Maven/cmd/mavsttd/whisper_handler.go

package main

/*
#cgo CFLAGS: -I${SRCDIR}/../../deps/include -I${SRCDIR}/../../deps/whisper.cpp/ggml/include
#cgo LDFLAGS: -L${SRCDIR}/../../deps/lib -Wl,-rpath,${SRCDIR}/../../deps/lib -lwhisper -lggml -lggml-base -lggml-cpu -lggml-vulkan -lm -lstdc++ -fopenmp
#include <whisper.h>
#include <stdlib.h>
*/
import "C"

import (
	"context"
	"fmt"
	"math"
	"unsafe"

	"github.com/kami/maven/internal/worker"
)

type whisperHandler struct {
	ctx *C.struct_whisper_context
}

func newWhisperHandler(modelPath string) (*whisperHandler, error) {
	cparams := C.whisper_context_default_params()
	cPath := C.CString(modelPath)
	defer C.free(unsafe.Pointer(cPath))

	ctx := C.whisper_init_from_file_with_params(cPath, cparams)
	if ctx == nil {
		return nil, fmt.Errorf("whisper: failed to init from %s", modelPath)
	}
	return &whisperHandler{ctx: ctx}, nil
}

func (h *whisperHandler) Transcribe(ctx context.Context, req worker.TranscribeReq) (worker.TranscribeResp, error) {
	if err := ctx.Err(); err != nil {
		return worker.TranscribeResp{}, fmt.Errorf("whisper: context done before transcribe: %w", err)
	}
	a := req.Audio
	if len(a.Bytes) == 0 {
		return worker.TranscribeResp{}, fmt.Errorf("whisper: empty audio")
	}

	nSamples := len(a.Bytes) / 2
	samples := make([]float32, nSamples)
	for i := 0; i < nSamples; i++ {
		s := int16(a.Bytes[i*2]) | int16(a.Bytes[i*2+1])<<8
		samples[i] = float32(s) / 32768.0
	}

	params := C.whisper_full_default_params(C.WHISPER_SAMPLING_GREEDY)
	params.print_progress = false
	params.print_realtime = false
	params.print_timestamps = false
	params.print_special = false
	params.n_threads = C.int(4)
	params.single_segment = true

	lang := C.CString(req.Lang)
	defer C.free(unsafe.Pointer(lang))
	params.language = lang
	params.detect_language = false

	// CGo blocks the goroutine; whisper has no portable CGo-friendly abort
	// callback. Run in a goroutine so the caller's context cancellation at
	// least returns promptly — the CGo goroutine leaks until whisper finishes
	// but the caller doesn't hang.
	type result struct {
		code int
	}
	ch := make(chan result, 1)
	cSamples := (*C.float)(unsafe.Pointer(&samples[0]))
	go func() {
		ch <- result{code: int(C.whisper_full(h.ctx, params, cSamples, C.int(nSamples)))}
	}()
	select {
	case r := <-ch:
		if r.code != 0 {
			return worker.TranscribeResp{}, fmt.Errorf("whisper: full failed: %d", r.code)
		}
	case <-ctx.Done():
		return worker.TranscribeResp{}, fmt.Errorf("whisper: %w", ctx.Err())
	}

	nSegments := int(C.whisper_full_n_segments(h.ctx))
	if nSegments == 0 {
		return worker.TranscribeResp{Text: "", Confidence: 0}, nil
	}

	var text string
	totalLogProb := float64(0)
	totalTokens := 0

	for i := 0; i < nSegments; i++ {
		cSeg := C.whisper_full_get_segment_text(h.ctx, C.int(i))
		if cSeg != nil {
			text += C.GoString(cSeg)
		}

		nTokens := int(C.whisper_full_n_tokens(h.ctx, C.int(i)))
		for j := 0; j < nTokens; j++ {
			p := float64(C.whisper_full_get_token_p(h.ctx, C.int(i), C.int(j)))
			if p > 0 {
				totalLogProb += math.Log(p)
				totalTokens++
			}
		}
	}

	confidence := 0.0
	if totalTokens > 0 {
		avgLogProb := totalLogProb / float64(totalTokens)
		confidence = math.Exp(avgLogProb)
	}

	noSpeechProb := float64(C.whisper_full_get_segment_no_speech_prob(h.ctx, 0))
	if noSpeechProb > 0.9 {
		confidence = 0
	}

	if math.IsNaN(confidence) || math.IsInf(confidence, 0) {
		confidence = 0
	}

	return worker.TranscribeResp{
		Text:       text,
		Confidence: confidence,
	}, nil
}

func (h *whisperHandler) Close() {
	if h.ctx != nil {
		C.whisper_free(h.ctx)
		h.ctx = nil
	}
}