feat(videogen): canonical video-generation surface + llama-swap client
New videogen/ contract package (ADR-0019): Request/Result/Model/Provider
with the imagegen conventions. Text-to-video and image-to-video are one
surface (Request.InitImage, nil = t2v) since hybrid checkpoints like
Wan 2.2 TI2V serve both from one model; Result carries a single clip.
provider/llamaswap gains VideoModel(id) targeting the blocking
POST {base}/v1/videos/sync (multipart, model-routed by the fork's new
video routes): vLLM-Omni parameter names, OpenAI-style input_reference
file part, optional fields stay off the wire so per-model launch-flag
defaults apply. CLAUDE.md package map picks up audio/ (missed in #12)
and videogen/.
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01AXQxVhXBw8PwFAtsVrXSmj
This commit is contained in:
@@ -0,0 +1,177 @@
|
||||
package llamaswap
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"mime"
|
||||
"mime/multipart"
|
||||
"net/http"
|
||||
"strconv"
|
||||
"strings"
|
||||
|
||||
"gitea.stevedudenhoeffer.com/steve/majordomo/llm"
|
||||
"gitea.stevedudenhoeffer.com/steve/majordomo/videogen"
|
||||
)
|
||||
|
||||
// VideoModel implements videogen.Provider, binding a video-generation model
|
||||
// served by llama-swap (routed to a vLLM-Omni-style upstream, or any shim
|
||||
// exposing the same /v1/videos/sync shape). The id is passed through verbatim
|
||||
// and selects which upstream llama-swap loads.
|
||||
func (p *Provider) VideoModel(id string, opts ...videogen.ModelOption) (videogen.Model, error) {
|
||||
if err := p.requireBaseURL(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
_ = videogen.ApplyModelOptions(opts)
|
||||
return &videoModel{p: p, id: id}, nil
|
||||
}
|
||||
|
||||
type videoModel struct {
|
||||
p *Provider
|
||||
id string
|
||||
}
|
||||
|
||||
// Generate implements videogen.Model via POST {base}/v1/videos/sync
|
||||
// (multipart/form-data — llama-swap routes by the `model` form field). The
|
||||
// blocking sync endpoint answers with the encoded video itself, so the
|
||||
// response body is the result; there is no job id to poll. Generation runs
|
||||
// for minutes — the provider client carries no timeout by design, and callers
|
||||
// bound the call with a context deadline.
|
||||
//
|
||||
// Parameter names follow vLLM-Omni's videos API (num_frames, fps,
|
||||
// num_inference_steps, guidance_scale); the conditioning frame is sent as an
|
||||
// `input_reference` file part, following OpenAI's videos API. Upstreams
|
||||
// ignore fields they don't understand, and optional fields stay off the wire
|
||||
// entirely so the model's own defaults apply.
|
||||
func (m *videoModel) Generate(ctx context.Context, req videogen.Request, opts ...videogen.Option) (*videogen.Result, error) {
|
||||
req = req.Apply(opts...)
|
||||
if strings.TrimSpace(req.Prompt) == "" {
|
||||
return nil, fmt.Errorf("%w: video generation requires a prompt", llm.ErrUnsupported)
|
||||
}
|
||||
if req.NumFrames < 0 {
|
||||
return nil, fmt.Errorf("%w: video frame count must be >= 0, got %d", llm.ErrUnsupported, req.NumFrames)
|
||||
}
|
||||
if req.FPS < 0 {
|
||||
return nil, fmt.Errorf("%w: video fps must be >= 0, got %d", llm.ErrUnsupported, req.FPS)
|
||||
}
|
||||
if req.InitImage != nil && len(req.InitImage.Data) == 0 {
|
||||
return nil, fmt.Errorf("%w: video init image has no bytes", llm.ErrUnsupported)
|
||||
}
|
||||
width, height, err := parseSize(req.Size)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("%w: %v", llm.ErrUnsupported, err)
|
||||
}
|
||||
|
||||
var buf bytes.Buffer
|
||||
w := multipart.NewWriter(&buf)
|
||||
// Required fields always go on the wire (an empty model id should fail
|
||||
// loudly upstream, not silently vanish); optional ones only when set.
|
||||
fields := []struct {
|
||||
key, value string
|
||||
required bool
|
||||
}{
|
||||
{"model", m.id, true},
|
||||
{"prompt", req.Prompt, true},
|
||||
{"negative_prompt", req.NegativePrompt, false},
|
||||
{"width", formatInt(width), false},
|
||||
{"height", formatInt(height), false},
|
||||
{"num_frames", formatNonZero(req.NumFrames), false},
|
||||
{"fps", formatNonZero(req.FPS), false},
|
||||
{"num_inference_steps", formatInt(req.Steps), false},
|
||||
{"guidance_scale", formatFloat(req.GuidanceScale), false},
|
||||
{"seed", formatInt64(req.Seed), false},
|
||||
}
|
||||
for _, f := range fields {
|
||||
if !f.required && f.value == "" {
|
||||
continue
|
||||
}
|
||||
if err := w.WriteField(f.key, f.value); err != nil {
|
||||
return nil, fmt.Errorf("llama-swap: build video form: %w", err)
|
||||
}
|
||||
}
|
||||
if req.InitImage != nil {
|
||||
fw, err := w.CreateFormFile("input_reference", initImageFilename(req.InitImage.MIME))
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("llama-swap: build video form: %w", err)
|
||||
}
|
||||
if _, err := fw.Write(req.InitImage.Data); err != nil {
|
||||
return nil, fmt.Errorf("llama-swap: build video form: %w", err)
|
||||
}
|
||||
}
|
||||
if err := w.Close(); err != nil {
|
||||
return nil, fmt.Errorf("llama-swap: build video form: %w", err)
|
||||
}
|
||||
|
||||
videoBytes, contentType, err := m.p.doRaw(ctx, http.MethodPost, "/v1/videos/sync", m.id, w.FormDataContentType(), &buf)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if len(videoBytes) == 0 {
|
||||
return nil, &llm.APIError{Provider: m.p.name, Model: m.id, Message: "video response contained no video"}
|
||||
}
|
||||
return &videogen.Result{Video: videogen.Video{Data: videoBytes, MIME: videoMIME(contentType, videoBytes)}}, nil
|
||||
}
|
||||
|
||||
// videoMIME resolves the result MIME type: the response Content-Type when it
|
||||
// is a concrete video type, else content sniffing, else video/mp4 (the
|
||||
// endpoint's default container).
|
||||
func videoMIME(contentType string, data []byte) string {
|
||||
if mt, _, err := mime.ParseMediaType(contentType); err == nil && strings.HasPrefix(mt, "video/") {
|
||||
return mt
|
||||
}
|
||||
if mt := http.DetectContentType(data); strings.HasPrefix(mt, "video/") {
|
||||
return mt
|
||||
}
|
||||
return "video/mp4"
|
||||
}
|
||||
|
||||
// initImageFilename picks the multipart filename hint for the conditioning
|
||||
// frame from its MIME subtype. The name is provider-chosen (never
|
||||
// caller-supplied), so no sanitization is needed.
|
||||
func initImageFilename(mimeType string) string {
|
||||
mt := strings.ToLower(strings.TrimSpace(mimeType))
|
||||
if parsed, _, err := mime.ParseMediaType(mt); err == nil {
|
||||
mt = parsed
|
||||
}
|
||||
switch mt {
|
||||
case "image/jpeg", "image/jpg":
|
||||
return "frame.jpg"
|
||||
case "image/webp":
|
||||
return "frame.webp"
|
||||
default: // stable-diffusion outputs and unknowns: PNG is the safe hint
|
||||
return "frame.png"
|
||||
}
|
||||
}
|
||||
|
||||
// formatInt renders an optional int pointer for a form field; nil = "" (omit).
|
||||
func formatInt(v *int) string {
|
||||
if v == nil {
|
||||
return ""
|
||||
}
|
||||
return strconv.Itoa(*v)
|
||||
}
|
||||
|
||||
// formatInt64 renders an optional int64 pointer for a form field; nil = "" (omit).
|
||||
func formatInt64(v *int64) string {
|
||||
if v == nil {
|
||||
return ""
|
||||
}
|
||||
return strconv.FormatInt(*v, 10)
|
||||
}
|
||||
|
||||
// formatFloat renders an optional float pointer for a form field; nil = "" (omit).
|
||||
func formatFloat(v *float64) string {
|
||||
if v == nil {
|
||||
return ""
|
||||
}
|
||||
return strconv.FormatFloat(*v, 'g', -1, 64)
|
||||
}
|
||||
|
||||
// formatNonZero renders a non-negative int for a form field; 0 = "" (omit,
|
||||
// backend default).
|
||||
func formatNonZero(v int) string {
|
||||
if v == 0 {
|
||||
return ""
|
||||
}
|
||||
return strconv.Itoa(v)
|
||||
}
|
||||
Reference in New Issue
Block a user