Make WhisperX uploads replayable

This commit is contained in:
2026-09-22 00:15:48 +00:00
parent d872edca94
commit dc3288d15a
5 changed files with 634 additions and 175 deletions

View File

@@ -6,14 +6,11 @@ import (
"errors"
"fmt"
"io"
"mime/multipart"
"net"
"net/http"
"net/url"
"os"
"path/filepath"
"strings"
"sync"
"time"
"gitea.maximumdirect.net/eric/narratio/internal/fileops"
@@ -21,7 +18,6 @@ import (
const (
defaultMaxWhisperXResponseBytes int64 = 10 * 1024 * 1024
whisperXUploadBufferSize = 32 * 1024
)
// HTTPClientConfig contains parsed, deterministic WhisperX HTTP client settings.
@@ -192,21 +188,32 @@ func (c *HTTPClient) Transcribe(ctx context.Context, req TranscribeRequest) (Tra
}
func (c *HTTPClient) doTranscribeAttempt(ctx context.Context, audioPath string) (int, []byte, error) {
upload := newMultipartUpload(ctx, audioPath, c.language, c.openAudio)
defer upload.Close()
upload, err := newMultipartBodyFactory(ctx, audioPath, c.language, c.openAudio)
if err != nil {
return 0, nil, fmt.Errorf("prepare whisperx request body: %w", err)
}
defer func() { _ = upload.Close() }()
body, err := upload.NewBody()
if err != nil {
return 0, nil, fmt.Errorf("open whisperx request body: %w", err)
}
req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.url.String(), upload)
req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.url.String(), body)
if err != nil {
return 0, nil, fmt.Errorf("build whisperx request: %w", err)
}
req.Header.Set("Content-Type", upload.contentType)
req.Header.Set("Expect", "100-continue")
req.GetBody = upload.NewBody
resp, err := c.httpClient.Do(req)
if err != nil {
_ = upload.Close()
if producerErr := upload.Wait(); producerErr != nil {
return 0, nil, fmt.Errorf("stream whisperx request body: %w", producerErr)
if sourceErr := upload.SourceError(true); sourceErr != nil {
return 0, nil, fmt.Errorf("stream whisperx request body: %w", sourceErr)
}
if ctxErr := ctx.Err(); ctxErr != nil {
return 0, nil, fmt.Errorf("perform whisperx request: %w", ctxErr)
}
return 0, nil, fmt.Errorf("perform whisperx request: %w", err)
}
@@ -214,8 +221,8 @@ func (c *HTTPClient) doTranscribeAttempt(ctx context.Context, audioPath string)
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
_ = upload.Close()
if producerErr := upload.Wait(); producerErr != nil {
return resp.StatusCode, nil, fmt.Errorf("stream whisperx request body: %w", producerErr)
if sourceErr := upload.SourceError(false); sourceErr != nil {
return resp.StatusCode, nil, fmt.Errorf("stream whisperx request body: %w", sourceErr)
}
if _, err := readWhisperXResponse(resp.Body, c.maxResponseBytes); err != nil {
return resp.StatusCode, nil, fmt.Errorf("read whisperx response body: %w", err)
@@ -223,13 +230,19 @@ func (c *HTTPClient) doTranscribeAttempt(ctx context.Context, audioPath string)
return resp.StatusCode, nil, fmt.Errorf("whisperx returned status %d", resp.StatusCode)
}
if err := upload.Wait(); err != nil {
return resp.StatusCode, nil, fmt.Errorf("stream whisperx request body: %w", err)
data, responseErr := readWhisperXResponse(resp.Body, c.maxResponseBytes)
_ = upload.Close()
if sourceErr := upload.SourceError(false); sourceErr != nil {
if responseErr != nil {
return resp.StatusCode, nil, fmt.Errorf("stream whisperx request body: %w", errors.Join(sourceErr, responseErr))
}
return resp.StatusCode, nil, fmt.Errorf("stream whisperx request body: %w", sourceErr)
}
data, err := readWhisperXResponse(resp.Body, c.maxResponseBytes)
if err != nil {
return resp.StatusCode, nil, fmt.Errorf("read whisperx response body: %w", err)
if err := ctx.Err(); err != nil {
return resp.StatusCode, nil, fmt.Errorf("complete whisperx request: %w", err)
}
if responseErr != nil {
return resp.StatusCode, nil, fmt.Errorf("read whisperx response body: %w", responseErr)
}
return resp.StatusCode, data, nil
}
@@ -286,161 +299,6 @@ func isHTTPURLScheme(scheme string) bool {
}
}
type multipartUpload struct {
reader *io.PipeReader
writer *io.PipeWriter
contentType string
done chan struct{}
mu sync.Mutex
audio io.Closer
err error
aborted bool
}
func newMultipartUpload(ctx context.Context, audioPath, language string, openAudio func(string) (io.ReadCloser, error)) *multipartUpload {
reader, writer := io.Pipe()
multipartWriter := multipart.NewWriter(writer)
upload := &multipartUpload{
reader: reader,
writer: writer,
contentType: multipartWriter.FormDataContentType(),
done: make(chan struct{}),
}
go func() {
err := upload.write(ctx, multipartWriter, audioPath, language, openAudio)
if err != nil {
_ = writer.CloseWithError(err)
} else {
_ = writer.Close()
}
upload.mu.Lock()
upload.err = err
upload.audio = nil
upload.mu.Unlock()
close(upload.done)
}()
go func() {
select {
case <-ctx.Done():
upload.abort()
case <-upload.done:
}
}()
return upload
}
func (u *multipartUpload) Read(p []byte) (int, error) {
return u.reader.Read(p)
}
func (u *multipartUpload) Close() error {
u.abort()
return nil
}
func (u *multipartUpload) Wait() error {
<-u.done
u.mu.Lock()
defer u.mu.Unlock()
return u.err
}
func (u *multipartUpload) write(ctx context.Context, writer *multipart.Writer, audioPath, language string, openAudio func(string) (io.ReadCloser, error)) error {
fileWriter, err := writer.CreateFormFile("file", filepath.Base(audioPath))
if err != nil {
return u.producerError(ctx, fmt.Errorf("create multipart file field: %w", err))
}
audioFile, err := openAudio(audioPath)
if err != nil {
return u.producerError(ctx, fmt.Errorf("open audio file %q: %w", audioPath, err))
}
u.setAudio(audioFile)
_, copyErr := io.CopyBuffer(fileWriter, &contextReader{ctx: ctx, reader: audioFile}, make([]byte, whisperXUploadBufferSize))
closeErr := audioFile.Close()
u.clearAudio(audioFile)
if copyErr != nil {
return u.producerError(ctx, fmt.Errorf("copy audio file %q: %w", audioPath, copyErr))
}
if closeErr != nil {
return u.producerError(ctx, fmt.Errorf("close audio file %q: %w", audioPath, closeErr))
}
if err := writer.WriteField("language", language); err != nil {
return u.producerError(ctx, fmt.Errorf("write language form field: %w", err))
}
if err := writer.Close(); err != nil {
return u.producerError(ctx, fmt.Errorf("close multipart writer: %w", err))
}
return nil
}
func (u *multipartUpload) producerError(ctx context.Context, err error) error {
if ctx.Err() != nil {
return ctx.Err()
}
u.mu.Lock()
aborted := u.aborted
u.mu.Unlock()
if aborted {
return nil
}
return err
}
func (u *multipartUpload) setAudio(audio io.Closer) {
u.mu.Lock()
u.audio = audio
aborted := u.aborted
u.mu.Unlock()
if aborted {
_ = audio.Close()
}
}
func (u *multipartUpload) clearAudio(audio io.Closer) {
u.mu.Lock()
if u.audio == audio {
u.audio = nil
}
u.mu.Unlock()
}
func (u *multipartUpload) abort() {
u.mu.Lock()
if u.aborted {
u.mu.Unlock()
return
}
u.aborted = true
audio := u.audio
u.mu.Unlock()
_ = u.reader.Close()
if audio != nil {
_ = audio.Close()
}
}
type contextReader struct {
ctx context.Context
reader io.Reader
}
func (r *contextReader) Read(p []byte) (int, error) {
select {
case <-r.ctx.Done():
return 0, r.ctx.Err()
default:
return r.reader.Read(p)
}
}
func writeFileAtomic(path string, data []byte, perm os.FileMode) error {
if strings.TrimSpace(path) == "" {
return fmt.Errorf("path is required")