feat(server): implement vid_gen async API and mode-aware capabilities (#1437)

This commit is contained in:
leejet
2026-04-18 15:06:36 +08:00
committed by GitHub
parent f3f69e2fbe
commit 4d626d24b2
14 changed files with 1345 additions and 339 deletions

View File

@@ -45,6 +45,44 @@ std::string normalize_output_format(std::string output_format) {
return output_format;
}
std::vector<std::string> supported_img_output_formats(bool allow_webp) {
std::vector<std::string> formats = {"png", "jpeg"};
#ifdef SD_USE_WEBP
if (allow_webp) {
formats.push_back("webp");
}
#else
(void)allow_webp;
#endif
return formats;
}
std::vector<std::string> supported_vid_output_formats() {
std::vector<std::string> formats;
#ifdef SD_USE_WEBM
formats.push_back("webm");
#endif
#ifdef SD_USE_WEBP
formats.push_back("webp");
#endif
formats.push_back("avi");
return formats;
}
static std::string valid_vid_output_formats_message() {
const std::vector<std::string> formats = supported_vid_output_formats();
std::string message = "invalid output_format, must be one of [";
for (size_t i = 0; i < formats.size(); ++i) {
if (i > 0) {
message += ", ";
}
message += formats[i];
}
message += "]";
return message;
}
bool assign_output_options(ImgGenJobRequest& request,
std::string output_format,
int output_compression,
@@ -53,19 +91,88 @@ bool assign_output_options(ImgGenJobRequest& request,
request.output_format = normalize_output_format(std::move(output_format));
request.output_compression = std::clamp(output_compression, 0, 100);
const bool valid_format = request.output_format == "png" ||
request.output_format == "jpeg" ||
(allow_webp && request.output_format == "webp");
const std::vector<std::string> valid_formats = supported_img_output_formats(allow_webp);
const bool valid_format = std::find(valid_formats.begin(),
valid_formats.end(),
request.output_format) != valid_formats.end();
if (!valid_format) {
error_message = allow_webp
? "invalid output_format, must be one of [png, jpeg, webp]"
: "invalid output_format, must be one of [png, jpeg]";
error_message = "invalid output_format, must be one of [";
for (size_t i = 0; i < valid_formats.size(); ++i) {
if (i > 0) {
error_message += ", ";
}
error_message += valid_formats[i];
}
error_message += "]";
return false;
}
return true;
}
bool assign_output_options(VidGenJobRequest& request,
std::string output_format,
int output_compression,
std::string& error_message) {
request.output_format = normalize_output_format(std::move(output_format));
request.output_compression = std::clamp(output_compression, 0, 100);
if (request.output_format == "avi") {
return true;
}
if (request.output_format == "webm") {
#ifdef SD_USE_WEBM
return true;
#else
error_message = valid_vid_output_formats_message();
return false;
#endif
}
if (request.output_format == "webp") {
#ifdef SD_USE_WEBP
return true;
#else
error_message = valid_vid_output_formats_message();
return false;
#endif
}
error_message = valid_vid_output_formats_message();
return false;
}
std::string video_mime_type(const std::string& output_format) {
if (output_format == "webm") {
return "video/webm";
}
if (output_format == "webp") {
return "image/webp";
}
return "video/x-msvideo";
}
bool runtime_supports_generation_mode(const ServerRuntime& runtime, SDMode mode) {
if (mode == VID_GEN) {
return sd_ctx_supports_video_generation(runtime.sd_ctx);
}
if (mode == IMG_GEN) {
return sd_ctx_supports_image_generation(runtime.sd_ctx);
}
return true;
}
std::string unsupported_generation_mode_error(SDMode mode) {
if (mode == VID_GEN) {
return "loaded model does not support vid_gen";
}
if (mode == IMG_GEN) {
return "loaded model does not support img_gen";
}
return "loaded model does not support requested mode";
}
ArgOptions SDSvrParams::get_options() {
ArgOptions options;