Fabio
c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
145 lines
6.6 KiB
C++
145 lines
6.6 KiB
C++
#pragma once
|
|
|
|
#include "config.h"
|
|
#include "http.h"
|
|
#include "multipart.h"
|
|
|
|
#include "engine/framework/io/json.h"
|
|
#include "engine/framework/runtime/session.h"
|
|
|
|
#include <chrono>
|
|
#include <filesystem>
|
|
#include <functional>
|
|
#include <optional>
|
|
#include <sstream>
|
|
#include <string>
|
|
#include <string_view>
|
|
#include <unordered_map>
|
|
#include <vector>
|
|
|
|
namespace minitts::server {
|
|
|
|
using ServerClock = std::chrono::steady_clock;
|
|
|
|
std::optional<int>
|
|
parse_busy_timeout_override(const engine::io::json::Value &body);
|
|
bool model_accepts_request_option(
|
|
std::string_view family, std::string_view option,
|
|
const std::optional<std::filesystem::path> &model_spec_override,
|
|
const std::filesystem::path &model_path);
|
|
std::string model_request_params_json(
|
|
std::string_view family,
|
|
const std::optional<std::filesystem::path> &model_spec_override,
|
|
const std::filesystem::path &model_path);
|
|
std::string json_quote(std::string_view value);
|
|
std::filesystem::path resolve_path(const std::filesystem::path &base,
|
|
const std::filesystem::path &path);
|
|
std::optional<std::string>
|
|
load_voice_library_text(const std::filesystem::path &voice_dir,
|
|
const std::string &voice_id);
|
|
std::optional<std::filesystem::path>
|
|
resolve_voice_library_wav(const std::filesystem::path &voice_dir,
|
|
const std::string &voice_id);
|
|
std::unordered_map<std::string, std::string>
|
|
options_from_object(const engine::io::json::Value *value);
|
|
|
|
struct RuntimeVoicePreset {
|
|
std::optional<std::string> voice_id;
|
|
std::optional<engine::runtime::AudioBuffer> audio;
|
|
std::optional<std::string> reference_text;
|
|
std::unordered_map<std::string, std::string> options;
|
|
};
|
|
|
|
struct SpeechRequestContext {
|
|
const std::filesystem::path &request_base;
|
|
const std::optional<std::filesystem::path> &voice_dir;
|
|
const std::unordered_map<std::string, RuntimeVoicePreset> &voice_presets;
|
|
const std::optional<RuntimeVoicePreset> &default_voice_preset;
|
|
bool accepts_reference_text = true;
|
|
bool accepts_speed = true;
|
|
bool accepts_speaking_rate = true;
|
|
const std::unordered_map<std::string, std::string> &default_options;
|
|
};
|
|
|
|
engine::runtime::TaskRequest build_speech_request(
|
|
const engine::io::json::Value &body,
|
|
const SpeechRequestContext &context);
|
|
engine::runtime::TaskRequest apply_default_request_options(
|
|
const std::unordered_map<std::string, std::string> &default_options,
|
|
engine::runtime::TaskRequest request);
|
|
std::string safe_upload_name(std::string value);
|
|
ServerModelConfig
|
|
model_config_from_json(const engine::io::json::Value &body,
|
|
const std::filesystem::path &request_base, bool lazy);
|
|
std::string query_param(const std::string &query, const std::string &key);
|
|
std::string decoded_query_param(const std::string &query,
|
|
const std::string &key);
|
|
const char *backend_name(engine::core::BackendType type);
|
|
HttpResponse ui_service_worker_retirement_response();
|
|
std::string
|
|
options_json(const std::unordered_map<std::string, std::string> &options);
|
|
void add_option_from_json(std::unordered_map<std::string, std::string> &options,
|
|
const engine::io::json::Value &object,
|
|
const std::string &field,
|
|
const std::string &option_key);
|
|
std::vector<uint8_t>
|
|
encode_pcm16_wav(const engine::runtime::AudioBuffer &audio);
|
|
std::vector<uint8_t>
|
|
encode_pcm16_samples(const engine::runtime::AudioBuffer &audio);
|
|
void write_sse(HttpStreamWriter &writer, const std::string &json);
|
|
void write_sse_done(HttpStreamWriter &writer);
|
|
bool bool_field(const engine::io::json::Value &object, const std::string &key,
|
|
bool default_value);
|
|
HttpResponse sse_response(std::function<void(HttpStreamWriter &)> stream);
|
|
HttpResponse
|
|
chunked_audio_response(std::function<void(HttpStreamWriter &)> stream);
|
|
bool is_wav_upload_filename(const std::string &filename);
|
|
void log_request_body_if_enabled(const ServerConfig &config,
|
|
const HttpRequest &request);
|
|
void log_multipart_request_summary_if_enabled(
|
|
const ServerConfig &config, const std::vector<MultipartPart> &parts);
|
|
double elapsed_ms(ServerClock::time_point started);
|
|
double audio_duration_ms(const engine::runtime::AudioBuffer &audio);
|
|
double audio_rtf(double wall_ms, double duration_ms);
|
|
std::string timing_json(double wall_ms);
|
|
std::string timing_json(double wall_ms,
|
|
const engine::runtime::AudioBuffer &audio);
|
|
std::string ttft_timing_json(double ttft_ms);
|
|
std::string live_speech_timing_json(double request_start_to_first_audio_ms,
|
|
double input_end_ms);
|
|
bool stream_event_has_output(const engine::runtime::StreamEvent &event, bool include_artifacts = false);
|
|
bool task_result_has_output(const engine::runtime::TaskResult &result);
|
|
double require_ttft_ms(const std::optional<double> &ttft_ms);
|
|
double require_input_end_ms(const std::optional<double> &input_end_ms);
|
|
std::unordered_map<std::string, std::string>
|
|
timing_headers(double wall_ms, const engine::runtime::AudioBuffer &audio);
|
|
std::string
|
|
diarization_event_json(const std::vector<engine::runtime::SpeakerTurn> &turns,
|
|
int sample_rate, bool final,
|
|
const std::optional<double> &ttft_ms = std::nullopt);
|
|
void write_transcript_detail_fields(std::ostringstream &out,
|
|
const engine::runtime::TaskResult &result);
|
|
std::string
|
|
task_result_json_with_timing(const engine::runtime::TaskResult &result,
|
|
const std::string &timing);
|
|
std::string task_result_json(const engine::runtime::TaskResult &result,
|
|
double wall_ms);
|
|
std::string alignment_result_json(const engine::runtime::TaskResult &result,
|
|
const engine::runtime::AudioBuffer &audio,
|
|
double wall_ms);
|
|
std::string
|
|
streaming_task_result_json(const engine::runtime::TaskResult &result,
|
|
const std::optional<double> &ttft_ms);
|
|
std::string stream_event_json(const engine::runtime::StreamEvent &event,
|
|
bool diarization, bool include_artifacts = false);
|
|
const engine::runtime::AudioBuffer &
|
|
select_audio_output(const engine::runtime::TaskResult &result);
|
|
engine::runtime::TaskRequest build_openai_transcription_request(
|
|
const engine::io::json::Value &body, const std::filesystem::path &base_dir,
|
|
bool accepts_language_option,
|
|
const std::string *uploaded_audio_bytes = nullptr);
|
|
std::optional<std::filesystem::path>
|
|
find_repository_root(const std::filesystem::path &request_base,
|
|
const std::filesystem::path &resource_anchor);
|
|
|
|
} // namespace minitts::server
|