feat: media expansion surfaces — edit mask, upscale, background removal, interpolation, diarization, meshgen (ADR-0020)

- imagegen.EditRequest.Mask -> sd-server img2img inpainting (white=repaint)
- imagegen.Upscaler + BackgroundRemover, videogen.Interpolator,
  audio.DiarizationModel: new optional provider-minted surfaces
- NEW meshgen leaf package (image->3D, glb/stl/obj)
- provider/llamaswap: all five via the /upstream/<model>/<path> passthrough
  (upstreamPath helper, shared one-file multipart builder); binary success
  bodies validated (non-image, non-video, JSON-mesh rejection); diarization
  pins output=json (vtt/srt drop speaker labels)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

meshgen/meshgen.go

@@ -0,0 +1,152 @@
+// Package meshgen is majordomo's canonical image-to-3D surface. Like
+// imagegen/audio/videogen, it is a deliberately separate leaf contract from
+// the llm package: mesh generation shares none of the chat machinery, so it
+// gets its own small Provider/Model interfaces (ADR-0020, following the
+// ADR-0016→0019 lineage: functional options, zero values = backend default,
+// bytes-only I/O, Raw escape hatch).
+//
+// The first implementation is provider/llamaswap, which targets a
+// Hunyuan3D-2.1-style api_server reached through the /upstream passthrough.
+// The surface is image-to-3D only: text-to-3D is the composition
+// imagegen.Generate → meshgen.Generate, owned by the caller.
+package meshgen
+
+import (
+	"context"
+
+	"gitea.stevedudenhoeffer.com/steve/majordomo/llm"
+)
+
+// Image is the conditioning input (bytes + MIME), aliased to llm.ImagePart so
+// chat- or imagegen-sourced images feed mesh generation without conversion.
+type Image = llm.ImagePart
+
+// Mesh is one generated 3D asset: raw encoded bytes plus its format.
+type Mesh struct {
+	// Data is the encoded mesh (binary GLB/STL/OBJ container).
+	Data []byte
+
+	// Format is the lowercase container name ("glb", "stl", "obj").
+	Format string
+
+	// MIME is the corresponding MIME type, e.g. "model/gltf-binary",
+	// "model/stl".
+	MIME string
+}
+
+// Request is an image-to-3D generation request. Zero values mean "backend
+// default".
+type Request struct {
+	// Image is the source image the mesh is reconstructed from. Required.
+	Image Image
+
+	// Format is the requested output container ("glb", "stl", "obj");
+	// "" = backend default (glb).
+	Format string
+
+	// Texture asks the backend to also synthesize surface textures. Much
+	// slower and much more VRAM-hungry than shape-only on the reference
+	// backend; leave false for print-pipeline geometry.
+	Texture bool
+
+	// RemoveBackground lets the backend cut the subject out first; nil =
+	// backend default (true on Hunyuan3D). Set explicitly to false when the
+	// input is already a clean cutout.
+	RemoveBackground *bool
+
+	// OctreeResolution is the shape-decoder grid resolution; 0 = backend
+	// default.
+	OctreeResolution int
+
+	// Steps is the number of diffusion steps; nil = backend default.
+	Steps *int
+
+	// GuidanceScale is the guidance strength; nil = backend default.
+	GuidanceScale *float64
+
+	// Seed fixes the RNG seed for reproducible output; nil = backend default.
+	Seed *int64
+
+	// FaceCount caps the output mesh's face count; 0 = backend default.
+	FaceCount int
+}
+
+// Result is the canonical mesh-generation result.
+type Result struct {
+	// Mesh is the generated asset.
+	Mesh Mesh
+
+	// Raw is the provider-native response object. May be nil.
+	Raw any
+}
+
+// Option mutates a Request before it is sent. Options passed to Generate are
+// applied to a copy of the request, so a Request value can be reused.
+type Option func(*Request)
+
+// WithFormat sets the output container ("glb", "stl", "obj").
+func WithFormat(f string) Option { return func(r *Request) { r.Format = f } }
+
+// WithTexture enables texture synthesis.
+func WithTexture() Option { return func(r *Request) { r.Texture = true } }
+
+// WithRemoveBackground sets the backend's background-removal preprocessing.
+func WithRemoveBackground(v bool) Option {
+	return func(r *Request) { r.RemoveBackground = &v }
+}
+
+// WithOctreeResolution sets the shape-decoder grid resolution.
+func WithOctreeResolution(n int) Option { return func(r *Request) { r.OctreeResolution = n } }
+
+// WithSteps overrides the number of diffusion steps.
+func WithSteps(n int) Option { return func(r *Request) { r.Steps = &n } }
+
+// WithGuidanceScale overrides the guidance strength.
+func WithGuidanceScale(s float64) Option { return func(r *Request) { r.GuidanceScale = &s } }
+
+// WithSeed fixes the RNG seed.
+func WithSeed(seed int64) Option { return func(r *Request) { r.Seed = &seed } }
+
+// WithFaceCount caps the output mesh's face count.
+func WithFaceCount(n int) Option { return func(r *Request) { r.FaceCount = n } }
+
+// Apply returns a copy of the request with all options applied. Providers
+// call this once at the top of Generate.
+func (r Request) Apply(opts ...Option) Request {
+	for _, opt := range opts {
+		opt(&r)
+	}
+	return r
+}
+
+// Model generates 3D meshes from images.
+type Model interface {
+	// Generate reconstructs a mesh from the request's image.
+	Generate(ctx context.Context, req Request, opts ...Option) (*Result, error)
+}
+
+// ModelOption configures a Model at construction time. Reserved for future
+// per-model settings.
+type ModelOption func(*ModelConfig)
+
+// ModelConfig carries per-model construction settings.
+type ModelConfig struct{}
+
+// ApplyModelOptions folds options into a config.
+func ApplyModelOptions(opts []ModelOption) ModelConfig {
+	var cfg ModelConfig
+	for _, opt := range opts {
+		opt(&cfg)
+	}
+	return cfg
+}
+
+// Provider mints mesh models bound to one backend.
+type Provider interface {
+	// Name is the registry identifier for the provider.
+	Name() string
+
+	// MeshModel returns a Model bound to the given id (passed through to the
+	// backend verbatim; no catalog validation).
+	MeshModel(id string, opts ...ModelOption) (Model, error)
+}