feat: audio surfaces (TTS + transcription), imagegen.Editor, llamaswap health probe

- New leaf package `audio` (ADR-0017): SpeechModel/SpeechProvider and
  TranscriptionModel/TranscriptionProvider with imagegen conventions
  (zero value = backend default, functional options + Apply, bytes
  in/out, never URLs). Root re-exports added.
- imagegen.Editor (ADR-0018): optional image-to-image interface —
  EditRequest carries the generation knobs plus Init image and
  denoising Strength; separate interface so existing Models keep
  compiling.
- provider/llamaswap implements all of it: POST /v1/audio/speech (JSON,
  raw-audio response, MIME from Content-Type with format fallback),
  POST /v1/audio/transcriptions (multipart, response_format=json),
  ListVoices (GET /v1/audio/voices?model=, tolerant of string-list and
  object-list shapes), POST /sdapi/v1/img2img (txt2img wire +
  init_images/denoising_strength, shared image decode), and Health(ctx)
  (GET /health) — a cheap liveness probe for often-offline hosts.
- Hermetic httptest coverage for every new wire shape and validation
  path; README sections + support-matrix footnote updated in the same
  commit (also corrects the stale /v1/images/generations claim — the
  image path has been SDAPI since the seed fix).

First consumer: mort's llamaswap media tool cluster (status / image /
TTS / STT agent tools against the netherstorm host).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXQxVhXBw8PwFAtsVrXSmj

imagegen/edit.go

@@ -0,0 +1,91 @@
+package imagegen
+
+import "context"
+
+// EditRequest is an image-to-image (edit) request: a prompt applied to an
+// initial image. As with Request, zero values mean "backend default"
+// (ADR-0018).
+type EditRequest struct {
+	// Prompt is the text description of the desired edit.
+	Prompt string
+
+	// Init is the initial image the edit starts from. Required.
+	Init Image
+
+	// Strength is the denoising strength in [0,1] — how far the result may
+	// depart from Init (0 = return the input, 1 = ignore it); nil = backend
+	// default.
+	Strength *float64
+
+	// N is the number of images to generate; 0 = provider default.
+	N int
+
+	// Size is the requested resolution, e.g. "1024x1024"; "" = provider
+	// default (usually the init image's own resolution).
+	Size string
+
+	// Steps is the number of diffusion steps; nil = backend default.
+	Steps *int
+
+	// CFGScale is the classifier-free-guidance scale; nil = backend default.
+	CFGScale *float64
+
+	// NegativePrompt steers generation away from concepts; "" = none.
+	NegativePrompt string
+
+	// Sampler selects the sampling method (e.g. "euler", "euler_a");
+	// "" = backend default.
+	Sampler string
+
+	// Seed fixes the RNG seed for reproducible output; nil = random.
+	Seed *int64
+}
+
+// EditOption mutates an EditRequest before it is sent. Options passed to Edit
+// are applied to a copy of the request, so an EditRequest value can be reused.
+type EditOption func(*EditRequest)
+
+// WithEditStrength sets the denoising strength in [0,1].
+func WithEditStrength(s float64) EditOption { return func(r *EditRequest) { r.Strength = &s } }
+
+// WithEditN sets the number of images to generate.
+func WithEditN(n int) EditOption { return func(r *EditRequest) { r.N = n } }
+
+// WithEditSize sets the requested resolution (e.g. "1024x1024").
+func WithEditSize(size string) EditOption { return func(r *EditRequest) { r.Size = size } }
+
+// WithEditSteps overrides the number of diffusion steps.
+func WithEditSteps(n int) EditOption { return func(r *EditRequest) { r.Steps = &n } }
+
+// WithEditCFGScale overrides the classifier-free-guidance scale.
+func WithEditCFGScale(s float64) EditOption { return func(r *EditRequest) { r.CFGScale = &s } }
+
+// WithEditNegativePrompt sets a negative prompt.
+func WithEditNegativePrompt(s string) EditOption {
+	return func(r *EditRequest) { r.NegativePrompt = s }
+}
+
+// WithEditSampler overrides the sampling method.
+func WithEditSampler(s string) EditOption { return func(r *EditRequest) { r.Sampler = s } }
+
+// WithEditSeed fixes the RNG seed for reproducible output.
+func WithEditSeed(seed int64) EditOption { return func(r *EditRequest) { r.Seed = &seed } }
+
+// Apply returns a copy of the request with all options applied. Providers
+// call this once at the top of Edit.
+func (r EditRequest) Apply(opts ...EditOption) EditRequest {
+	for _, opt := range opts {
+		opt(&r)
+	}
+	return r
+}
+
+// Editor is the image-to-image surface. It is a separate, optional interface
+// rather than a method on Model so existing Model implementations keep
+// compiling; callers type-assert (`m.(imagegen.Editor)`) or require it
+// explicitly.
+type Editor interface {
+	// Edit produces one or more images derived from the request's init image
+	// under the request's prompt.
+	Edit(ctx context.Context, req EditRequest, opts ...EditOption) (*Result, error)
+}