add full source: transcribe pipeline, HTTP API, OpenAI endpoint, web UI, Docker

2026-09-09 16:04:10 +08:00
parent 757861f1dd
commit 7010da8855
19 changed files with 2918 additions and 0 deletions
@@ -0,0 +1,197 @@
+// Package config loads all runtime configuration from environment variables.
+//
+// Every knob the transcriber exposes is controlled by an ASR_* environment
+// variable so the same binary can be reconfigured without a rebuild.
+package config
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// Config holds all runtime configuration derived from environment variables.
+type Config struct {
+	// ASR_PASSWORD: shared secret required on every API request. If empty,
+	// auth is disabled and a loud warning is logged (intended for local use).
+	Password string
+	// ASR_ADDR: listen address (default ":8080").
+	Addr string
+	// ASR_DATA_DIR: root for persisted state (default "./data").
+	DataDir string
+	// ASR_MODEL_DIR: where models live / are auto-downloaded
+	// (default "<DataDir>/models").
+	ModelDir string
+	// ASR_NUM_THREADS: ONNX compute threads (default: min(4, NumCPU)).
+	NumThreads int
+	// ASR_LANGUAGE: SenseVoice language hint. "auto" detects per segment.
+	Language string
+	// ASR_USE_ITN: enable inverse text normalization (numbers/punctuation).
+	UseITN bool
+	// ASR_MAX_RETRIES: automatic retry attempts before a task is marked failed.
+	MaxRetries int
+	// ASR_MAX_UPLOAD_MB: per-upload size cap (default 512).
+	MaxUploadBytes int64
+	// ASR_WORKERS: concurrent transcription workers (default 1).
+	Workers int
+	// ASR_MAX_CONCURRENT: cap on simultaneously in-flight synchronous
+	// transcriptions on the OpenAI endpoint. Bounds onnxruntime thread
+	// contention under load. Defaults to ASR_NUM_THREADS. Excess requests
+	// wait for a slot (or get 503 if the client disconnects while waiting).
+	MaxConcurrent int
+
+	// VAD (silero) tuning.
+	VadThreshold          float32
+	VadMinSpeechDuration  float32
+	VadMinSilenceDuration float32
+	VadMaxSpeechDuration  float32
+	VadWindowSize         int
+	VadBufferSeconds      float32
+
+	// ASR_DOWNLOAD_TIMEOUT: timeout for the initial model download.
+	DownloadTimeout time.Duration
+	// ASR_FFMPEG: path to the ffmpeg binary used to decode formats beep
+	// can't (m4a/aac/opus/wma …). Defaults to "ffmpeg" on PATH. Only
+	// required when such formats are actually uploaded.
+	FFmpegPath string
+	// ASR_DECODE_TIMEOUT: per-file decode timeout (beep + ffmpeg fallback).
+	DecodeTimeout time.Duration
+}
+
+// Load reads the environment and returns a validated Config.
+func Load() (Config, error) {
+	c := Config{
+		Password:              strings.TrimSpace(os.Getenv("ASR_PASSWORD")),
+		Addr:                  envStr("ASR_ADDR", ":8080"),
+		DataDir:               envStr("ASR_DATA_DIR", "./data"),
+		NumThreads:            envInt("ASR_NUM_THREADS", defaultThreads()),
+		Language:              envStr("ASR_LANGUAGE", "auto"),
+		UseITN:                envBool("ASR_USE_ITN", true),
+		MaxRetries:            envInt("ASR_MAX_RETRIES", 3),
+		MaxUploadBytes:        int64(envInt("ASR_MAX_UPLOAD_MB", 512)) * 1024 * 1024,
+		Workers:               envInt("ASR_WORKERS", 1),
+		MaxConcurrent:         envInt("ASR_MAX_CONCURRENT", 0),
+		VadThreshold:          float32(envFloat("ASR_VAD_THRESHOLD", 0.5)),
+		VadMinSpeechDuration:  float32(envFloat("ASR_VAD_MIN_SPEECH", 0.5)),
+		VadMinSilenceDuration: float32(envFloat("ASR_VAD_MIN_SILENCE", 0.5)),
+		VadMaxSpeechDuration:  float32(envFloat("ASR_VAD_MAX_SPEECH", 20)),
+		VadWindowSize:         envInt("ASR_VAD_WINDOW", 512),
+		VadBufferSeconds:      float32(envFloat("ASR_VAD_BUFFER_SECS", 60)),
+		DownloadTimeout:       time.Duration(envInt("ASR_DOWNLOAD_TIMEOUT", 1800)) * time.Second,
+		FFmpegPath:            envStr("ASR_FFMPEG", "ffmpeg"),
+		DecodeTimeout:         time.Duration(envInt("ASR_DECODE_TIMEOUT", 1800)) * time.Second,
+	}
+	c.ModelDir = envStr("ASR_MODEL_DIR", filepath.Join(c.DataDir, "models"))
+
+	if c.Workers < 1 {
+		c.Workers = 1
+	}
+	if c.MaxConcurrent < 1 {
+		c.MaxConcurrent = c.NumThreads
+	}
+	if c.NumThreads < 1 {
+		c.NumThreads = 1
+	}
+	if c.MaxRetries < 0 {
+		c.MaxRetries = 0
+	}
+	if c.VadMaxSpeechDuration < 1 {
+		c.VadMaxSpeechDuration = 1
+	}
+	if c.DecodeTimeout < time.Minute {
+		c.DecodeTimeout = 30 * time.Minute
+	}
+	if c.VadWindowSize <= 0 {
+		c.VadWindowSize = 512
+	}
+	if c.MaxUploadBytes < 1<<20 {
+		c.MaxUploadBytes = 512 << 20
+	}
+	if c.DownloadTimeout < time.Minute {
+		c.DownloadTimeout = 30 * time.Minute
+	}
+
+	// Absolutize data dir so relative file paths in persisted state survive
+	// regardless of the process working directory at request time.
+	if abs, err := filepath.Abs(c.DataDir); err == nil {
+		c.DataDir = abs
+	}
+	if abs, err := filepath.Abs(c.ModelDir); err == nil {
+		c.ModelDir = abs
+	}
+	return c, nil
+}
+
+// AuthEnabled reports whether API auth is enforced.
+func (c Config) AuthEnabled() bool { return c.Password != "" }
+
+func (c Config) String() string {
+	return fmt.Sprintf("addr=%s data=%s models=%s threads=%d lang=%s itn=%t workers=%d maxconcurrent=%d retries=%d vad{thr=%.2f max=%.0fs win=%d}",
+		c.Addr, c.DataDir, c.ModelDir, c.NumThreads, c.Language, c.UseITN, c.Workers,
+		c.MaxConcurrent, c.MaxRetries, c.VadThreshold, c.VadMaxSpeechDuration, c.VadWindowSize)
+}
+
+func defaultThreads() int {
+	n := runtime.NumCPU()
+	if n > 4 {
+		n = 4
+	}
+	if n < 1 {
+		n = 1
+	}
+	return n
+}
+
+func envStr(key, def string) string {
+	if v := strings.TrimSpace(os.Getenv(key)); v != "" {
+		return v
+	}
+	return def
+}
+
+func envInt(key string, def int) int {
+	v := strings.TrimSpace(os.Getenv(key))
+	if v == "" {
+		return def
+	}
+	n, err := strconv.Atoi(v)
+	if err != nil {
+		return def
+	}
+	return n
+}
+
+func envFloat(key string, def float64) float64 {
+	v := strings.TrimSpace(os.Getenv(key))
+	if v == "" {
+		return def
+	}
+	f, err := strconv.ParseFloat(v, 64)
+	if err != nil {
+		return def
+	}
+	return f
+}
+
+func envBool(key string, def bool) bool {
+	v := strings.TrimSpace(os.Getenv(key))
+	if v == "" {
+		return def
+	}
+	b, err := strconv.ParseBool(v)
+	if err != nil {
+		// Accept "0"/"1" and "yes"/"no" too.
+		switch strings.ToLower(v) {
+		case "yes", "on", "true", "1":
+			return true
+		case "no", "off", "false", "0":
+			return false
+		}
+		return def
+	}
+	return b
+}