update from base repository

Dockerfile

@@ -1,14 +1,8 @@
 FROM swaggerapi/swagger-ui:v4.18.2 AS swagger-ui
 FROM nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu22.04
 
-ARG SERVICE_USER=service
-ARG SERVICE_UID=1001
-ARG SERVICE_GID=1001
-
 ENV PYTHON_VERSION=3.10
 ENV POETRY_VENV=/app/.venv
-ENV HF_HOME="/app/.cache"
-ENV ASR_MODEL_PATH="/app/.cache"
 
 RUN export DEBIAN_FRONTEND=noninteractive \
     && apt-get -qq update \
@@ -29,44 +23,28 @@ RUN ln -s -f /usr/bin/python${PYTHON_VERSION} /usr/bin/python3 && \
     ln -s -f /usr/bin/python${PYTHON_VERSION} /usr/bin/python && \
     ln -s -f /usr/bin/pip3 /usr/bin/pip
 
-RUN groupadd -g $SERVICE_GID $SERVICE_USER && \
-    useradd -u $SERVICE_UID -g $SERVICE_GID -d /app -s /usr/sbin/nologin $SERVICE_USER
-
-RUN getent group $SERVICE_USER
-RUN getent passwd $SERVICE_USER
-
-COPY --chown=$SERVICE_UID:$SERVICE_GID . /app
-COPY --chown=$SERVICE_UID:$SERVICE_GID --from=swagger-ui /usr/share/nginx/html/swagger-ui.css /app/swagger-ui-assets/swagger-ui.css
-COPY --chown=$SERVICE_UID:$SERVICE_GID --from=swagger-ui /usr/share/nginx/html/swagger-ui-bundle.js /app/swagger-ui-assets/swagger-ui-bundle.js
-
-RUN chown -R $SERVICE_UID:$SERVICE_GID /app
+RUN python3 -m venv $POETRY_VENV \
+    && $POETRY_VENV/bin/pip install -U pip setuptools \
+    && $POETRY_VENV/bin/pip install poetry==1.6.1
 
-RUN ls -la /app
-RUN mkdir -p /app/.cache && chown -R $SERVICE_UID:$SERVICE_GID /app/.cache && ls -la /app/.cache
-
-USER $SERVICE_USER
+ENV PATH="${PATH}:${POETRY_VENV}/bin"
 
 WORKDIR /app
 
-RUN python3 -m venv $POETRY_VENV && $POETRY_VENV/bin/pip install -U pip setuptools && \
-    $POETRY_VENV/bin/pip install poetry==1.6.1
-
-ENV PATH="${PATH}:${POETRY_VENV}/bin"
-
-COPY --chown=$SERVICE_UID:$SERVICE_GID poetry.lock pyproject.toml ./
+COPY poetry.lock pyproject.toml ./
 
 RUN poetry config virtualenvs.in-project true
 RUN poetry install --no-root
 
-RUN poetry install && rm -rf /app/.cache/pypoetry
+COPY . .
+COPY --from=swagger-ui /usr/share/nginx/html/swagger-ui.css swagger-ui-assets/swagger-ui.css
+COPY --from=swagger-ui /usr/share/nginx/html/swagger-ui-bundle.js swagger-ui-assets/swagger-ui-bundle.js
+
+RUN poetry install && rm -rf /root/.cache/pypoetry
 RUN $POETRY_VENV/bin/pip install --no-cache-dir torch==1.13.1+cu117 -f https://download.pytorch.org/whl/torch
 
 WORKDIR /app/reascripts/ReaSpeech
-
-RUN ls -la /app/reascripts/ReaSpeech
-
 RUN make publish
-
 WORKDIR /app
 RUN rm -rf reascripts
 

app/faster_whisper/core.py

@@ -13,6 +13,7 @@ from .utils import ResultWriter, WriteTXT, WriteSRT, WriteVTT, WriteTSV, WriteJS
 ASR_ENGINE_OPTIONS = frozenset([
     "task",
     "language",
+    "hotwords",
     "initial_prompt",
     "vad_filter",
     "word_timestamps",

app/run.py

@@ -30,6 +30,9 @@ argmap = {
     '--asr-model': {
         'default': os.getenv('ASR_MODEL', 'small'),
         'help': 'ASR model to use (default: %(default)s)' },
+    '--build-reascripts': {
+        'action': 'store_true',
+        'help': 'Build ReaScripts before starting' },
 }
 
 parser = argparse.ArgumentParser()
@@ -46,26 +49,53 @@ os.environ['FFMPEG_BIN'] = args.ffmpeg_bin
 os.environ['ASR_ENGINE'] = args.asr_engine
 os.environ['ASR_MODEL'] = args.asr_model
 
+if args.build_reascripts:
+    if os.system('cd reascripts/ReaSpeech && make') != 0:
+        print('ReaScript build failed', file=sys.stderr)
+        sys.exit(1)
+
+processes = {}
+
 # Start Redis
 print('Starting database...', file=sys.stderr)
-subprocess.Popen([args.redis_bin], stdout=subprocess.DEVNULL)
+processes['redis'] = subprocess.Popen([args.redis_bin], stdout=subprocess.DEVNULL)
 
 # Start Celery
 print('Starting worker...', file=sys.stderr)
-subprocess.Popen(['celery', '-A', 'app.worker.celery', 'worker', '--pool=solo', '--loglevel=info'])
+processes['celery'] = subprocess.Popen(['celery', '-A', 'app.worker.celery', 'worker', '--pool=solo', '--loglevel=info'])
 
 # Start Gunicorn
 print('Starting application...', file=sys.stderr)
-subprocess.Popen(['gunicorn', '--bind', '0.0.0.0:9000', '--workers', '1', '--timeout', '0', 'app.webservice:app', '-k', 'uvicorn.workers.UvicornWorker'])
+processes['gunicorn'] = subprocess.Popen(['gunicorn', '--bind', '0.0.0.0:9000', '--workers', '1', '--timeout', '0', 'app.webservice:app', '-k', 'uvicorn.workers.UvicornWorker'])
 
 # Wait for any process to exit
-status = os.WEXITSTATUS(os.wait()[1])
-print('Process exited with status', status, file=sys.stderr)
+pid, waitstatus = os.wait()
+exitcode = os.waitstatus_to_exitcode(waitstatus)
+process_name = '<unknown>'
+for name, p in processes.items():
+    if p.pid == pid:
+        process_name = name
+        break
+if exitcode < 0:
+    print('Process', process_name, 'received signal', -exitcode, file=sys.stderr)
+else:
+    print('Process', process_name, 'exited with status', exitcode, file=sys.stderr)
 
 # Terminate any child processes
 print('Terminating child processes...', file=sys.stderr)
-os.system('pkill -P %d' % os.getpid())
+for name, p in processes.items():
+    try:
+        print('Terminating', name, file=sys.stderr)
+
+        # kinda bass-ackwards, but poll() returns None if process is still running
+        if not p.poll():
+            p.terminate()
+        else:
+            print(name, "already exited", file=sys.stderr)
+    except Exception as e:
+        print(e, file=sys.stderr)
 
 # Exit with status of process that exited
+status = 1 if exitcode < 0 else exitcode
 print('Exiting with status', status, file=sys.stderr)
 sys.exit(status)

app/webservice.py

@@ -24,6 +24,7 @@ ASR_ENGINE = os.getenv("ASR_ENGINE", "faster_whisper")
 ASR_OPTIONS = frozenset([
     "task",
     "language",
+    "hotwords",
     "initial_prompt",
     "encode",
     "output",
@@ -36,6 +37,8 @@ DEFAULT_MODEL_NAME = os.getenv("ASR_MODEL", "small")
 
 LANGUAGE_CODES = sorted(list(tokenizer.LANGUAGES.keys()))
 
+TASK_EXPIRATION_SECONDS = 30
+
 projectMetadata = importlib.metadata.metadata('reaspeech')
 app = FastAPI(
     # docs_url=None,
@@ -111,6 +114,7 @@ async def reascript(request: Request, name: str, host: str):
 async def asr(
     task: Union[str, None] = Query(default="transcribe", enum=["transcribe", "translate"]),
     language: Union[str, None] = Query(default=None, enum=LANGUAGE_CODES),
+    hotwords: Union[str, None] = Query(default=None),
     initial_prompt: Union[str, None] = Query(default=None),
     audio_file: UploadFile = File(...),
     encode: bool = Query(default=True, description="Encode audio first through ffmpeg"),
@@ -137,7 +141,7 @@ async def asr(
     transcriber = transcribe.si(temp_file_path, audio_file.filename, asr_options)
 
     if use_async:
-        job = transcriber.apply_async()
+        job = transcriber.apply_async(expires=TASK_EXPIRATION_SECONDS)
         return JSONResponse({"job_id": job.id})
 
     else:

reascripts/ReaSpeech/source/ColumnLayout.lua

@@ -0,0 +1,70 @@
+--[[
+
+ColumnLayout.lua - Fixed-width column layout helper
+
+]]--
+
+ColumnLayout = Polo {
+  DEFAULT_COLUMN_PADDING = 15,
+  DEFAULT_NUM_COLUMNS = 3,
+}
+
+function ColumnLayout:init()
+  assert(self.render_column, 'render_column function must be provided')
+  self.column_padding = self.column_padding or self.DEFAULT_COLUMN_PADDING
+  self.margin_bottom = self.margin_bottom or 0
+  self.margin_left = self.margin_left or 0
+  self.margin_right = self.margin_right or 0
+  self.margin_top = self.margin_top or 0
+  self.num_columns = self.num_columns or self.DEFAULT_NUM_COLUMNS
+  self.width = self.width or 0
+end
+
+function ColumnLayout:render()
+  local total_padding = (self.num_columns - 1) * self.column_padding
+  local total_width = self.width
+  if total_width == 0 then
+    total_width = self:_get_avail_width()
+  end
+  local content_width = total_width - self.margin_left - self.margin_right
+  local column_width = (content_width - total_padding) / self.num_columns
+
+  self:_horiz_margin(self.margin_left)
+
+  for i = 1, self.num_columns do
+    local column = {num = i, width = column_width}
+
+    self:_with_group(function ()
+      self:_vert_margin(self.margin_top, column_width)
+      self.render_column(column)
+      self:_vert_margin(self.margin_bottom, column_width)
+    end)
+
+    if i < self.num_columns then
+      self:_column_gap(self.column_padding)
+    end
+  end
+end
+
+function ColumnLayout:_column_gap(padding)
+  ImGui.SameLine(ctx, 0, padding)
+end
+
+function ColumnLayout:_get_avail_width()
+  local avail_width, _ = ImGui.GetContentRegionAvail(ctx)
+  return avail_width
+end
+
+function ColumnLayout:_horiz_margin(margin)
+  ImGui.SetCursorPosX(ctx, ImGui.GetCursorPosX(ctx) + margin)
+end
+
+function ColumnLayout:_vert_margin(margin, width)
+  ImGui.Dummy(ctx, width, margin)
+end
+
+function ColumnLayout:_with_group(f)
+  ImGui.BeginGroup(ctx)
+  app:trap(f)
+  ImGui.EndGroup(ctx)
+end

reascripts/ReaSpeech/source/ReaSpeechAPI.lua

@@ -30,8 +30,10 @@ end
 
 -- Fetch simple JSON responses. Will block until result or curl timeout.
 -- For large amounts of data, use fetch_large instead.
-function ReaSpeechAPI:fetch_json(url_path, http_method, error_handler)
+function ReaSpeechAPI:fetch_json(url_path, http_method, error_handler, timeout_handler)
   http_method = http_method or 'GET'
+  error_handler = error_handler or function(_msg) end
+  timeout_handler = timeout_handler or function() end
 
   local curl = self:get_curl_cmd()
   local api_url = self:get_api_url(url_path)
@@ -63,8 +65,13 @@ function ReaSpeechAPI:fetch_json(url_path, http_method, error_handler)
   end
 
   local status, output = exec_result:match("(%d+)\n(.*)")
+  status = tonumber(status)
 
-  if tonumber(status) ~= 0 then
+  if status == 28 then
+    app:debug("Curl timeout reached")
+    timeout_handler()
+    return nil
+  elseif status ~= 0 then
     local msg = "Curl failed with status " .. status
     app:debug(msg)
     error_handler(msg)
@@ -190,7 +197,7 @@ function ReaSpeechAPI.http_status_and_body(response)
 
   local status = last_status_line:match("^HTTP/%d%.%d%s+(%d+)")
   if not status then
-    return -1, 'Unable to parse response'
+    return -1, 'Status not found in headers'
   end
 
   local body = {}

reascripts/ReaSpeech/source/ReaSpeechActionsUI.lua

@@ -60,9 +60,13 @@ function ReaSpeechActionsUI:render()
     end
 
     ImGui.SameLine(ctx)
-    ImGui.ProgressBar(ctx, progress)
+    local overlay = string.format("%.0f%%", progress * 100)
+    local status = self.worker:status()
+    if status then
+      overlay = overlay .. ' - ' .. status
+    end
+    ImGui.ProgressBar(ctx, progress, nil, nil, overlay)
   end
-  ImGui.Dummy(ctx,0, 5)
 end
 
 function ReaSpeechActionsUI.make_job(media_item, take)

reascripts/ReaSpeech/source/ReaSpeechControlsUI.lua

@@ -7,178 +7,305 @@ ReaSpeechControlsUI.lua - UI elements for configuring ASR services
 ReaSpeechControlsUI = Polo {
   -- Copied from whisper.tokenizer.LANGUAGES
   LANGUAGES = {
-    en = 'english', zh = 'chinese', de = 'german',
-    es = 'spanish', ru = 'russian', ko = 'korean',
-    fr = 'french', ja = 'japanese', pt = 'portuguese',
-    tr = 'turkish', pl = 'polish', ca = 'catalan',
-    nl = 'dutch', ar = 'arabic', sv = 'swedish',
-    it = 'italian', id = 'indonesian', hi = 'hindi',
-    fi = 'finnish', vi = 'vietnamese', he = 'hebrew',
-    uk = 'ukrainian', el = 'greek', ms = 'malay',
-    cs = 'czech', ro = 'romanian', da = 'danish',
-    hu = 'hungarian', ta = 'tamil', no = 'norwegian',
-    th = 'thai', ur = 'urdu', hr = 'croatian',
-    bg = 'bulgarian', lt = 'lithuanian', la = 'latin',
-    mi = 'maori', ml = 'malayalam', cy = 'welsh',
-    sk = 'slovak', te = 'telugu', fa = 'persian',
-    lv = 'latvian', bn = 'bengali', sr = 'serbian',
-    az = 'azerbaijani', sl = 'slovenian', kn = 'kannada',
-    et = 'estonian', mk = 'macedonian', br = 'breton',
-    eu = 'basque', is = 'icelandic', hy = 'armenian',
-    ne = 'nepali', mn = 'mongolian', bs = 'bosnian',
-    kk = 'kazakh', sq = 'albanian', sw = 'swahili',
-    gl = 'galician', mr = 'marathi', pa = 'punjabi',
-    si = 'sinhala', km = 'khmer', sn = 'shona',
-    yo = 'yoruba', so = 'somali', af = 'afrikaans',
-    oc = 'occitan', ka = 'georgian', be = 'belarusian',
-    tg = 'tajik', sd = 'sindhi', gu = 'gujarati',
-    am = 'amharic', yi = 'yiddish', lo = 'lao',
-    uz = 'uzbek', fo = 'faroese', ht = 'haitian creole',
-    ps = 'pashto', tk = 'turkmen', nn = 'nynorsk',
-    mt = 'maltese', sa = 'sanskrit', lb = 'luxembourgish',
-    my = 'myanmar', bo = 'tibetan', tl = 'tagalog',
-    mg = 'malagasy', as = 'assamese', tt = 'tatar',
-    haw = 'hawaiian', ln = 'lingala', ha = 'hausa',
-    ba = 'bashkir', jw = 'javanese', su = 'sundanese'
+    en = 'English', zh = 'Chinese', de = 'German',
+    es = 'Spanish', ru = 'Russian', ko = 'Korean',
+    fr = 'French', ja = 'Japanese', pt = 'Portuguese',
+    tr = 'Turkish', pl = 'Polish', ca = 'Catalan',
+    nl = 'Dutch', ar = 'Arabic', sv = 'Swedish',
+    it = 'Italian', id = 'Indonesian', hi = 'Hindi',
+    fi = 'Finnish', vi = 'Vietnamese', he = 'Hebrew',
+    uk = 'Ukrainian', el = 'Greek', ms = 'Malay',
+    cs = 'Czech', ro = 'Romanian', da = 'Danish',
+    hu = 'Hungarian', ta = 'Tamil', no = 'Norwegian',
+    th = 'Thai', ur = 'Urdu', hr = 'Croatian',
+    bg = 'Bulgarian', lt = 'Lithuanian', la = 'Latin',
+    mi = 'Maori', ml = 'Malayalam', cy = 'Welsh',
+    sk = 'Slovak', te = 'Telugu', fa = 'Persian',
+    lv = 'Latvian', bn = 'Bengali', sr = 'Serbian',
+    az = 'Azerbaijani', sl = 'Slovenian', kn = 'Kannada',
+    et = 'Estonian', mk = 'Macedonian', br = 'Breton',
+    eu = 'Basque', is = 'Icelandic', hy = 'Armenian',
+    ne = 'Nepali', mn = 'Mongolian', bs = 'Bosnian',
+    kk = 'Kazakh', sq = 'Albanian', sw = 'Swahili',
+    gl = 'Galician', mr = 'Marathi', pa = 'Punjabi',
+    si = 'Sinhala', km = 'Khmer', sn = 'Shona',
+    yo = 'Yoruba', so = 'Somali', af = 'Afrikaans',
+    oc = 'Occitan', ka = 'Georgian', be = 'Belarusian',
+    tg = 'Tajik', sd = 'Sindhi', gu = 'Gujarati',
+    am = 'Amharic', yi = 'Yiddish', lo = 'Lao',
+    uz = 'Uzbek', fo = 'Faroese', ht = 'Haitian Creole',
+    ps = 'Pashto', tk = 'Turkmen', nn = 'Nynorsk',
+    mt = 'Maltese', sa = 'Sanskrit', lb = 'Luxembourgish',
+    my = 'Myanmar', bo = 'Tibetan', tl = 'Tagalog',
+    mg = 'Malagasy', as = 'Assamese', tt = 'Tatar',
+    haw = 'Hawaiian', ln = 'Lingala', ha = 'Hausa',
+    ba = 'Bashkir', jw = 'Javanese', su = 'Sundanese'
   },
+
   LANGUAGE_CODES = {},
-  DEFAULT_LANGUAGE = 'en',
 
-  ITEM_WIDTH = 125,
-  LARGE_ITEM_WIDTH = 375,
+  DEFAULT_LANGUAGE = '',
+  DEFAULT_MODEL_NAME = 'small',
+
+  SIMPLE_MODEL_SIZES = {
+    {'Small', 'small'},
+    {'Medium', 'medium'},
+    {'Large', 'distil-large-v3'},
+  },
+
+  COLUMN_PADDING = 15,
+  MARGIN_BOTTOM = 5,
+  MARGIN_LEFT = 115,
+  MARGIN_RIGHT = 0,
+  NARROW_COLUMN_WIDTH = 150,
 }
 
+ReaSpeechControlsUI._init_languages = function ()
+  for code, _ in pairs(ReaSpeechControlsUI.LANGUAGES) do
+    table.insert(ReaSpeechControlsUI.LANGUAGE_CODES, code)
+  end
+
+  table.sort(ReaSpeechControlsUI.LANGUAGE_CODES, function (a, b)
+    return ReaSpeechControlsUI.LANGUAGES[a] < ReaSpeechControlsUI.LANGUAGES[b]
+  end)
+
+  table.insert(ReaSpeechControlsUI.LANGUAGE_CODES, 1, '')
+  ReaSpeechControlsUI.LANGUAGES[''] = 'Detect'
+end
+
+ReaSpeechControlsUI._init_languages()
+
 function ReaSpeechControlsUI:init()
+  self.tab = 'simple'
+
   self.log_enable = false
   self.log_debug = false
 
   self.language = self.DEFAULT_LANGUAGE
   self.translate = false
+  self.hotwords = ''
   self.initial_prompt = ''
-  self.model_name = nil
+  self.model_name = self.DEFAULT_MODEL_NAME
+  self.vad_filter = true
+
+  self:init_layouts()
 end
 
 function ReaSpeechControlsUI:get_request_data()
   return {
     language = self.language,
     translate = self.translate,
+    hotwords = self.hotwords,
     initial_prompt = self.initial_prompt,
     model_name = self.model_name,
+    vad_filter = self.vad_filter,
   }
 end
 
-ReaSpeechControlsUI._init_languages = function ()
-  for code, _ in pairs(ReaSpeechControlsUI.LANGUAGES) do
-    table.insert(ReaSpeechControlsUI.LANGUAGE_CODES, code)
+function ReaSpeechControlsUI:init_layouts()
+  self:init_simple_layouts()
+  self:init_advanced_layouts()
+end
+
+function ReaSpeechControlsUI:init_simple_layouts()
+  local with_button_color = function (selected, f)
+    if selected then
+      ImGui.PushStyleColor(ctx, ImGui.Col_Button(), Theme.colors.dark_gray_translucent)
+      app:trap(f)
+      ImGui.PopStyleColor(ctx)
+    else
+      f()
+    end
   end
 
-  table.sort(ReaSpeechControlsUI.LANGUAGE_CODES, function (a, b)
-    return ReaSpeechControlsUI.LANGUAGES[a] < ReaSpeechControlsUI.LANGUAGES[b]
-  end)
+  self.model_sizes_layout = ColumnLayout.new {
+    column_padding = self.COLUMN_PADDING,
+    margin_bottom = self.MARGIN_BOTTOM,
+    margin_left = self.MARGIN_LEFT,
+    margin_right = self.MARGIN_RIGHT,
+    num_columns = #self.SIMPLE_MODEL_SIZES,
 
-  table.insert(ReaSpeechControlsUI.LANGUAGE_CODES, 1, '')
-  ReaSpeechControlsUI.LANGUAGES[''] = 'detect'
+    render_column = function (column)
+      self:render_input_label(column.num == 1 and 'Model Size' or '')
+      local label, model_name = table.unpack(self.SIMPLE_MODEL_SIZES[column.num])
+      with_button_color(self.model_name == model_name, function ()
+        if ImGui.Button(ctx, label, column.width) then
+          self.model_name = model_name
+        end
+      end)
+    end
+  }
 end
 
-ReaSpeechControlsUI._init_languages()
+function ReaSpeechControlsUI:init_advanced_layouts()
+  local renderers = {
+    {self.render_model_name, self.render_hotwords, self.render_language},
+    {self.render_options, self.render_initial_prompt, self.render_logging},
+  }
+
+  self.advanced_layouts = {}
+
+  for row = 1, #renderers do
+    self.advanced_layouts[row] = ColumnLayout.new {
+      column_padding = self.COLUMN_PADDING,
+      margin_bottom = self.MARGIN_BOTTOM,
+      margin_left = self.MARGIN_LEFT,
+      margin_right = self.MARGIN_RIGHT,
+      num_columns = #renderers[row],
+
+      render_column = function (column)
+        ImGui.PushItemWidth(ctx, column.width)
+        app:trap(function () renderers[row][column.num](self, column) end)
+        ImGui.PopItemWidth(ctx)
+      end
+    }
+  end
+end
 
 function ReaSpeechControlsUI:render()
-  --start input table so logo and inputs sit side-by-side
-  if ImGui.BeginTable(ctx, 'InputTable', 2) then
-    app:trap(function()
-      --column settings
-      ImGui.TableSetupColumn(ctx, 'Logo', ImGui.TableColumnFlags_WidthFixed())
-      ImGui.TableSetupColumn(ctx, 'Inputs', ImGui.TableColumnFlags_WidthFixed())
-      -- first column
-      ImGui.TableNextColumn(ctx)
-      ImGui.SameLine(ctx, -10)
-      app.png_from_bytes('reaspeech-logo-small')
-      -- second column
-      ImGui.TableNextColumn(ctx)
-      -- start language selection
-      self:render_language_controls()
-      ImGui.Dummy(ctx,0, 10)
-      self:render_advanced_controls()
-    end)
-    ImGui.EndTable(ctx)
+  self:render_heading()
+  if self.tab == 'advanced' then
+    self:render_advanced()
+  else
+    self:render_simple()
   end
-  -- end input table
-  ImGui.SameLine(ctx, ImGui.GetWindowWidth(ctx) - self.ITEM_WIDTH + 65)
+  ImGui.Separator(ctx)
+  ImGui.Dummy(ctx, 0, 5)
+end
+
+function ReaSpeechControlsUI:render_heading()
+  local init_x, init_y = ImGui.GetCursorPos(ctx)
+
+  ImGui.SetCursorPosX(ctx, init_x - 20)
+  app.png_from_bytes('reaspeech-logo-small')
+
+  ImGui.SetCursorPos(ctx, init_x + self.MARGIN_LEFT + 2, init_y)
+  self:render_tabs()
+
+  ImGui.SetCursorPos(ctx, ImGui.GetWindowWidth(ctx) - 55, init_y)
   app.png_from_bytes('heading-logo-tech-audio')
+
+  ImGui.SetCursorPos(ctx, init_x, init_y + 40)
 end
 
-function ReaSpeechControlsUI:render_language_controls()
-  if ImGui.TreeNode(ctx, 'Language Options', ImGui.TreeNodeFlags_DefaultOpen()) then
-    app:trap(function()
-      ImGui.Dummy(ctx, 0, 25)
-      ImGui.SameLine(ctx)
-      if ImGui.BeginCombo(ctx, "language", self.LANGUAGES[self.language]) then
-        app:trap(function()
-          local combo_items = self.LANGUAGE_CODES
-          for _, combo_item in pairs(combo_items) do
-            local is_selected = (combo_item == self.language)
-            if ImGui.Selectable(ctx, self.LANGUAGES[combo_item], is_selected) then
-              self.language = combo_item
-            end
-          end
+function ReaSpeechControlsUI:render_tabs()
+  if ImGui.BeginTabBar(ctx, '##tabs', ImGui.TabBarFlags_None()) then
+    app:trap(function ()
+      if ImGui.BeginTabItem(ctx, 'Simple') then
+        app:trap(function ()
+          self.tab = 'simple'
         end)
-        ImGui.EndCombo(ctx)
+        ImGui.EndTabItem(ctx)
       end
-      local rv, value
-      ImGui.SameLine(ctx)
-      rv, value = ImGui.Checkbox(ctx, "translate", self.translate)
-      if rv then
-        self.translate = value
+      if ImGui.BeginTabItem(ctx, 'Advanced') then
+        app:trap(function ()
+          self.tab = 'advanced'
+        end)
+        ImGui.EndTabItem(ctx)
       end
     end)
-
-    ImGui.TreePop(ctx)
+    ImGui.EndTabBar(ctx)
   end
 end
 
-function ReaSpeechControlsUI:render_advanced_controls()
-  local rv, value
+function ReaSpeechControlsUI:render_simple()
+  self:render_model_sizes()
+end
 
-  if ImGui.TreeNode(ctx, 'Advanced Options') then
-    app:trap(function()
-      ImGui.Dummy(ctx, 0, 25)
-
-      ImGui.SameLine(ctx)
-      ImGui.PushItemWidth(ctx, self.LARGE_ITEM_WIDTH)
-      app:trap(function ()
-        rv, value = ImGui.InputText(ctx, 'initial prompt', self.initial_prompt)
-        if rv then
-          self.initial_prompt = value
-        end
-      end)
-      ImGui.PopItemWidth(ctx)
-
-      ImGui.SameLine(ctx)
-      ImGui.PushItemWidth(ctx, 100)
-      app:trap(function ()
-        rv, value = ImGui.InputTextWithHint(ctx, 'model name', self.model_name or "<default>")
-        if rv then
-          self.model_name = value
-        end
-      end)
-      ImGui.PopItemWidth(ctx)
+function ReaSpeechControlsUI:render_advanced()
+  for row = 1, #self.advanced_layouts do
+    self.advanced_layouts[row]:render()
+  end
+end
 
-      ImGui.SameLine(ctx)
-      rv, value = ImGui.Checkbox(ctx, "log", self.log_enable)
-      if rv then
-        self.log_enable = value
-      end
+function ReaSpeechControlsUI:render_input_label(text)
+  ImGui.Text(ctx, text)
+  ImGui.Dummy(ctx, 0, 0)
+end
+
+function ReaSpeechControlsUI:render_language(column)
+  self:render_input_label('Language')
 
-      if self.log_enable then
-        ImGui.SameLine(ctx)
-        rv, value = ImGui.Checkbox(ctx, "debug", self.log_debug)
-        if rv then
-          self.log_debug = value
+  if ImGui.BeginCombo(ctx, "##language", self.LANGUAGES[self.language]) then
+    app:trap(function()
+      local combo_items = self.LANGUAGE_CODES
+      for _, combo_item in pairs(combo_items) do
+        local is_selected = (combo_item == self.language)
+        if ImGui.Selectable(ctx, self.LANGUAGES[combo_item], is_selected) then
+          self.language = combo_item
         end
       end
     end)
+    ImGui.EndCombo(ctx)
+  end
+
+  local translate_label = "Translate to English"
+  if column.width < self.NARROW_COLUMN_WIDTH then
+    translate_label = "Translate"
+  end
+  local rv, value = ImGui.Checkbox(ctx, translate_label, self.translate)
+  if rv then
+    self.translate = value
+  end
+end
+
+function ReaSpeechControlsUI:render_model_name()
+  self:render_input_label('Model Name')
+
+  local rv, value = ImGui.InputTextWithHint(ctx, '##model_name', self.model_name or "<default>")
+  if rv then
+    self.model_name = value
+  end
+end
+
+function ReaSpeechControlsUI:render_model_sizes()
+  self.model_sizes_layout:render()
+end
+
+function ReaSpeechControlsUI:render_hotwords()
+  self:render_input_label('Hot Words')
+
+  local rv, value = ImGui.InputText(ctx, '##hotwords', self.hotwords)
+  if rv then
+    self.hotwords = value
+  end
+end
+
+function ReaSpeechControlsUI:render_options(column)
+  self:render_input_label('Options')
+
+  local vad_label = "Voice Activity Detection"
+  if column.width < self.NARROW_COLUMN_WIDTH then
+    vad_label = "VAD"
+  end
+  local rv, value = ImGui.Checkbox(ctx, vad_label, self.vad_filter)
+  if rv then
+    self.vad_filter = value
+  end
+end
+
+function ReaSpeechControlsUI:render_logging()
+  self:render_input_label('Logging')
+
+  local rv, value = ImGui.Checkbox(ctx, "Enable", self.log_enable)
+  if rv then
+    self.log_enable = value
+  end
+
+  if self.log_enable then
+    ImGui.SameLine(ctx)
+    rv, value = ImGui.Checkbox(ctx, "Debug", self.log_debug)
+    if rv then
+      self.log_debug = value
+    end
+  end
+end
+
+function ReaSpeechControlsUI:render_initial_prompt()
+  self:render_input_label('Initial Prompt')
 
-    ImGui.TreePop(ctx)
-    ImGui.Spacing(ctx)
+  rv, value = ImGui.InputText(ctx, '##initial_prompt', self.initial_prompt)
+  if rv then
+    self.initial_prompt = value
   end
 end

reascripts/ReaSpeech/source/ReaSpeechWorker.lua

@@ -89,6 +89,12 @@ function ReaSpeechWorker:progress()
   return completed_job_count / job_count
 end
 
+function ReaSpeechWorker:status()
+  if self.active_job and self.active_job.job then
+    return self.active_job.job.job_status
+  end
+end
+
 function ReaSpeechWorker:cancel()
   if self.active_job then
     if self.active_job.job and self.active_job.job.job_id then
@@ -123,7 +129,7 @@ function ReaSpeechWorker:handle_request(request)
     task = request.translate and 'translate' or 'transcribe',
     output = 'json',
     use_async = 'true',
-    vad_filter = 'true',
+    vad_filter = request.vad_filter and 'true' or 'false',
     word_timestamps = 'true',
     model_name = request.model_name,
   }
@@ -132,6 +138,10 @@ function ReaSpeechWorker:handle_request(request)
     data.language = request.language
   end
 
+  if request.hotwords and request.hotwords ~= '' then
+    data.hotwords = request.hotwords
+  end
+
   if request.initial_prompt and request.initial_prompt ~= '' then
     data.initial_prompt = request.initial_prompt
   end
@@ -156,6 +166,7 @@ function ReaSpeechWorker:handle_job_status(active_job, response)
   end
 
   active_job.job.job_id = response.job_id
+  active_job.job.job_status = response.job_status
 
   if not response.job_status then
     return false
@@ -252,21 +263,23 @@ function ReaSpeechWorker:handle_response_json(output_file, sentinel_file, succes
 
   local f = io.open(output_file, 'r')
   if not f then
-    fail_f("Couldn't open output_filename: " .. tostring(output_file))
+    fail_f("Couldn't open output file: " .. tostring(output_file))
+    Tempfile:remove(sentinel_file)
     return
   end
 
   local http_status, body = ReaSpeechAPI.http_status_and_body(f)
   f:close()
 
-  if #body < 1 then
-    fail_f("Empty response from server.")
+  if http_status == -1 then
+    app:debug(body .. ", trying again later")
     return
   end
 
+  Tempfile:remove(output_file)
+  Tempfile:remove(sentinel_file)
+
   if http_status ~= 200 then
-    Tempfile:remove(sentinel_file)
-    Tempfile:remove(output_file)
     local msg = "Server responded with status " .. http_status
     fail_f(msg)
     app:log(msg)
@@ -274,15 +287,18 @@ function ReaSpeechWorker:handle_response_json(output_file, sentinel_file, succes
     return
   end
 
+  if #body < 1 then
+    fail_f("Empty response from server")
+    return
+  end
+
   local response = nil
   if app:trap(function ()
     response = json.decode(body)
   end) then
-    Tempfile:remove(sentinel_file)
-    Tempfile:remove(output_file)
     success_f(response)
   else
-    app:debug("JSON parse error, trying again later")
+    fail_f("Error parsing response JSON")
   end
 end
 

reascripts/ReaSpeech/source/Theme.lua

@@ -33,7 +33,10 @@ function Theme.init()
       { ImGui.Col_CheckMark(), Theme.colors.pink_opaque },
       { ImGui.Col_HeaderHovered(), Theme.colors.dark_gray_semi_opaque },
       { ImGui.Col_HeaderActive(), Theme.colors.dark_gray_semi_transparent },
-      { ImGui.Col_Header(), Theme.colors.dark_gray_semi_opaque }
+      { ImGui.Col_Header(), Theme.colors.dark_gray_semi_opaque },
+      { ImGui.Col_Tab(), Theme.colors.dark_gray_opaque },
+      { ImGui.Col_TabActive(), Theme.colors.medium_gray_opaque },
+      { ImGui.Col_TabHovered(), Theme.colors.dark_gray_translucent },
     },
 
     styles = {
@@ -47,4 +50,4 @@ function Theme.init()
   })
 
   return Theme.theme
-end
+end

reascripts/ReaSpeech/tests/TestColumnLayout.lua

@@ -0,0 +1,202 @@
+package.path = '../common/libs/?.lua;../common/vendor/?.lua;' .. package.path
+
+local lu = require('luaunit')
+
+require('Polo')
+
+require('source/ColumnLayout')
+
+
+
+TestColumnLayout = {
+  AVAIL_WIDTH = 100
+}
+
+function TestColumnLayout:stub(layout)
+  layout._column_gap_calls = {}
+  layout._column_gap = function (self, padding)
+    table.insert(self._column_gap_calls, {padding})
+  end
+
+  layout._get_avail_width_calls = 0
+  layout._get_avail_width = function (self)
+    self._get_avail_width_calls = self._get_avail_width_calls + 1
+    return TestColumnLayout.AVAIL_WIDTH
+  end
+
+  layout._horiz_margin_calls = {}
+  layout._horiz_margin = function (self, margin)
+    table.insert(self._horiz_margin_calls, {margin})
+  end
+
+  layout._vert_margin_calls = {}
+  layout._vert_margin = function (self, margin, width)
+    table.insert(self._vert_margin_calls, {margin, width})
+  end
+
+  layout._with_group_calls = 0
+  layout._with_group = function (self, f)
+    self._with_group_calls = self._with_group_calls + 1
+    f()
+  end
+end
+
+function TestColumnLayout:testInitDefault()
+  local layout = ColumnLayout.new {
+    render_column = function () end
+  }
+  lu.assertEquals(layout.column_padding, ColumnLayout.DEFAULT_COLUMN_PADDING)
+  lu.assertEquals(layout.margin_bottom, 0)
+  lu.assertEquals(layout.margin_left, 0)
+  lu.assertEquals(layout.margin_right, 0)
+  lu.assertEquals(layout.margin_top, 0)
+  lu.assertEquals(layout.num_columns, ColumnLayout.DEFAULT_NUM_COLUMNS)
+  lu.assertEquals(layout.width, 0)
+end
+
+function TestColumnLayout:testRender()
+  local column_padding = 20
+  local margin_bottom = 15
+  local margin_left = 10
+  local margin_right = 30
+  local margin_top = 5
+  local num_columns = 2
+
+  local expected_column_width = (
+    self.AVAIL_WIDTH
+    - margin_left
+    - margin_right
+    - column_padding
+  ) / num_columns
+
+  local render_column_calls = {}
+
+  local layout = ColumnLayout.new {
+    column_padding = column_padding,
+    margin_bottom = margin_bottom,
+    margin_left = margin_left,
+    margin_right = margin_right,
+    margin_top = margin_top,
+    num_columns = num_columns,
+    render_column = function (column)
+      table.insert(render_column_calls, {column})
+    end
+  }
+
+  self:stub(layout)
+  layout:render()
+
+  lu.assertEquals(layout._column_gap_calls, {{column_padding}})
+  lu.assertEquals(layout._get_avail_width_calls, 1)
+  lu.assertEquals(layout._horiz_margin_calls, {{margin_left}})
+  lu.assertEquals(layout._vert_margin_calls, {
+    {margin_top, expected_column_width},
+    {margin_bottom, expected_column_width},
+    {margin_top, expected_column_width},
+    {margin_bottom, expected_column_width},
+  })
+  lu.assertEquals(layout._with_group_calls, num_columns)
+
+  lu.assertEquals(render_column_calls, {
+    {{num = 1, width = expected_column_width}},
+    {{num = 2, width = expected_column_width}},
+  })
+end
+
+function TestColumnLayout:testRenderWithWidth()
+  local column_padding = 15
+  local margin_bottom = 10
+  local margin_left = 5
+  local margin_right = 20
+  local margin_top = 25
+  local num_columns = 3
+  local width = 205
+
+  local expected_column_width = 50
+
+  local render_column_calls = {}
+
+  local layout = ColumnLayout.new {
+    column_padding = column_padding,
+    margin_bottom = margin_bottom,
+    margin_left = margin_left,
+    margin_right = margin_right,
+    margin_top = margin_top,
+    num_columns = num_columns,
+    width = width,
+    render_column = function (column)
+      table.insert(render_column_calls, {column})
+    end
+  }
+
+  self:stub(layout)
+  layout:render()
+
+  lu.assertEquals(layout._column_gap_calls, {{column_padding}, {column_padding}})
+  lu.assertEquals(layout._get_avail_width_calls, 0)
+  lu.assertEquals(layout._horiz_margin_calls, {{margin_left}})
+  lu.assertEquals(layout._vert_margin_calls, {
+    {margin_top, expected_column_width},
+    {margin_bottom, expected_column_width},
+    {margin_top, expected_column_width},
+    {margin_bottom, expected_column_width},
+    {margin_top, expected_column_width},
+    {margin_bottom, expected_column_width},
+  })
+  lu.assertEquals(layout._with_group_calls, num_columns)
+
+  lu.assertEquals(render_column_calls, {
+    {{num = 1, width = expected_column_width}},
+    {{num = 2, width = expected_column_width}},
+    {{num = 3, width = expected_column_width}},
+  })
+end
+
+function TestColumnLayout:testRenderNested()
+  local inner_renders = {}
+  local outer_renders = {}
+
+  local outer_layout = ColumnLayout.new {
+    column_padding = 0,
+    num_columns = 2,
+    width = 60,
+
+    render_column = function (outer_column)
+      table.insert(outer_renders, outer_column)
+
+      local inner_layout = ColumnLayout.new {
+        column_padding = 0,
+        num_columns = 3,
+        width = outer_column.width,
+
+        render_column = function (inner_column)
+          table.insert(inner_renders, inner_column)
+        end
+      }
+      self:stub(inner_layout)
+
+      inner_layout:render()
+    end
+  }
+  self:stub(outer_layout)
+
+  outer_layout:render()
+
+  lu.assertEquals(inner_renders, {
+    {num=1, width=10},
+    {num=2, width=10},
+    {num=3, width=10},
+    {num=1, width=10},
+    {num=2, width=10},
+    {num=3, width=10},
+  })
+
+  lu.assertEquals(outer_renders, {
+    {num=1, width=30},
+    {num=2, width=30},
+  })
+end
+
+
+
+os.exit(lu.LuaUnit.run())

reascripts/ReaSpeech/tests/TestReaSpeechUI.lua

@@ -8,6 +8,7 @@ require('ReaUtil')
 require('mock_reaper')
 
 require('source/AlertPopup')
+require('source/ColumnLayout')
 require('source/ReaSpeechActionsUI')
 require('source/ReaSpeechAPI')
 require('source/ReaSpeechProductActivation')