andodeki/makepad
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
makepad/libs/flow/recipes/prelude_recipes.splash
Admin 83a00d2801 flow: archive generated assets and add live source editing 
Host or attach the asset service, archive generated and terminal outputs, and provide image/video/audio/text/3D viewers with fullscreen and gallery navigation. Load only visible bounded previews, generate video posters and text excerpts, and normalize legacy thumbnails off the UI thread.

Expose video frame counts and Sandbox chat providers, make failed queue cleanup idempotent, compact the panels, and correct short-wire direction. Replace App view with a movable, resizable CodeEditor overlay with debounced source saves, live graph updates, invalid-source preservation and close/reopen geometry.

Validation: release Flow, Flow UI and flowgraph suites passed 335 tests (two ignored). Updated Flow UI suite passed 107 tests; asset route suites passed nine tests after legacy preview fixes. The rebuilt release app passed remote editing, dragging, resizing, invalid-source, close/reopen and App removal checks; final All assets scrolling measured 16.5 ms median and 18.3 ms maximum over 12 inputs. Test instances were closed through the remote protocol.
2026-09-05 01:47:53 +02:00

355 lines
15 KiB
Text
Raw Permalink Blame History

/**
Recipe prototypes for the asset creator's generation kinds.
`text.expand` deliberately uses the existing `mod.flow.Llm` node instead of
adding an `Expand` prototype: expansion is a pure text turn, and `Llm` exposes
the system instruction that makes the rewrite inspectable in each template.
*/
/** Ask a vision model a question about an image. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Vision = mod.flow.Gen{
kind: @gen
domain: "vision"
type_name: "Vision"
/** 16..512 */ max_tokens: 512
/** Empty selects the hub's best model for the vision domain. */ model: ""
ports: { in: {prompt: @text, image: @image}, out: {text: @text} }
text: fn() { self.out(@text) }
/** Generic parameter and text-result face. */ ui: mod.flow.ui.GenFace
}
/** Edit a primary image with an instruction. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.ImageEdit = mod.flow.Gen{
kind: @gen
domain: "edit"
type_name: "ImageEdit"
/** one of: 1.0, 0.85, 0.7, 0.55, 0.4, 0.25 */ strength: 1.0
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the edit domain. */ model: ""
ports: {
in: {
prompt: @text
image: @image
/** Optional additional reference image; up to three total. */ reference_1: @image
/** Optional additional reference image; up to three total. */ reference_2: @image
/** Optional additional reference image; up to three total. */ reference_3: @image
}
out: { image: @image }
}
image: fn() { self.out(@image) }
/** Image preview with generic edit parameters. */ ui: mod.flow.ui.ImageFace
}
/** Render an image constrained by a control image. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.Control = mod.flow.Gen{
kind: @gen
domain: "control"
type_name: "Control"
/** 1..200; the control backend default is 30. */ steps: 30
/** 0..2000; Canny default is 50. */ canny_low: 50.0
/** 0..2000; Canny default is 200. */ canny_high: 200.0
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects depth or Canny control by hub affinity. */ model: ""
ports: { in: {prompt: @text, control: @image}, out: {image: @image} }
image: fn() { self.out(@image) }
/** Image preview with generic control parameters. */ ui: mod.flow.ui.ImageFace
}
/** Fill or extend a masked region of an image. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.Inpaint = mod.flow.Gen{
kind: @gen
domain: "inpaint"
type_name: "Inpaint"
/** 1..200; the inpaint backend default is 50. */ steps: 50
/** 0..100; the inpaint backend default is 30. */ guidance: 30.0
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the inpaint domain. */ model: ""
ports: {
in: {
prompt: @text
image: @image
/** White marks the region to regenerate; black is kept. */ mask: @image
}
out: { image: @image }
}
image: fn() { self.out(@image) }
/** Image preview with generic inpaint parameters. */ ui: mod.flow.ui.ImageFace
}
/** Extract a soft alpha matte from an image. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.Matte = mod.flow.Gen{
kind: @gen
domain: "matte"
type_name: "Matte"
/** Empty selects the hub's best model for the matte domain. */ model: ""
ports: { in: {image: @image}, out: {image: @image} }
image: fn() { self.out(@image) }
/** Image preview with generic matte parameters. */ ui: mod.flow.ui.ImageFace
}
/** Estimate a metric depth map from an image. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.Depth = mod.flow.Gen{
kind: @gen
domain: "depth"
type_name: "Depth"
/** Empty selects the hub's best model for the depth domain. */ model: ""
ports: { in: {image: @image}, out: {image: @image} }
image: fn() { self.out(@image) }
/** Image preview with generic depth parameters. */ ui: mod.flow.ui.ImageFace
}
/** Generate a video, optionally from a first image keyframe. Default face: `mod.flow.ui.VideoFace`. */
mod.flow.Video = mod.flow.Gen{
kind: @gen
domain: "video"
type_name: "Video"
/** 640..960 step 32 paired sizes: 640x352, 864x480, 960x544 */ width: 640
/** 352..544 step 32 paired sizes: 640x352, 864x480, 960x544 */ height: 352
/** 5..4096 step 1. Requested native frames; H3 rounds up to 17n+5. Larger clips require more VRAM. */ frames: 39
/** Denoising steps. */ steps: 30
/** one of: h264, h265, hevc */ codec: "h264"
/** Decode and mux the jointly generated audio track. */ audio: true
/** one of: 1, 2, 4 */ interpolate: 1
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the video domain. */ model: ""
ports: {
in: {
prompt: @text
image: @image
/** Optional last-frame keyframe, for a closed loop back to `image`. */ last_frame: @image
}
out: { video: @video }
}
video: fn() { self.out(@video) }
/** Video settings with an editable frame count. */ ui: mod.flow.ui.VideoFace
}
/** Upscale, interpolate, or add motion metadata to a video. Default face: `mod.flow.ui.GenFace`. */
mod.flow.VideoEnhance = mod.flow.Gen{
kind: @gen
domain: "enhance"
type_name: "VideoEnhance"
/** one of: 1, 2, 4 */ upscale: 2
/** one of: 1, 2, 4 */ interpolate: 2
/** Append the `mkfl` motion-field sidecar. */ flow_map: true
/** Empty selects the hub's best model for the enhance domain. */ model: ""
ports: { in: {video: @video}, out: {video: @video} }
video: fn() { self.out(@video) }
/** Generic parameter and video-result face. */ ui: mod.flow.ui.GenFace
}
/** Generate a sound effect from a description. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Sfx = mod.flow.Gen{
kind: @gen
domain: "audio"
type_name: "Sfx"
/** 0.5..120; the current creator default is 4 seconds. */ seconds: 4.0
/** 1..200; the SA3 default is 8. */ steps: 8
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the audio domain. */ model: ""
ports: { in: {prompt: @text}, out: {audio: @audio} }
audio: fn() { self.out(@audio) }
/** Generic parameter and audio-result face. */ ui: mod.flow.ui.GenFace
}
/** Generate a song from a production brief and optional structured lyrics and reference audio. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Music = mod.flow.Gen{
kind: @gen
domain: "music"
type_name: "Music"
/** one of: 60, 120, 180, 240, 300; accepted range 5..300 */ seconds: 180
/** 0..1; 0.8 matches the backend's default reference cadence. */ strength: 0.8
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the music domain. */ model: ""
ports: {
in: {
prompt: @text
/** Optional section-tagged lyrics; empty renders instrumental. */ lyrics: @text
audio: @audio
}
out: { audio: @audio }
}
audio: fn() { self.out(@audio) }
/** Generic parameter and audio-result face. */ ui: mod.flow.ui.GenFace
}
/** Speak text, optionally cloning a reference-audio voice. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Speech = mod.flow.Gen{
kind: @gen
domain: "speech"
type_name: "Speech"
/** Empty uses the selected backend's default voice. */ voice: ""
/** 0.25..4 */ speed: 1.0
/** Empty uses the model's default language. */ language: ""
/** Eight values in 0..1.2; empty is neutral/reference affect. */ emotion: []
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the speech domain. */ model: ""
ports: { in: {text: @text, audio: @audio}, out: {audio: @audio} }
audio: fn() { self.out(@audio) }
/** Generic parameter and audio-result face. */ ui: mod.flow.ui.GenFace
}
/** Reconstruct a GLB mesh from a reference image. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Mesh = mod.flow.Gen{
kind: @gen
domain: "mesh"
type_name: "Mesh"
/** 0 for raw decode, otherwise 16..512; creator chains use 512. */ remesh_resolution: 512
/** Keep TRELLIS's own baked texture. */ texture: true
/** one of: 12000, 20000, 40000, 80000, 160000; creator object default is 12000. */ decimation_target: 12000
/** one of: 1024, 2048, 4096 */ texture_size: 1024
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the mesh domain. */ model: ""
ports: { in: {prompt: @text, image: @image}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Paint PBR materials onto a mesh from a reference image. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Paint = mod.flow.Gen{
kind: @gen
domain: "paint"
type_name: "Paint"
/** one of: 1024, 2048, 4096 */ texture_size: 1024
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the paint domain. */ model: ""
ports: { in: {prompt: @text, mesh: @mesh, reference_image: @image}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Skin and rig a GLB mesh. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Rig = mod.flow.Gen{
kind: @gen
domain: "rig"
type_name: "Rig"
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the rig domain. */ model: ""
ports: { in: {prompt: @text, mesh: @mesh}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Animate a rigged GLB mesh. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Motion = mod.flow.Gen{
kind: @gen
domain: "motion"
type_name: "Motion"
/** one of: playable, prompt */ motion_mode: "playable"
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the motion domain. */ model: ""
ports: { in: {prompt: @text, mesh: @mesh}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Reconstruct one object as a Gaussian-splat PLY. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Splat = mod.flow.Gen{
kind: @gen
domain: "splat"
type_name: "Splat"
/** 1..200; the TripoSplat default is 20. */ steps: 20
/** The TripoSplat default is 3. */ guidance: 3.0
/** 32768..262144 step 32 */ gaussians: 262144
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the splat domain. */ model: ""
ports: { in: {prompt: @text, image: @image}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Generate a walkable world as a Gaussian-splat PLY. Default face: `mod.flow.ui.GenFace`. */
mod.flow.World = mod.flow.Gen{
kind: @gen
domain: "world"
type_name: "World"
/** 0 is the deterministic default seed. */ seed: 0
/** Empty selects the hub's best model for the world domain. */ model: ""
ports: { in: {prompt: @text, image: @image}, out: {mesh: @mesh} }
mesh: fn() { self.out(@mesh) }
/** Generic parameter and mesh-result face. */ ui: mod.flow.ui.GenFace
}
/** Recover structured human pose data from an image. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Body = mod.flow.Gen{
kind: @gen
domain: "body"
type_name: "Body"
/** `hands`, `detect`, and `persons=N` may be combined. */ prompt: ""
/** Empty selects the hub's best model for the body domain. */ model: ""
ports: { in: {image: @image}, out: {json: @json} }
json: fn() { self.out(@json) }
/** Generic parameter and structured-result face. */ ui: mod.flow.ui.GenFace
}
/** Segment one or more prompted subjects from an image. Default face: `mod.flow.ui.ImageFace`. */
mod.flow.Segment = mod.flow.Gen{
kind: @gen
domain: "segment"
type_name: "Segment"
/** Multiplex prompt such as `person:1` or `cat:2`. */ prompt: "person:1"
/** Empty selects the hub's best model for the segment domain. */ model: ""
ports: { in: {image: @image}, out: {mask: @image, cutout: @image} }
mask: fn() { self.out(@mask) }
cutout: fn() { self.out(@cutout) }
/** Image preview with generic segmentation parameters. */ ui: mod.flow.ui.ImageFace
}
/** Transcribe speech audio into a timed JSON transcript. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Stt = mod.flow.Gen{
kind: @gen
domain: "stt"
type_name: "Stt"
/** Optional BCP-47 language hint; empty enables automatic detection. */ language: ""
/** Empty selects the hub's best model for speech transcription. */ model: ""
ports: { in: {audio: @audio}, out: {json: @json} }
json: fn() { self.out(@json) }
/** Generic parameter and transcript-result face. */ ui: mod.flow.ui.GenFace
}
/** Detect beats and downbeats in audio. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Beats = mod.flow.Gen{
kind: @gen
domain: "beats"
type_name: "Beats"
/** Empty selects the hub's best beat-analysis model. */ model: ""
ports: { in: {audio: @audio}, out: {json: @json} }
json: fn() { self.out(@json) }
/** Generic parameter and analysis-result face. */ ui: mod.flow.ui.GenFace
}
/** Separate stereo music into drums, bass, other, and vocals. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Stems = mod.flow.Gen{
kind: @gen
domain: "stems"
type_name: "Stems"
/** Empty selects the hub's best source-separation model. */ model: ""
ports: { in: {audio: @audio}, out: {stems: @bytes} }
stems: fn() { self.out(@stems) }
/** Generic parameter and stem-bundle face. */ ui: mod.flow.ui.GenFace
}
/** Transcribe polyphonic notes to JSON and MIDI. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Notes = mod.flow.Gen{
kind: @gen
domain: "notes"
type_name: "Notes"
/** Empty selects the hub's best note-transcription model. */ model: ""
ports: { in: {audio: @audio}, out: {json: @json, midi: @bytes} }
json: fn() { self.out(@json) }
midi: fn() { self.out(@midi) }
/** Generic parameter and transcription-result face. */ ui: mod.flow.ui.GenFace
}
/** Apply a vision answer to an existing asset's annotation record. Default face: `mod.flow.ui.GenFace`. */
mod.flow.Annotate = mod.flow.Gen{
kind: @gen
domain: "vision"
type_name: "Annotate"
/** The current nine-line annotation record needs about 200 tokens. */ max_tokens: 200
/** Empty selects the hub's best model for the vision domain. */ model: ""
ports: { in: {prompt: @text, image: @image}, out: {json: @json} }
json: fn() { self.out(@json) }
/** Generic parameter and annotation-receipt face. */ ui: mod.flow.ui.GenFace
}
mod.flow.freeze_module()
Reference in a new issue View git blame Copy permalink