correccion VAD en formato binario pcm16 - agrego transcripcion al live

This commit is contained in:
2025-08-01 21:22:11 +00:00
parent e43a8c01a7
commit 976e350436
11 changed files with 133 additions and 123 deletions

View File

@ -26,8 +26,8 @@ defmodule Whisper.Application do
{Phoenix.PubSub, name: Whisper.PubSub},
WhisperWeb.Endpoint,
Whisper.Counter,
AudioBuffer,
AudioFilesList
AudioBuffer
# AudioFilesList
]
opts = [strategy: :one_for_one, name: Whisper.Supervisor]

View File

@ -1,66 +1,66 @@
defmodule AudioFilesList do
use GenServer
require Logger
alias Phoenix.PubSub
# defmodule AudioFilesList do
# use GenServer
# require Logger
# alias Phoenix.PubSub
def start_link(_opts) do
GenServer.start_link(__MODULE__, :idle, name: __MODULE__)
end
# def start_link(_opts) do
# GenServer.start_link(__MODULE__, :idle, name: __MODULE__)
# end
def add_file(path) do
Logger.debug("add file")
GenServer.cast(__MODULE__, {:new_file, path})
end
# def add_file(path) do
# Logger.debug("add file")
# GenServer.cast(__MODULE__, {:new_file, path})
# end
def init(:idle) do
Logger.info("AudioFilesList iniciado")
{:ok, %{queue: [], processing: false}}
end
# def init(:idle) do
# Logger.info("AudioFilesList iniciado")
# {:ok, %{queue: [], processing: false}}
# end
def handle_cast({:new_file, path}, %{queue: queue, processing: false} = state) do
Logger.info("📥 Archivo encolado: #{path}")
queue = queue ++ [path]
[next | rest] = queue
{:noreply, %{queue: rest, processing: false}}
end
# def handle_cast({:new_file, path}, %{queue: queue, processing: false} = state) do
# Logger.info("📥 Archivo encolado: #{path}")
# queue = queue ++ [path]
# [next | rest] = queue
# {:noreply, %{queue: rest, processing: false}}
# end
def handle_cast({:new_file, path}, %{queue: queue, processing: true} = state) do
{:noreply, %{state | queue: queue ++ [path]}}
end
# def handle_cast({:new_file, path}, %{queue: queue, processing: true} = state) do
# {:noreply, %{state | queue: queue ++ [path]}}
# end
def handle_info(:done, %{queue: []} = state) do
{:noreply, %{state | processing: false}}
end
# def handle_info(:done, %{queue: []} = state) do
# {:noreply, %{state | processing: false}}
# end
def handle_info(:done, %{queue: [next | rest]} = state) do
{:noreply, %{state | queue: rest}}
end
# def handle_info(:done, %{queue: [next | rest]} = state) do
# {:noreply, %{state | queue: rest}}
# end
# def handle_cast({:done_processing, path}, %{queue: queue} = state) do
# new_queue = Enum.reject(queue, fn p -> p == path end)
# Logger.info("🗑️ Archivo eliminado y removido de la cola: #{path}")
# {:noreply, %{state | queue: new_queue}}
# end
# # def handle_cast({:done_processing, path}, %{queue: queue} = state) do
# # new_queue = Enum.reject(queue, fn p -> p == path end)
# # Logger.info("🗑️ Archivo eliminado y removido de la cola: #{path}")
# # {:noreply, %{state | queue: new_queue}}
# # end
# defp process_file(path) do
# Logger.info("▶️ Inicia procesamiento realtime: #{path}")
# # defp process_file(path) do
# # Logger.info("▶️ Inicia procesamiento realtime: #{path}")
# Task.start(fn ->
# case Whisper.SendToModel.realtime(path) do
# {:ok, text} when is_binary(text) ->
# message = %{"chunks" => [%{"text" => text}]}
# Phoenix.PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription, Jason.encode!(message)})
# Logger.info("✅ Transcripción (realtime): #{text}")
# File.rm!(path)
# # AudioFilesList.done_processing(path)
# {:error, reason} ->
# Logger.error("❌ Error transcribiendo: #{inspect(reason)}")
# end
# # Task.start(fn ->
# # case Whisper.SendToModel.realtime(path) do
# # {:ok, text} when is_binary(text) ->
# # message = %{"chunks" => [%{"text" => text}]}
# # Phoenix.PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription, Jason.encode!(message)})
# # Logger.info("✅ Transcripción (realtime): #{text}")
# # File.rm!(path)
# # # AudioFilesList.done_processing(path)
# # {:error, reason} ->
# # Logger.error("❌ Error transcribiendo: #{inspect(reason)}")
# # end
# send(__MODULE__, :done)
# # send(__MODULE__, :done)
# end)
# end
# # end)
# # end
end
# end

View File

@ -26,22 +26,20 @@ defmodule WhisperWeb.AudioChannel do
<<header_len::16, rest::binary>> = raw_binary
<<header::binary-size(header_len), audio::binary>> = rest
IO.inspect(header, label: "HEADER BINARIO RECIBIDO")
%{"sample_rate" => rate} = Jason.decode!(header)
ref = socket_id(socket)
case Jason.decode(header) do
{:ok, %{"sample_rate" => rate}} ->
Logger.info("Chunk recibido: #{byte_size(audio)} bytes, sample_rate: #{rate}")
AudioBuffer.append(socket_id(socket), {rate, audio})
{:noreply, socket}
Logger.info("Chunk recibido: #{byte_size(audio)} bytes, sample_rate: #{rate}")
AudioBuffer.append(ref, {rate, audio})
{:error, reason} ->
Logger.error("Error decodificando header JSON: #{inspect(reason)}")
{:noreply, socket}
end
# {:ok, path} = AudioSaver.save_chunk_as_wav(ref, audio, rate, "part")
# AudioFilesList.add_file(path)
{:noreply, socket}
end
@doc """
Recupera todos los chunks acumulados en el buffer, los concatena y guarda un archivo WAV final (sufijo `"final"`).
"""
@ -60,8 +58,8 @@ defmodule WhisperWeb.AudioChannel do
Task.start(fn ->
transcription = Whisper.SendToModel.large(path)
Logger.info("✅ Transcripción completa:\n#{transcription}")
# message = %{"chunks" => [%{"text" => transcription}]}
# Phoenix.PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription_m, Jason.encode!(message)})
message = %{"chunks" => [%{"text" => transcription}]}
Phoenix.PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription, Jason.encode!(message)})
File.rm!(path)
end)
end

View File

@ -1,28 +1,48 @@
defmodule WhisperWeb.VadLive do
use WhisperWeb, :live_view
alias Phoenix.PubSub
def mount(_, _, socket) do
{:ok, assign(socket, started: false)}
PubSub.subscribe(Whisper.PubSub, "transcription")
socket =
socket
|> assign(:transcription, "")
|> assign(:started, false)
{:ok, socket}
end
def handle_event("start_vad", _params, socket) do
push_event(socket, "init-vad", %{})
{:noreply, assign(socket, started: true)}
end
def handle_info({:transcription, raw_json}, socket) do
new_text =
raw_json
|> Jason.decode!()
|> get_in(["chunks", Access.at(0), "text"])
{:noreply, update(socket, :transcription, &(&1 <> " " <> new_text))}
end
def render(assigns) do
~H"""
<div id="vad-container" phx-hook="VadHook">
<button phx-click="start_vad" class="btn btn-primary">🎙 Iniciar VAD</button>
<button phx-click="stop_vad" class="btn btn-danger">🛑 Detener VAD</button>
<div id="vad-status" class="mt-4 text-sm text-gray-700"></div>
</div>
<div id="transcriptionContainer" class="w-full max-w-2xl space-y-4">
<%= if @transcription != "" do %>
<div class="p-4 bg-gray-100 rounded shadow-md">
<h2 class="text-sm font-semibold text-gray-700 mb-2">✅ Transcripción</h2>
<p class="text-green-600 whitespace-pre-wrap break-words text-sm leading-relaxed"><%= @transcription %></p>
</div>
<% end %>
</div>
"""
end
def handle_event("start_vad", _, socket) do
push_event(socket, "init-vad", %{})
{:noreply, socket}
end
def handle_event("stop_vad", _, socket) do
push_event(socket, "stop-vad", %{})
{:noreply, socket}
end
end