Fix whipser setup (#7)

* Create new PodcastBuddy.cache_dir, fix whisper setup, and use the cache dir for all whisper calls
* Move ShowAssistant and CoHost initialization into CLI.run so that it doesn't setup the session before it's configured

lib/podcast_buddy.rb

@@ -39,7 +39,12 @@ def start_session(name: nil)
       @session = Session.new(name: name)
     end
 
+    def cache_dir
+      @cache_dir ||= "#{ENV["HOME"]}/.buddy"
+    end
+
     def setup
+      Dir.mkdir PodcastBuddy.cache_dir unless Dir.exist?(PodcastBuddy.cache_dir)
       SystemDependency.auto_install!(:git)
       SystemDependency.auto_install!(:sdl2)
       SystemDependency.auto_install!(:whisper)

lib/podcast_buddy/cli.rb

@@ -20,13 +20,14 @@ class CLI
     def initialize(argv)
       @options = parse_options(argv)
       @tasks = []
-      @show_assistant = PodcastBuddy::ShowAssistant.new
     end
 
     def run
       configure_logger
       configure_session
       setup_dependencies
+      @show_assistant = PodcastBuddy::ShowAssistant.new
+      @co_host = PodcastBuddy::CoHost.new(listener: @show_assistant.listener)
       start_recording
     rescue Interrupt
       handle_shutdown
@@ -77,9 +78,12 @@ def configure_session
       if @options[:name]
         base_path = "#{PodcastBuddy.root}/tmp/#{@options[:name]}"
         FileUtils.mkdir_p base_path
-        PodcastBuddy.session = @options[:name]
+        PodcastBuddy.start_session(name: @options[:name])
         PodcastBuddy.logger.info PodcastBuddy.to_human("Using custom session name: #{@options[:name]}", :info)
-        PodcastBuddy.logger.info PodcastBuddy.to_human("  Saving files to: #{PodcastBuddy.session}", :info)
+        PodcastBuddy.logger.info PodcastBuddy.to_human("  Saving files to: #{PodcastBuddy.session.base_path}", :info)
+      else
+        PodcastBuddy.start_session
+        PodcastBuddy.logger.info PodcastBuddy.to_human("Saving files to: #{PodcastBuddy.session.base_path}", :info)
       end
     end
 
@@ -92,7 +96,6 @@ def setup_dependencies
 
     def start_recording
       Sync do |task|
-        @co_host = PodcastBuddy::CoHost.new(listener: @show_assistant.listener)
         show_assistant_task = task.async { @show_assistant.start }
         co_host_task = task.async { @co_host.start }
         @tasks = [

lib/podcast_buddy/configuration.rb

@@ -86,7 +86,7 @@ def openai_client
 
     # @return [String]
     def whisper_command
-      "./whisper.cpp/stream -m ./whisper.cpp/models/ggml-#{whisper_model}.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
+      "#{PodcastBuddy.cache_dir}/whisper.cpp/build/bin/stream -m #{PodcastBuddy.cache_dir}/whisper.cpp/models/ggml-#{whisper_model}.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
     end
   end
 end

lib/podcast_buddy/system_dependency.rb

@@ -21,12 +21,12 @@ def self.resolve_whisper_model(model)
     end
 
     def self.model_downloaded?(model)
-      File.exist?(File.join(PodcastBuddy.root, "whisper.cpp", "models", "ggml-#{model}.bin"))
+      File.exist?(File.join(PodcastBuddy.cache_dir, "whisper.cpp", "models", "ggml-#{model}.bin"))
     end
 
     def self.download_model(model)
       originating_directory = Dir.pwd
-      Dir.chdir("whisper.cpp")
+      Dir.chdir("#{PodcastBuddy.cache_dir}/whisper.cpp")
       PodcastBuddy.logger.info "Downloading GGML model: #{PodcastBuddy.whisper_model}"
       PodcastBuddy.logger.info `bash ./models/download-ggml-model.sh #{PodcastBuddy.whisper_model}`
     ensure
@@ -36,7 +36,7 @@ def self.download_model(model)
     def installed?
       PodcastBuddy.logger.info "Checking for system dependency: #{name}..."
       if name.to_s == "whisper"
-        Dir.exist?("whisper.cpp")
+        Dir.exist?("#{PodcastBuddy.cache_dir}/whisper.cpp")
       else
         system("brew list -1 #{name} > /dev/null") || system("type -a #{name}")
       end
@@ -46,12 +46,14 @@ def install
       PodcastBuddy.logger.info "Installing #{name}..."
       originating_directory = Dir.pwd
       if name.to_s == "whisper"
-        Dir.chdir(PodcastBuddy.root)
+        Dir.chdir(PodcastBuddy.cache_dir)
+        PodcastBuddy.logger.info "Setting up whipser.cpp in #{PodcastBuddy.cache_dir}/whipser.cpp"
         PodcastBuddy.logger.info `git clone https://github.com/ggerganov/whisper.cpp`
         Dir.chdir("whisper.cpp")
         PodcastBuddy.logger.info "Downloading GGML model: #{PodcastBuddy.whisper_model}"
-        PodcastBuddy.logger.info `bash ./models/download-ggml-model.sh #{PodcastBuddy.whisper_model}`
-        PodcastBuddy.logger.info `make && make stream`
+        PodcastBuddy.logger.info `bash #{PodcastBuddy.cache_dir}/whisper.cpp/models/download-ggml-model.sh #{PodcastBuddy.whisper_model}`
+        PodcastBuddy.logger.info "Building whipser.cpp with streaming support..."
+        PodcastBuddy.logger.info `cmake -B build -DWHISPER_SDL2=ON && cmake --build build --config Release`
       else
         `brew list #{name} || brew install #{name}`
       end

spec/cli_spec.rb

@@ -24,6 +24,7 @@
 
   describe "#run" do
     let(:show_assistant) { instance_double(PodcastBuddy::ShowAssistant) }
+    let(:listener) { instance_double(PodcastBuddy::Listener) }
 
     before do
       allow(PodcastBuddy).to receive(:logger).and_return(Logger.new(nil))
@@ -33,6 +34,8 @@
       allow(show_assistant).to receive(:start)
       allow(show_assistant).to receive(:stop)
       allow(show_assistant).to receive(:generate_show_notes)
+      allow(show_assistant).to receive(:listener).and_return(listener)
+      allow(listener).to receive(:subscribe)
     end
 
     it "configures logger and session" do

spec/configuration_spec.rb

@@ -63,13 +63,13 @@
 
   describe "#whisper_command" do
     it "returns the expected command string" do
-      expected_command = "./whisper.cpp/stream -m ./whisper.cpp/models/ggml-small.en-q5_1.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
+      expected_command = "#{PodcastBuddy.cache_dir}/whisper.cpp/build/bin/stream -m #{PodcastBuddy.cache_dir}/whisper.cpp/models/ggml-small.en-q5_1.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
       expect(config.whisper_command).to eq(expected_command)
     end
 
     it "incorporates changes to whisper_model" do
       config.whisper_model = "test_model"
-      expected_command = "./whisper.cpp/stream -m ./whisper.cpp/models/ggml-test_model.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
+      expected_command = "#{PodcastBuddy.cache_dir}/whisper.cpp/build/bin/stream -m #{PodcastBuddy.cache_dir}/whisper.cpp/models/ggml-test_model.bin -t 8 --step 0 --length 5000 --keep 500 --vad-thold 0.75 --audio-ctx 0 --keep-context -c 1 -l en"
       expect(config.whisper_command).to eq(expected_command)
     end
   end

spec/fixtures/tmp/2024-12-19_13-39-47/transcript.log

@@ -0,0 +1,2 @@
+test transcription
+test transcription

spec/fixtures/tmp/session/transcript.log

@@ -1 +1,2 @@
-Existing transcript
+Existing transcripttest transcription
+test transcription