pub mod list_api_keys;
pub mod update_api_key;

View File

@ -0,0 +1,15 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
// ── PUT /v1/api_keys/{api_key_token} ──
#[derive(Deserialize, ToSchema)]
pub struct UpdateApiKeyRequest {
/// New description for the key. `None` clears it.
pub maybe_description: Option<String>,
}
#[derive(Serialize, ToSchema)]
pub struct UpdateApiKeySuccessResponse {
pub success: bool,
}

View File

@ -0,0 +1,32 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use enums::common::generation::common_model_type::CommonModelType;
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
/// Request body for creating a character.
#[derive(Deserialize, ToSchema)]
pub struct CreateCharacterRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// The model to create the character for.
pub model: CommonModelType,
/// The reference image media file token.
pub image_media_token: MediaFileToken,
/// Name of the character.
pub character_name: String,
/// Description of the character.
pub character_description: Option<String>,
}
/// Response body for creating a character.
#[derive(Serialize, ToSchema)]
pub struct CreateCharacterResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,16 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use tokens::tokens::characters::CharacterToken;
/// Path parameters for deleting a character.
#[derive(Deserialize, ToSchema)]
pub struct DeleteCharacterPathInfo {
pub character_token: CharacterToken,
}
/// Response body for deleting a character.
#[derive(Serialize, ToSchema)]
pub struct DeleteCharacterResponse {
pub success: bool,
}

View File

@ -0,0 +1,23 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use tokens::tokens::characters::CharacterToken;
/// Request body for editing a character.
#[derive(Deserialize, ToSchema)]
pub struct EditCharacterRequest {
/// The character to edit.
pub token: CharacterToken,
/// Updated name (if changing).
pub updated_name: Option<String>,
/// Updated description (if changing).
pub updated_description: Option<String>,
}
/// Response body for editing a character.
#[derive(Serialize, ToSchema)]
pub struct EditCharacterResponse {
pub success: bool,
}

View File

@ -0,0 +1,36 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
use crate::common::responses::media_links::MediaLinks;
use enums::common::generation::common_model_type::CommonModelType;
use tokens::tokens::characters::CharacterToken;
/// Path parameters for getting a character.
#[derive(Deserialize, ToSchema)]
pub struct GetCharacterPathInfo {
pub character_token: CharacterToken,
}
/// Response body for getting a character.
#[derive(Serialize, ToSchema)]
pub struct GetCharacterResponse {
pub success: bool,
pub character: GetCharacterDetails,
}
/// Full character details returned by the get endpoint.
#[derive(Serialize, ToSchema)]
pub struct GetCharacterDetails {
pub token: CharacterToken,
/// Which models this character can be used with.
pub models: Vec<CommonModelType>,
pub name: String,
pub maybe_description: Option<String>,
pub maybe_avatar: Option<MediaLinks>,
pub maybe_full_image: Option<MediaLinks>,
}

View File

@ -0,0 +1,38 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::{IntoParams, ToSchema};
use crate::common::responses::media_links::MediaLinks;
use enums::common::generation::common_model_type::CommonModelType;
use tokens::tokens::characters::CharacterToken;
/// Query string parameters for listing characters.
#[derive(Deserialize, IntoParams)]
pub struct ListCharactersQuery {
/// Optional cursor for pagination.
pub cursor: Option<u64>,
}
/// Response body for listing characters in the current session.
#[derive(Serialize, ToSchema)]
pub struct ListCharactersResponse {
pub success: bool,
pub characters: Vec<ListCharactersEntry>,
pub next_cursor: Option<u64>,
}
/// A character entry in the list response.
#[derive(Serialize, ToSchema)]
pub struct ListCharactersEntry {
pub token: CharacterToken,
/// Which models this character can be used with.
pub models: Vec<CommonModelType>,
pub name: String,
pub maybe_description: Option<String>,
pub maybe_avatar: Option<MediaLinks>,
pub maybe_full_image: Option<MediaLinks>,
}

View File

@ -0,0 +1,5 @@
pub mod create_character;
pub mod delete_character;
pub mod edit_character;
pub mod get_character;
pub mod list_characters;

View File

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

View File

@ -0,0 +1,19 @@
use serde_derive::{Deserialize, Serialize};
use url::Url;
use utoipa::ToSchema;
// TODO(bt,2025-08-20): Replace the storyteller-web version of this.
// All it needs is the impl ported into a builder.
/// Cover image links can only be jpg, png, etc. No videos.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct CoverImageLinks {
/// Primary link to the cover image via the CDN.
pub cdn_url: Url,
/// Template to construct thumbnail URLs.
/// Replace the string `{WIDTH}` with the desired width.
/// Only relevant for image media files. (Video media files instead have
/// video previews, which, in turn, have their own thumbnail templates.)
pub thumbnail_template: String,
}

View File

@ -0,0 +1,33 @@
use serde::Deserialize;
use serde::Serialize;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
/// Details about submitted live portrait jobs (request arguments only)
#[derive(Serialize, Deserialize, ToSchema)]
pub struct JobDetailsLivePortraitRequest {
/// Source media token
/// This can be an image or a video media file token
/// This is what constitutes the "portrait" or the overall final video.
/// This video or image must contain a face.
pub source_media_file_token: MediaFileToken,
/// Driving media token
/// This must be a video media file token.
/// This drives the animation of the face, but the actor will disappear
/// and their facial expressions will be transferred to the source.
/// This video must contain a face.
pub face_driver_media_file_token: MediaFileToken,
}
/// Details about submitted lipsync jobs (request arguments only)
#[derive(Serialize, Deserialize, ToSchema)]
pub struct JobDetailsLipsyncRequest {
/// Media file token for the source audio.
/// This is probably an audio file, but in the future we might pull audio from video.
pub audio_source_token: MediaFileToken,
/// Media file token for the source visuals.
/// This is either an image or video.
pub image_or_video_source_token: MediaFileToken,
}

View File

@ -0,0 +1,44 @@
use crate::common::responses::cover_image_links::CoverImageLinks;
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
// TODO(bt,2025-08-20): Replace the storyteller-web version of this.
// We can't drop the old version just yet due to the `impl` (easy)
// and a few differing fields.
/// Everything we need to create a cover image.
/// Cover images are small descriptive images that can be set for any media file.
/// If a cover image is set, this is the path to the asset.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct MediaFileCoverImageDetails {
// /// (DEPRECATED) URL path to the media file
// #[deprecated(note="This field doesn't point to the full URL. Use media_links instead to leverage the CDN.")]
// pub maybe_cover_image_public_bucket_path: Option<String>,
// /// (DEPRECATED) Full URL to the media file
// #[deprecated(note="This points to the bucket. Use media_links instead to leverage the CDN.")]
// pub maybe_cover_image_public_bucket_url: Option<Url>,
// NB(bt,2024-09-19): I accidentally rolled this field out to production.
// I don't think this field is in use, but maybe ...
// /// (DEPRECATED) Use maybe_links instead.
// #[deprecated(note="Use `maybe_links` instead.")]
// pub maybe_media_links: Option<MediaLinks>,
/// Links to the cover image (CDN direct link, thumbnail template)
/// If a cover image is set, this is the path to the asset.
/// If a cover image is not set, use the information in `default_cover` instead.
/// Rich CDN links to the media, including thumbnails, previews, and more.
pub maybe_links: Option<CoverImageLinks>,
/// For items without a cover image, we can use one of our own.
pub default_cover: MediaFileDefaultCover,
}
/// The default cover is composed of an image and color pair that are
/// predefined by the frontend.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct MediaFileDefaultCover {
pub image_index: u8,
pub color_index: u8,
}

View File

@ -0,0 +1,39 @@
use serde::Deserialize;
use serde::Serialize;
use url::Url;
use utoipa::ToSchema;
/// Links to media file locations (bucket, CDN, etc.)
#[derive(Serialize, Deserialize, ToSchema, Debug, Clone, Eq, PartialEq)]
pub struct MediaLinks {
/// Primary link to the asset via the CDN.
pub cdn_url: Url,
/// Template to construct thumbnail URLs.
/// Replace the string `{WIDTH}` with the desired width.
/// Only relevant for image media files. (Video media files instead have
/// video previews, which, in turn, have their own thumbnail templates.)
pub maybe_thumbnail_template: Option<String>,
/// Video preview images (still and animated gif) for mp4 video files.
/// These are only set for video media files.
/// These are not set for "cover images" that are gif or webms, but rather video files.
pub maybe_video_previews: Option<VideoPreviews>,
}
#[derive(Serialize, Deserialize, ToSchema, Debug, Clone, Eq, PartialEq)]
pub struct VideoPreviews {
/// A static single frame preview image of the video.
pub still: Url,
/// An animated gif preview of the video.
pub animated: Url,
/// A template used to construct the still thumbnail URL.
/// Replace the string `{WIDTH}` with the desired width.
pub still_thumbnail_template: String,
/// A template used to construct the animated thumbnail URL.
/// Replace the string `{WIDTH}` with the desired width.
pub animated_thumbnail_template: String,
}

View File

@ -0,0 +1,8 @@
pub mod cover_image_links;
pub mod job_details;
pub mod media_file_cover_image_details;
pub mod media_links;
pub mod pagination_cursors;
pub mod simple_entity_stats;
pub mod simple_generic_json_success;
pub mod user_details_light;

View File

@ -0,0 +1,22 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
// TODO(bt,2026-07-09): Replace the storyteller-web version of this.
/// Pagination by cursors
/// These types of pagination are for "infinite scrolling", which do not reveal a number of pages.
/// This is good so that investors and competitors cannot reveal how many database records we have.
/// This is typically used for "discovery" type pages, not user profiles.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct PaginationCursors {
/// The "next" cursor.
/// This is an opaque (typically even encrypted) handle.
pub maybe_next: Option<String>,
/// The "previous" cursor.
/// This is an opaque (typically even encrypted) handle.
pub maybe_previous: Option<String>,
/// Details whether we're walking forward or backward.
pub cursor_is_reversed: bool,
}

View File

@ -0,0 +1,12 @@
use serde_derive::{Deserialize, Serialize};
// TODO(bt,2025-08-20): Replace the storyteller-web version of this.
/// Simple stats that can be attached to any entity
#[derive(Serialize, Deserialize, Debug)]
pub struct SimpleEntityStats {
/// Number of positive ratings (or "likes") for this item
pub positive_rating_count: u32,
/// Number of bookmarks for this item
pub bookmark_count: u32,
}

View File

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

View File

@ -0,0 +1,34 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::users::UserToken;
use utoipa::ToSchema;
/// Everything we need to refer to a user on the public web interface.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct UserDetailsLight {
/// The token for the user
pub user_token: UserToken,
/// The unique username someone logs in with
/// As of 2023-08-23, this is always lowercase
pub username: String,
/// As of 2023-08-23, this is the username with capitalization
/// (In the future, a display name can be customized by the user.)
pub display_name: String,
/// Email hash for Gravatar
/// Always set for now since login is email/username+password.
/// In the future this will need to become an optional *OR* be filled with a bogus hash.
pub gravatar_hash: String,
/// For users without a gravatar, we show one of our own avatars.
pub default_avatar: UserDefaultAvatarInfo,
// In the future, we'll also support user-uploaded avatars that we store on our servers.
}
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct UserDefaultAvatarInfo {
pub image_index: u8,
pub color_index: u8,
}

View File

@ -0,0 +1,20 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::ToSchema;
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GetSessionCreditsResponse {
pub success: bool,
/// Any free credits the user might have
/// We might support daily free credits.
pub free_credits: u64,
/// Any monthly refilling credits the user might have
pub monthly_credits: u64,
/// Any banked credits the user might have.
pub banked_credits: u64,
/// All the credit amounts added together.
pub sum_total_credits: u64,
}

View File

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

View File

@ -0,0 +1,94 @@
use chrono::{DateTime, Utc};
use serde_derive::{Deserialize, Serialize};
use url::Url;
use utoipa::ToSchema;
use enums::by_table::media_files::media_file_class::MediaFileClass;
use enums::by_table::media_files::media_file_type::MediaFileType;
use tokens::tokens::folders::FolderToken;
use tokens::tokens::media_files::MediaFileToken;
use tokens::tokens::users::UserToken;
// NB: The "folder media file" list-item shape lives in storyteller_web
// (see `endpoints/folders/media_files/list_folder_media_files_handler.rs`)
// because it embeds `MediaFileCoverImageDetails` and other domain types
// that depend on the request's `MediaDomain` + `ServerEnvironment`.
/// Canonical wire shape for a folder. Used by single-folder GETs, create
/// responses, list rows, and subfolder list rows.
#[derive(Clone, Debug, Serialize, Deserialize, ToSchema)]
pub struct FolderInfo {
pub token: FolderToken,
pub name: String,
pub owner_user_token: UserToken,
pub maybe_parent_folder_token: Option<FolderToken>,
/// The four most-recent media files added to the folder, in slot order
/// (the schema stores them as `maybe_last_media_file_token_1..4`).
/// Slots whose media file has been deleted are skipped, so the vec
/// may contain 0-4 entries.
pub last_media_thumbnails: Vec<FolderThumbnail>,
/// User-chosen cover image. When present, the frontend should use this
/// as the folder's primary thumbnail instead of the
/// `last_media_thumbnails` grid.
pub maybe_custom_cover_thumbnail: Option<FolderThumbnail>,
pub maybe_color_code: Option<String>,
pub has_star: bool,
pub created_at: DateTime<Utc>,
pub updated_at: DateTime<Utc>,
/// True when `maybe_parent_folder_token` is set but the referenced
/// parent row is missing or soft-deleted.
pub is_orphaned: bool,
}
/// A compact thumbnail descriptor for a media file referenced by a
/// folder (either one of the `last_media_*` slots or the custom cover).
///
/// Intentionally narrower than `MediaLinks` / `CoverImageLinks` — folders
/// only need the bare minimum to render a thumbnail card. The CDN URL is
/// the direct asset link; `maybe_thumbnail_template` carries the resizing
/// template URL (with `{{width}}` / `{{height}}` placeholders) when the
/// media class supports it.
#[derive(Clone, Debug, Serialize, Deserialize, ToSchema)]
pub struct FolderThumbnail {
pub token: MediaFileToken,
pub media_class: MediaFileClass,
pub media_type: MediaFileType,
/// Direct CDN URL to the underlying media asset.
pub cdn_url: Url,
/// Resizing template URL (with `{{width}}` / `{{height}}` placeholders)
/// when the media class supports it. `None` for media classes without
/// a thumbnail variant. (Video media files instead have video previews,
/// which, in turn, have their own thumbnail templates.)
pub maybe_thumbnail_template: Option<String>,
/// Video preview images (still and animated gif) for mp4 video files.
/// These are only set for video media files.
pub maybe_video_previews: Option<FolderThumbnailVideoPreviews>,
}
/// Video preview images for a video media file referenced by a folder.
/// Mirrors the shape of `MediaLinks`' video previews, but is its own type
/// so the folder wire shape can evolve independently.
#[derive(Clone, Debug, Serialize, Deserialize, ToSchema)]
pub struct FolderThumbnailVideoPreviews {
/// A static single frame preview image of the video.
pub still: Url,
/// An animated gif preview of the video.
pub animated: Url,
/// A template used to construct the still thumbnail URL.
/// Replace the string `{WIDTH}` with the desired width.
pub still_thumbnail_template: String,
/// A template used to construct the animated thumbnail URL.
/// Replace the string `{WIDTH}` with the desired width.
pub animated_thumbnail_template: String,
}

View File

@ -0,0 +1,116 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::{IntoParams, ToSchema};
use tokens::tokens::folders::FolderToken;
use tokens::tokens::media_files::MediaFileToken;
use crate::folders::common::FolderInfo;
// ── POST /v1/folders/create ──
#[derive(Deserialize, ToSchema)]
pub struct CreateFolderRequest {
pub name: String,
pub maybe_parent_folder_token: Option<FolderToken>,
pub maybe_color_code: Option<String>,
}
#[derive(Serialize, ToSchema)]
pub struct CreateFolderSuccessResponse {
pub success: bool,
pub folder: FolderInfo,
}
// ── GET /v1/folders/list_all ──
#[derive(Deserialize, ToSchema, IntoParams)]
pub struct ListFoldersQueryParams {
pub cursor: Option<String>,
pub limit: Option<u32>,
}
#[derive(Serialize, ToSchema)]
pub struct ListFoldersSuccessResponse {
pub success: bool,
pub folders: Vec<FolderInfo>,
pub maybe_cursor: Option<String>,
}
// ── GET /v1/folders/folder/{folder_token} ──
#[derive(Deserialize, ToSchema)]
pub struct FolderPathInfo {
pub folder_token: FolderToken,
}
#[derive(Serialize, ToSchema)]
pub struct GetFolderSuccessResponse {
pub success: bool,
pub folder: FolderInfo,
}
// ── PUT /v1/folders/folder/{folder_token}/rename ──
#[derive(Deserialize, ToSchema)]
pub struct RenameFolderRequest {
pub new_name: String,
}
#[derive(Serialize, ToSchema)]
pub struct RenameFolderSuccessResponse {
pub success: bool,
}
// ── PUT /v1/folders/folder/{folder_token}/star ──
#[derive(Deserialize, ToSchema)]
pub struct SetFolderStarRequest {
pub has_star: bool,
}
#[derive(Serialize, ToSchema)]
pub struct SetFolderStarSuccessResponse {
pub success: bool,
}
// ── PUT /v1/folders/folder/{folder_token}/color_code ──
#[derive(Deserialize, ToSchema)]
pub struct SetFolderColorCodeRequest {
/// `None` clears the color tag.
pub maybe_color_code: Option<String>,
}
#[derive(Serialize, ToSchema)]
pub struct SetFolderColorCodeSuccessResponse {
pub success: bool,
}
// ── PUT /v1/folders/folder/{folder_token}/cover_image ──
#[derive(Deserialize, ToSchema)]
pub struct SetFolderCoverImageRequest {
/// `None` clears the cover image. If set, the server resolves it to a
/// usable cover-image media file token: image/video media files are
/// used directly; other types fall back to the file's own
/// `maybe_cover_image_media_file_token` (and fail if neither path
/// applies).
pub maybe_media_file_token: Option<MediaFileToken>,
}
#[derive(Serialize, ToSchema)]
pub struct SetFolderCoverImageSuccessResponse {
pub success: bool,
/// The media file token that ended up being stored as the cover image
/// (may differ from the requested token if the server resolved through
/// the file's own cover image). `None` when the cover was cleared.
pub maybe_resolved_cover_media_file_token: Option<MediaFileToken>,
}
// ── DELETE /v1/folders/folder/{folder_token} ──
#[derive(Serialize, ToSchema)]
pub struct DeleteFolderSuccessResponse {
pub success: bool,
}

View File

@ -0,0 +1,111 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::{IntoParams, ToSchema};
use enums::by_table::media_files::media_file_class::MediaFileClass;
use tokens::tokens::folders::FolderToken;
use tokens::tokens::media_files::MediaFileToken;
// ── GET /v1/folders/media_files/{folder_token} ──
//
// NB: `ListFolderMediaFilesSuccessResponse` and `FolderMediaFileListItem`
// live in `storyteller_web`'s handler because the wire shape embeds
// `MediaFileCoverImageDetails` and `MediaLinks` constructors that depend
// on the request's `MediaDomain` + `ServerEnvironment`.
#[derive(Deserialize, ToSchema)]
pub struct FolderMediaFilesPathInfo {
pub folder_token: FolderToken,
}
#[derive(Deserialize, ToSchema, IntoParams)]
pub struct ListFolderMediaFilesQueryParams {
pub cursor: Option<String>,
pub limit: Option<u32>,
}
// ── GET /v1/folders/media_files_without_folder ──
//
// NB: `ListMediaFilesWithoutFolderSuccessResponse` and its list item live in
// `storyteller_web`'s handler because the wire shape embeds
// `MediaFileCoverImageDetails` and `MediaLinks` constructors that depend
// on the request's `MediaDomain` + `ServerEnvironment`.
#[derive(Deserialize, ToSchema, IntoParams)]
pub struct ListMediaFilesWithoutFolderQueryParams {
pub cursor: Option<String>,
pub limit: Option<u32>,
/// Optional filter on the coarse media class.
///
/// Usage:
/// - `?filter_media_class=image`
/// - `?filter_media_class=video`
/// - `?filter_media_class=mesh`
/// - `?filter_media_class=splat`
pub filter_media_class: Option<MediaFileClass>,
}
// ── PUT /v1/folders/media_files/{folder_token}/bulk_add ──
#[derive(Deserialize, ToSchema)]
pub struct BulkAddFolderMediaFilesRequest {
/// Media files to add to the folder. Tokens that don't exist or are
/// soft-deleted are silently skipped. Already-added rows are no-ops
/// (idempotent via INSERT IGNORE on the composite primary key).
pub media_file_tokens: Vec<MediaFileToken>,
}
#[derive(Serialize, ToSchema)]
pub struct BulkAddFolderMediaFilesSuccessResponse {
pub success: bool,
/// Tokens that pre-existed (existed + not deleted). Already-added rows
/// are included since the operation is idempotent.
pub accepted_media_file_tokens: Vec<MediaFileToken>,
}
// ── PUT /v1/folders/media_files/{folder_token}/bulk_remove ──
#[derive(Deserialize, ToSchema)]
pub struct BulkRemoveFolderMediaFilesRequest {
pub media_file_tokens: Vec<MediaFileToken>,
}
#[derive(Serialize, ToSchema)]
pub struct BulkRemoveFolderMediaFilesSuccessResponse {
pub success: bool,
/// How many membership rows were actually hard-deleted.
pub removed_count: u64,
}
// ── POST /v1/folders/media_files/{folder_token}/bulk_move ──
/// Atomically move the listed media files from `source_folder` to the
/// URL `folder_token` (destination). Idempotent on both sides: tokens
/// that aren't currently in the source are silently skipped, and tokens
/// that are already in the destination stay there.
#[derive(Deserialize, ToSchema)]
pub struct BulkMoveFolderMediaFilesRequest {
pub source_folder: FolderToken,
pub media_file_tokens: Vec<MediaFileToken>,
}
#[derive(Serialize, ToSchema)]
pub struct BulkMoveFolderMediaFilesSuccessResponse {
pub success: bool,
/// Number of membership rows actually deleted from the source folder.
pub removed_from_source_count: u64,
/// Number of membership rows actually inserted into the destination
/// folder. May be lower than the input length because (a) tokens that
/// were already present in the destination don't insert, and (b)
/// tokens whose media file doesn't exist / is soft-deleted are filtered
/// out before the insert.
pub added_to_destination_count: u64,
/// The media file tokens that ended up resident in the destination
/// folder after the move — i.e. the subset of the input that actually
/// referred to live media files. Includes both newly-inserted rows and
/// rows that were already there.
pub accepted_media_file_tokens: Vec<MediaFileToken>,
}

View File

@ -0,0 +1,4 @@
pub mod common;
pub mod folder;
pub mod media_files;
pub mod subfolder;

View File

@ -0,0 +1,61 @@
use serde_derive::{Deserialize, Serialize};
use utoipa::{IntoParams, ToSchema};
use tokens::tokens::folders::FolderToken;
use crate::folders::common::FolderInfo;
// ── GET /v1/folders/subfolders/{folder_token} ──
#[derive(Deserialize, ToSchema)]
pub struct SubfolderPathInfo {
pub folder_token: FolderToken,
}
#[derive(Deserialize, ToSchema, IntoParams)]
pub struct ListSubfoldersQueryParams {
pub cursor: Option<String>,
pub limit: Option<u32>,
}
#[derive(Serialize, ToSchema)]
pub struct ListSubfoldersSuccessResponse {
pub success: bool,
pub subfolders: Vec<FolderInfo>,
pub maybe_cursor: Option<String>,
}
// ── PUT /v1/folders/subfolders/{folder_token}/bulk_add ──
#[derive(Deserialize, ToSchema)]
pub struct BulkAddSubfoldersRequest {
/// The folders to reparent under the URL `folder_token`. Tokens that
/// don't exist, aren't owned by the caller, are soft-deleted, or are
/// the parent itself are silently skipped (idempotent).
pub subfolder_tokens: Vec<FolderToken>,
}
#[derive(Serialize, ToSchema)]
pub struct BulkAddSubfoldersSuccessResponse {
pub success: bool,
/// Tokens that were actually accepted (existed + owned + not deleted +
/// not the parent itself). Already-parented tokens are included here
/// since the operation is idempotent.
pub accepted_subfolder_tokens: Vec<FolderToken>,
}
// ── PUT /v1/folders/subfolders/{folder_token}/bulk_remove ──
#[derive(Deserialize, ToSchema)]
pub struct BulkRemoveSubfoldersRequest {
/// Folders to unparent. Only rows currently parented to the URL
/// `folder_token` are affected; others are silently skipped.
pub subfolder_tokens: Vec<FolderToken>,
}
#[derive(Serialize, ToSchema)]
pub struct BulkRemoveSubfoldersSuccessResponse {
pub success: bool,
/// How many rows were actually unparented.
pub removed_count: u64,
}

View File

@ -0,0 +1,122 @@
use serde::{Deserialize, Serialize};
use utoipa::ToSchema;
use enums::common::generation::common_aspect_ratio::CommonAspectRatio;
use enums::common::generation::common_image_model::CommonImageModel;
use enums::common::generation::common_quality::CommonQuality;
use enums::common::generation::common_resolution::CommonResolution;
use enums::common::generation_provider::GenerationProvider;
pub const ESTIMATE_IMAGE_COST_PATH: &str = "/v1/generate/cost_estimate/image";
/// Request body for the image cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct EstimateImageCostRequest {
/// The image model to estimate costs for.
pub model: CommonImageModel,
/// The provider to route the generation through.
pub provider: GenerationProvider,
/// The type of generation (determines whether input images are involved).
/// This is a tagged enum, so it looks like:
/// "generation_mode": {"type": "text_to_image"}
/// "generation_mode": {"type": "image_edit", "count": 2}
pub generation_mode: GenerationMode,
/// Optional aspect ratio.
pub aspect_ratio: Option<CommonAspectRatio>,
/// Optional resolution.
pub resolution: Option<CommonResolution>,
// Optional quality.
pub quality: Option<CommonQuality>,
/// Number of images to generate in parallel.
pub image_batch_count: Option<u16>,
}
/// Describes the type of image generation being requested.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
#[serde(tag = "type", rename_all = "snake_case")]
pub enum GenerationMode {
/// No input images — generate from prompt only.
TextToImage,
/// One or more input images are provided for editing/compositing.
ImageEdit { count: u32 },
}
/// Response body for the image cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateImageCostResponse {
pub success: bool,
/// Estimated cost in credits.
pub cost_in_credits: Option<u64>,
/// Estimated cost in USD cents.
pub cost_in_usd_cents: Option<u64>,
/// Whether the generation is free for this user/plan.
pub is_free: bool,
/// Whether the user has unlimited generations.
pub is_unlimited: bool,
/// Whether the user is rate limited.
pub is_rate_limited: bool,
/// Whether the output will have a watermark.
pub has_watermark: bool,
}
/// Error response for the image cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateImageCostError {
pub success: bool,
pub error_type: EstimateImageCostErrorType,
pub error_message: String,
}
#[derive(Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum EstimateImageCostErrorType {
InvalidProviderForModel,
InvalidInput,
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_serialization_text_to_image() {
let request = EstimateImageCostRequest {
model: CommonImageModel::NanoBananaPro,
provider: GenerationProvider::Artcraft,
generation_mode: GenerationMode::TextToImage,
aspect_ratio: None,
resolution: None,
quality: None,
image_batch_count: None,
};
let serialized = serde_json::to_string(&request).unwrap();
assert!(serialized.contains("\"generation_mode\":{\"type\":\"text_to_image\"}"));
}
#[test]
fn test_serialization_image_edit() {
let request = EstimateImageCostRequest {
model: CommonImageModel::NanoBananaPro,
provider: GenerationProvider::Artcraft,
generation_mode: GenerationMode::ImageEdit { count: 2 },
aspect_ratio: None,
resolution: None,
quality: None,
image_batch_count: None,
};
let serialized = serde_json::to_string(&request).unwrap();
assert!(serialized.contains("\"generation_mode\":{\"type\":\"image_edit\",\"count\":2}"));
}
}

View File

@ -0,0 +1,59 @@
use serde::{Deserialize, Serialize};
use utoipa::ToSchema;
use enums::common::generation::common_splat_model::CommonSplatModel;
use enums::common::generation_provider::GenerationProvider;
pub const ESTIMATE_SPLAT_COST_PATH: &str = "/v1/generate/cost_estimate/splat";
/// Request body for the splat cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateSplatCostRequest {
/// The splat model to estimate costs for.
pub model: CommonSplatModel,
/// The provider to route the generation through.
pub provider: GenerationProvider,
/// Whether the request includes a reference image.
pub has_reference_image: Option<bool>,
}
/// Response body for the splat cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateSplatCostResponse {
pub success: bool,
/// Estimated cost in credits.
pub cost_in_credits: Option<u64>,
/// Estimated cost in USD cents.
pub cost_in_usd_cents: Option<u64>,
/// Whether the generation is free for this user/plan.
pub is_free: bool,
/// Whether the user has unlimited generations.
pub is_unlimited: bool,
/// Whether the user is rate limited.
pub is_rate_limited: bool,
/// Whether the output will have a watermark.
pub has_watermark: bool,
}
/// Error response for the splat cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateSplatCostError {
pub success: bool,
pub error_type: EstimateSplatCostErrorType,
pub error_message: String,
}
#[derive(Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum EstimateSplatCostErrorType {
InvalidProviderForModel,
InvalidInput,
}

View File

@ -0,0 +1,130 @@
use serde::{Deserialize, Serialize};
use utoipa::ToSchema;
use enums::common::generation::common_aspect_ratio::CommonAspectRatio;
use enums::common::generation::common_video_model::CommonVideoModel;
use enums::common::generation::common_resolution::CommonResolution;
use enums::common::generation_provider::GenerationProvider;
pub const ESTIMATE_VIDEO_COST_PATH: &str = "/v1/generate/cost_estimate/video";
/// Request body for the video cost estimate endpoint.
///
/// Every model cares about its own set of parameters and not all
/// parameters may be relevant for every model.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct EstimateVideoCostRequest {
/// The video model to estimate costs for.
pub model: CommonVideoModel,
/// The provider to route the generation through.
pub provider: GenerationProvider,
/// The type of generation (determines how many input images are involved).
/// This is a tagged enum, so it looks like:
/// "generation_mode": {"type": "text_to_video"}
/// "generation_mode": {"type": "reference_image_to_video", "count": 1}
pub generation_mode: GenerationMode,
/// Optional aspect ratio.
pub aspect_ratio: Option<CommonAspectRatio>,
/// Optional resolution.
pub resolution: Option<CommonResolution>,
/// Duration in seconds.
pub duration_seconds: Option<u16>,
/// Number of videos to generate in parallel.
pub video_batch_count: Option<u16>,
/// Whether to generate audio for the video.
pub generate_audio: Option<bool>,
}
/// Describes the type of video generation being requested.
#[derive(Serialize, Deserialize, ToSchema, Debug)]
#[serde(tag = "type", rename_all = "snake_case")]
pub enum GenerationMode {
TextToVideo,
StartFrameToVideo,
StartAndEndFrameToVideo,
ReferenceImageToVideo { count: u32 },
}
/// Response body for the video cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateVideoCostResponse {
pub success: bool,
/// Estimated cost in credits.
pub cost_in_credits: Option<u64>,
/// Estimated cost in USD cents.
pub cost_in_usd_cents: Option<u64>,
/// Whether the generation is free for this user/plan.
pub is_free: bool,
/// Whether the user has unlimited generations.
pub is_unlimited: bool,
/// Whether the user is rate limited.
pub is_rate_limited: bool,
/// Whether the output will have a watermark.
pub has_watermark: bool,
}
/// Error response for the video cost estimate endpoint.
#[derive(Serialize, Deserialize, ToSchema)]
pub struct EstimateVideoCostError {
pub success: bool,
pub error_type: EstimateVideoCostErrorType,
pub error_message: String,
}
#[derive(Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum EstimateVideoCostErrorType {
InvalidProviderForModel,
InvalidInput,
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_serialization_tagging1() {
let request = EstimateVideoCostRequest {
model: CommonVideoModel::GrokVideo,
provider: GenerationProvider::Artcraft,
generation_mode: GenerationMode::TextToVideo,
aspect_ratio: None,
resolution: None,
duration_seconds: None,
video_batch_count: None,
generate_audio: None,
};
let serialized = serde_json::to_string(&request).unwrap();
assert!(serialized.contains("\"generation_mode\":{\"type\":\"text_to_video\"}"));
//assert_eq!(serialized, "{}");
}
#[test]
fn test_serialization_tagging2() {
let request = EstimateVideoCostRequest {
model: CommonVideoModel::GrokVideo,
provider: GenerationProvider::Artcraft,
generation_mode: GenerationMode::ReferenceImageToVideo { count: 1 },
aspect_ratio: None,
resolution: None,
duration_seconds: None,
video_batch_count: None,
generate_audio: None,
};
let serialized = serde_json::to_string(&request).unwrap();
assert!(serialized.contains("\"generation_mode\":{\"type\":\"reference_image_to_video\",\"count\":1}"));
}
}

View File

@ -0,0 +1,3 @@
pub mod estimate_image_cost;
pub mod estimate_splat_cost;
pub mod estimate_video_cost;

View File

@ -0,0 +1,56 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const FLUX_2_LORA_EDIT_IMAGE_ANGLE_PATH: &str = "/v1/generate/image/angle/flux_2_lora";
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct Flux2LoraEditImageAngleRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// The image to edit with angle adjustment.
pub image_media_token: MediaFileToken,
/// Horizontal rotation angle in degrees.
pub horizontal_angle: Option<f64>,
/// Vertical rotation angle in degrees.
pub vertical_angle: Option<f64>,
/// Zoom level.
pub zoom: Option<f64>,
/// Number of images to generate. Default is one.
pub num_images: Option<Flux2LoraEditImageAngleNumImages>,
/// Output image size. Default is square_hd.
pub image_size: Option<Flux2LoraEditImageAngleImageSize>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Flux2LoraEditImageAngleNumImages {
One,
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Flux2LoraEditImageAngleImageSize {
Square,
SquareHd,
PortraitFourThree,
PortraitSixteenNine,
LandscapeFourThree,
LandscapeSixteenNine,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct Flux2LoraEditImageAngleResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,2 @@
pub mod flux_2_lora_edit_image_angle;
pub mod qwen_edit_2511_edit_image_angle;

View File

@ -0,0 +1,59 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const QWEN_EDIT_2511_EDIT_IMAGE_ANGLE_PATH: &str = "/v1/generate/image/angle/qwen_edit_2511";
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct QwenEdit2511EditImageAngleRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// The image to edit with angle adjustment.
pub image_media_token: MediaFileToken,
/// Horizontal rotation angle in degrees.
pub horizontal_angle: Option<f64>,
/// Vertical rotation angle in degrees.
pub vertical_angle: Option<f64>,
/// Zoom level.
pub zoom: Option<f64>,
/// Additional text prompt to guide the generation.
pub additional_prompt: Option<String>,
/// Number of images to generate. Default is one.
pub num_images: Option<QwenEdit2511EditImageAngleNumImages>,
/// Output image size. Default is square_hd.
pub image_size: Option<QwenEdit2511EditImageAngleImageSize>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum QwenEdit2511EditImageAngleNumImages {
One,
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum QwenEdit2511EditImageAngleImageSize {
Square,
SquareHd,
PortraitFourThree,
PortraitSixteenNine,
LandscapeFourThree,
LandscapeSixteenNine,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct QwenEdit2511EditImageAngleResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

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

View File

@ -0,0 +1,22 @@
use serde::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const REMOVE_IMAGE_BACKGROUND_PATH: &str = "/v1/generate/image/remove_background";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct RemoveImageBackgroundRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the file to remove the background from.
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct RemoveImageBackgroundResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,36 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const FLUX_PRO_KONTEXT_MAX_EDIT_IMAGE_PATH: &str = "/v1/generate/image/edit/flux_pro_kontext_max";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxProKontextMaxEditImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// The image we're editing.
pub image_media_token: MediaFileToken,
/// Number of images to generate. Default is one.
pub num_images: Option<FluxProKontextMaxEditImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum FluxProKontextMaxEditImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxProKontextMaxEditImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -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/edit/gemini_25_flash";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct Gemini25FlashEditImageRequest {
/// 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<Gemini25FlashEditImageNumImages>,
/// Quality of the image to generate. Default is High.
pub image_quality: Option<Gemini25FlashEditImageImageQuality>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Gemini25FlashEditImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Gemini25FlashEditImageImageQuality {
Auto,
Low,
Medium,
High, // Default
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct Gemini25FlashEditImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -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 GPT_IMAGE_1_EDIT_IMAGE_PATH: &str = "/v1/generate/image/edit/gpt_image_1";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GptImage1EditImageRequest {
/// 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>>,
/// Size of the image to generate. Default is Square.
pub image_size: Option<GptImage1EditImageImageSize>,
/// Number of images to generate. Default is one.
pub num_images: Option<GptImage1EditImageNumImages>,
/// Quality of the image to generate. Default is High.
pub image_quality: Option<GptImage1EditImageImageQuality>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GptImage1EditImageImageSize {
Square, // 1:1, Default
Horizontal,
Vertical,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GptImage1EditImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GptImage1EditImageImageQuality {
Auto,
Low,
Medium,
High, // Default
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GptImage1EditImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,6 @@
pub mod flux_pro_kontext_max_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;

View File

@ -0,0 +1,51 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const QWEN_EDIT_IMAGE_PATH: &str = "/v1/generate/image/edit/qwen";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct QwenEditImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// The image we're editing.
pub image_media_token: MediaFileToken,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Text prompt to generate the image from.
pub negative_prompt: Option<String>,
/// Acceleration level for image generation.
/// Options: 'none', 'regular'. Higher acceleration increases speed.
/// 'regular' balances speed and quality. Default value: "none"
pub acceleration: Option<QwenEditImageAcceleration>,
/// Number of images to generate. Default is one.
pub num_images: Option<QwenEditImageNumImages>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub enum QwenEditImageAcceleration {
None,
Regular,
High,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum QwenEditImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct QwenEditImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,24 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const SEEDEDIT_3_EDIT_IMAGE_PATH: &str = "/v1/generate/image/edit/seededit_3";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct SeedEdit3EditImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// The image we're editing.
pub image_media_token: MediaFileToken,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct SeedEdit3EditImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,39 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const FLUX_DEV_JUGGERNAUT_INPAINT_PATH: &str = "/v1/generate/image/inpaint/flux_dev_juggernaut";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxDevJuggernautInpaintImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// The image we're editing.
pub image_media_token: MediaFileToken,
/// The mask to use against the image.
pub mask_media_token: MediaFileToken,
/// Number of images to generate. Default is one.
pub num_images: Option<FluxDevJuggernautInpaintImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum FluxDevJuggernautInpaintImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxDevJuggernautInpaintImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,39 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const FLUX_PRO_1_INPAINT_PATH: &str = "/v1/generate/image/inpaint/flux_pro_1";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxPro1InpaintImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// The image we're editing.
pub image_media_token: MediaFileToken,
/// The mask to use against the image.
pub mask_media_token: MediaFileToken,
/// Number of images to generate. Default is one.
pub num_images: Option<FluxPro1InpaintImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum FluxPro1InpaintImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct FluxPro1InpaintImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,2 @@
pub mod flux_dev_juggernaut_inpaint_image;
pub mod flux_pro_1_inpaint_image;

View File

@ -0,0 +1,6 @@
pub mod angle;
pub mod bg_removal;
pub mod edit;
pub mod inpaint;
pub mod multi_function;
pub mod text;

View File

@ -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_5_LITE_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/bytedance_seedream_5_lite";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct BytedanceSeedream5LiteMultiFunctionImageGenRequest {
/// 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<BytedanceSeedream5LiteMultiFunctionImageGenNumImages>,
pub max_images: Option<BytedanceSeedream5LiteMultiFunctionImageGenMaxImages>,
pub image_size: Option<BytedanceSeedream5LiteMultiFunctionImageGenImageSize>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum BytedanceSeedream5LiteMultiFunctionImageGenNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum BytedanceSeedream5LiteMultiFunctionImageGenMaxImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum BytedanceSeedream5LiteMultiFunctionImageGenImageSize {
// Square
Square,
SquareHd,
// Tall
PortraitFourThree,
PortraitSixteenNine,
// Wide
LandscapeFourThree,
LandscapeSixteenNine,
// Auto (NB: 5 Lite max is auto_3K, not auto_4K like V4.5)
Auto2k,
Auto3k,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct BytedanceSeedream5LiteMultiFunctionImageGenResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -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,
}

View File

@ -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,
}

View File

@ -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,
}

View File

@ -0,0 +1,7 @@
pub mod bytedance_seedream_v4_multi_function_image_gen;
pub mod bytedance_seedream_v4p5_multi_function_image_gen;
pub mod bytedance_seedream_5_lite_multi_function_image_gen;
pub mod gpt_image_1p5_multi_function_image_gen;
pub mod nano_banana_2_multi_function_image_gen;
pub mod nano_banana_multi_function_image_gen;
pub mod nano_banana_pro_multi_function_image_gen;

View File

@ -0,0 +1,74 @@
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_2_MULTI_FUNCTION_IMAGE_GEN_PATH: &str = "/v1/generate/image/multi_function/nano_banana_2";
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct NanaBanana2MultiFunctionImageGenRequest {
/// 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<NanaBanana2MultiFunctionImageGenNumImages>,
/// Resolution of the image to generate. Default is OneK (1K).
pub resolution: Option<NanaBanana2MultiFunctionImageGenImageResolution>,
/// Aspect ratio of the images to generate. Default is "1:1"
pub aspect_ratio: Option<NanaBanana2MultiFunctionImageGenAspectRatio>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum NanaBanana2MultiFunctionImageGenNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum NanaBanana2MultiFunctionImageGenImageResolution {
HalfK,
OneK,
TwoK,
FourK,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum NanaBanana2MultiFunctionImageGenAspectRatio {
// 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,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct NanaBanana2MultiFunctionImageGenResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -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,
}

View File

@ -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,
}

View File

@ -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,
}

View File

@ -0,0 +1,49 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use utoipa::ToSchema;
pub const GENERATE_FLUX_1_DEV_TEXT_TO_IMAGE_PATH: &str = "/v1/generate/image/flux_1_dev_text_to_image";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFlux1DevTextToImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Aspect ratio of the output images.
pub aspect_ratio: Option<GenerateFlux1DevTextToImageAspectRatio>,
/// Number of images to generate. Default is one.
pub num_images: Option<GenerateFlux1DevTextToImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFlux1DevTextToImageAspectRatio {
Square, // 1:1
SquareHd, // 1:1
LandscapeFourByThree, // 4:3
LandscapeSixteenByNine, // 16:9
PortraitThreeByFour, // 3:4
PortraitNineBySixteen, // 9:16
//Custom { width: u32, height: u32 }, // TODO
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFlux1DevTextToImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFlux1DevTextToImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,49 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use utoipa::ToSchema;
pub const GENERATE_FLUX_1_SCHNELL_TEXT_TO_IMAGE_PATH: &str = "/v1/generate/image/flux_1_schnell_text_to_image";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFlux1SchnellTextToImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Aspect ratio of the output images.
pub aspect_ratio: Option<GenerateFlux1SchnellTextToImageAspectRatio>,
/// Number of images to generate. Default is one.
pub num_images: Option<GenerateFlux1SchnellTextToImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFlux1SchnellTextToImageAspectRatio {
Square, // 1:1
SquareHd, // 1:1
LandscapeFourByThree, // 4:3
LandscapeSixteenByNine, // 16:9
PortraitThreeByFour, // 3:4
PortraitNineBySixteen, // 9:16
//Custom { width: u32, height: u32 }, // TODO
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFlux1SchnellTextToImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFlux1SchnellTextToImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,47 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use utoipa::ToSchema;
pub const GENERATE_FLUX_PRO_11_TEXT_TO_IMAGE_PATH: &str = "/v1/generate/image/flux_pro_1.1_text_to_image";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFluxPro11TextToImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Aspect ratio of the output images.
pub aspect_ratio: Option<GenerateFluxPro11TextToImageAspectRatio>,
/// Number of images to generate. Default is one.
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 {
Square, // 1:1
LandscapeFourByThree, // 4:3
LandscapeSixteenByNine, // 16:9
PortraitThreeByFour, // 3:4
PortraitNineBySixteen, // 9:16
//Custom { width: u32, height: u32 }, // TODO
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFluxPro11TextToImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFluxPro11TextToImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,50 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use utoipa::ToSchema;
pub const GENERATE_FLUX_PRO_11_ULTRA_TEXT_TO_IMAGE_PATH: &str = "/v1/generate/image/flux_pro_1.1_ultra_text_to_image";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFluxPro11UltraTextToImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Aspect ratio of the output images.
pub aspect_ratio: Option<GenerateFluxPro11UltraTextToImageAspectRatio>,
/// Number of images to generate. Default is one.
pub num_images: Option<GenerateFluxPro11UltraTextToImageNumImages>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[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
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateFluxPro11UltraTextToImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateFluxPro11UltraTextToImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,58 @@
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_1_TEXT_TO_IMAGE_PATH: &str = "/v1/generate/image/gpt_image_1_text_to_image";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateGptImage1TextToImageRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt to generate the image from.
pub prompt: Option<String>,
/// Size of the image to generate. Default is Square.
pub image_size: Option<GenerateGptImage1TextToImageImageSize>,
/// Number of images to generate. Default is one.
pub num_images: Option<GenerateGptImage1TextToImageNumImages>,
/// Quality of the image to generate. Default is High.
pub image_quality: Option<GenerateGptImage1TextToImageImageQuality>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateGptImage1TextToImageImageSize {
Square, // 1:1, Default
Horizontal,
Vertical,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateGptImage1TextToImageNumImages {
One, // Default
Two,
Three,
Four,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateGptImage1TextToImageImageQuality {
Auto,
Low,
Medium,
High, // Default
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateGptImage1TextToImageResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,6 @@
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;

View File

@ -0,0 +1,5 @@
pub mod cost_estimate;
pub mod image;
pub mod object;
pub mod splat;
pub mod video;

View File

@ -0,0 +1,22 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_HUNYUAN_2_0_IMAGE_TO_3D_URL_PATH: &str = "/v1/generate/object/hunyuan_2.0_image_to_3d";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateHunyuan20ImageTo3dRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the file to convert to 3D
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateHunyuan20ImageTo3dResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,22 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_HUNYUAN_2_1_IMAGE_TO_3D_URL_PATH: &str = "/v1/generate/object/hunyuan_2.1_image_to_3d";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateHunyuan21ImageTo3dRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the file to convert to 3D
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateHunyuan21ImageTo3dResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,3 @@
pub mod generate_hunyuan_2_0_image_to_3d;
pub mod generate_hunyuan_2_1_image_to_3d;
pub mod multi_function;

View File

@ -0,0 +1,84 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const HUNYUAN3D_V3_MULTI_FUNCTION_OBJECT_GEN_PATH: &str = "/v1/generate/object/multi_function/hunyuan3d_v3";
/// Hunyuan 3D v3 Multi-Function Request
///
/// This endpoint combines three modes:
/// - Text-to-3D: Provide only a prompt (no image_media_token)
/// - Image-to-3D: Provide an image_media_token (no prompt or empty prompt)
/// - Sketch-to-3D: Provide both a prompt AND an image_media_token
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct Hunyuan3dV3MultiFunctionObjectGenRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Text prompt for generation.
/// - Required for text-to-3d mode
/// - Required for sketch-to-3d mode
/// - Ignored for image-to-3d mode
pub prompt: Option<String>,
/// Primary image media token.
/// - Required for image-to-3d mode
/// - Required for sketch-to-3d mode
/// - Not used for text-to-3d mode
pub image_media_token: Option<MediaFileToken>,
/// Back view image for multi-view image-to-3d mode.
/// Only applicable when image_media_token is provided without a prompt.
pub back_image_media_token: Option<MediaFileToken>,
/// Left view image for multi-view image-to-3d mode.
/// Only applicable when image_media_token is provided without a prompt.
pub left_image_media_token: Option<MediaFileToken>,
/// Right view image for multi-view image-to-3d mode.
/// Only applicable when image_media_token is provided without a prompt.
pub right_image_media_token: Option<MediaFileToken>,
/// Target face count for the generated mesh.
/// Optional - if not provided, the model will use its default.
pub face_count: Option<u32>,
/// Type of generation to perform.
/// Default is Normal.
pub generate_type: Option<Hunyuan3dV3GenerateType>,
/// Type of polygon to use in the generated mesh.
/// Default is Triangle.
pub polygon_type: Option<Hunyuan3dV3PolygonType>,
/// Whether to enable PBR (Physically Based Rendering) materials.
/// Default is false.
pub enable_pbr: Option<bool>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Hunyuan3dV3GenerateType {
/// Standard generation quality (default)
Normal,
/// Low polygon count for game assets
LowPoly,
/// Geometry only, no textures
Geometry,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum Hunyuan3dV3PolygonType {
/// Triangle polygons (default)
Triangle,
/// Quadrilateral polygons
Quadrilateral,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct Hunyuan3dV3MultiFunctionObjectGenResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1 @@
pub mod hunyuan3d_v3_multi_function_object_gen;

View File

@ -0,0 +1,25 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_WORLDLABS_MARBLE_0P1_MINI_SPLAT_URL_PATH: &str = "/v1/generate/splat/worldlabs_marble_0p1_mini";
/// World Labs Marble 0.1-mini Splat Generation
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateWorldlabsMarble0p1MiniSplatRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Optional image to use as input for world generation.
pub image_media_file_token: Option<MediaFileToken>,
/// Optional text prompt for world generation.
pub prompt: Option<String>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateWorldlabsMarble0p1MiniSplatResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,25 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_WORLDLABS_MARBLE_0P1_PLUS_SPLAT_URL_PATH: &str = "/v1/generate/splat/worldlabs_marble_0p1_plus";
/// World Labs Marble 0.1-plus Splat Generation
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateWorldlabsMarble0p1PlusSplatRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Optional image to use as input for world generation.
pub image_media_file_token: Option<MediaFileToken>,
/// Optional text prompt for world generation.
pub prompt: Option<String>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateWorldlabsMarble0p1PlusSplatResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,2 @@
pub mod generate_worldlabs_marble_0p1_mini_splat;
pub mod generate_worldlabs_marble_0p1_plus_splat;

View File

@ -0,0 +1,20 @@
use serde_derive::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const BEEBLE_SWITCHX_EDIT_VIDEO_PATH: &str = "/v1/generate/video/edit/beeble_switchx";
#[derive(Serialize, Deserialize, ToSchema, Debug)]
pub struct BeebleSwitchXEditVideoRequest {
pub uuid_idempotency_token: String,
pub source_video_media_token: Option<MediaFileToken>,
pub reference_image_media_token: Option<MediaFileToken>,
pub prompt: Option<String>,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct BeebleSwitchXEditVideoResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

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

View File

@ -0,0 +1,53 @@
use serde::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_KLING_1_6_PRO_IMAGE_TO_VIDEO_URL_PATH: &str = "/v1/generate/video/kling_1.6_pro_image_to_video";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling16ProImageToVideoRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the image file to convert to video.
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
/// Optional.
/// Image media file; the image to remove the background from.
pub end_frame_image_media_token: Option<MediaFileToken>,
/// Optional text prompt.
pub prompt: Option<String>,
/// Optional: aspect ratio of the generated video.
pub aspect_ratio: Option<GenerateKling16ProAspectRatio>,
/// Optional: aspect ratio of the generated video.
pub duration: Option<GenerateKling16ProDuration>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling16ProAspectRatio {
/// 16:9 aspect ratio
WideSixteenNine,
/// 9:16 aspect ratio
TallNineSixteen,
/// 1:1 aspect ratio
Square,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling16ProDuration {
FiveSeconds,
TenSeconds,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling16ProImageToVideoResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,49 @@
use serde::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_KLING_2_1_MASTER_IMAGE_TO_VIDEO_URL_PATH: &str = "/v1/generate/video/kling_2.1_master_image_to_video";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling21MasterImageToVideoRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the image file to convert to video.
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
/// Optional text prompt.
pub prompt: Option<String>,
/// Optional: aspect ratio of the generated video.
pub aspect_ratio: Option<GenerateKling21MasterAspectRatio>,
/// Optional: aspect ratio of the generated video.
pub duration: Option<GenerateKling21MasterDuration>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling21MasterAspectRatio {
/// 16:9 aspect ratio
WideSixteenNine,
/// 9:16 aspect ratio
TallNineSixteen,
/// 1:1 aspect ratio
Square,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling21MasterDuration {
FiveSeconds,
TenSeconds,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling21MasterImageToVideoResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,53 @@
use serde::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_KLING_2_1_PRO_IMAGE_TO_VIDEO_URL_PATH: &str = "/v1/generate/video/kling_2.1_pro_image_to_video";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling21ProImageToVideoRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the image file to convert to video.
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
/// Optional.
/// Image media file; the image to remove the background from.
pub end_frame_image_media_token: Option<MediaFileToken>,
/// Optional text prompt.
pub prompt: Option<String>,
/// Optional: aspect ratio of the generated video.
pub aspect_ratio: Option<GenerateKling21ProAspectRatio>,
/// Optional: aspect ratio of the generated video.
pub duration: Option<GenerateKling21ProDuration>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling21ProAspectRatio {
/// 16:9 aspect ratio
WideSixteenNine,
/// 9:16 aspect ratio
TallNineSixteen,
/// 1:1 aspect ratio
Square,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateKling21ProDuration {
FiveSeconds,
TenSeconds,
}
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateKling21ProImageToVideoResponse {
pub success: bool,
pub inference_job_token: InferenceJobToken,
}

View File

@ -0,0 +1,65 @@
use serde::{Deserialize, Serialize};
use tokens::tokens::generic_inference_jobs::InferenceJobToken;
use tokens::tokens::media_files::MediaFileToken;
use utoipa::ToSchema;
pub const GENERATE_SEEDANCE_1_0_LITE_IMAGE_TO_VIDEO_URL_PATH: &str = "/v1/generate/video/seedance_1.0_lite_image_to_video";
#[derive(Serialize, Deserialize, ToSchema)]
pub struct GenerateSeedance10LiteImageToVideoRequest {
/// Idempotency token to prevent duplicate requests.
pub uuid_idempotency_token: String,
/// Source of the image file to convert to video.
/// It must be an image.
pub media_file_token: Option<MediaFileToken>,
/// Optional.
/// Image media file; the image to remove the background from.
pub end_frame_image_media_token: Option<MediaFileToken>,
/// Optional text prompt.
pub prompt: Option<String>,
/// Optional: resolution of the generated video.
pub resolution: Option<GenerateSeedance10LiteResolution>,
/// Optional: duration of the generated video.
pub duration: Option<GenerateSeedance10LiteDuration>,
/// Optional: aspect ratio of the generated video.
pub aspect_ratio: Option<GenerateSeedance10LiteAspectRatio>,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateSeedance10LiteResolution {
FourEightyP,
SevenTwentyP,
}
#[derive(Copy, Clone, Debug, Serialize, Deserialize, ToSchema)]
#[serde(rename_all = "snake_case")]
pub enum GenerateSeedance10LiteDuration {
FiveSeconds,
TenSeconds,
}
#[derive(Copy, Clone, Debug, 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,
pub inference_job_token: InferenceJobToken,
}

Some files were not shown because too many files have changed in this diff Show More