shakedzy / companion Goto Github PK

Currently, the tutor is only aware of the language taught and spoken (dialect are solely used for Google TTS service). Making it more aware of the dialects might enhance the lessons.

Currently there is no error handling, and the code assumes everything just works. Here are some suggestions:

• When ChatGPT fails, perhaps display a red message saying "oh no, try again"
• How to display a Whisper error message?
• How to handle Watson errors?
• When Google Translate fails, perhaps display a red message to instead of translation
• All continuous threads need error handling

See new API release notes: openai/openai-python#742

When alt+r are pressed, recording mode toggles. But it doesn't happen when the focus is on the text box.

Instead of configuring the app with a yaml file, add a new "init" page to select languages, voice and tutor.
Probably best handles once #2 is done

Audio files are generated, but no sound is heard. Might be due to bas installation of PyAudio or another dependency.

New lines (\n) are not displayed in chat messages, which sometime cause the tutor messages to look odd.

When testing new voices in /setup page, the engine is using the already-configured voice, which causes malfunctions in certain languages.

This can be converted to a Streamlit app, which can run online.
Another "welcome" page will have to be added, to ask the user for his API tokens.

Looking at split_to_sentences, this phrase:

Très bien! Vous avez bien formé la négation en utilisant \"ne...pas\". Maintenant, pouvez-vous me donner la forme interrogative de la phrase \"il parle espagnol\"?

Splits to:

['Très bien',
 ' Vous avez bien formé la négation en utilisant "ne',
 '',
 '',
 'pas"',
 ' Maintenant, pouvez-vous me donner la forme interrogative de la phrase "il parle espagnol"',
 '']

while it should be:

['Très bien',
 'Vous avez bien formé la négation en utilisant "ne...pas"',
 'Maintenant, pouvez-vous me donner la forme interrogative de la phrase "il parle espagnol"?']

It'll be great to have the option to load previous chats and continue them

This might spice up the lessons, as it will allow discussion on current events

I got this error for the first time on my mac

Was fixed by brew insstall pyaudio

I just thought you might want to add it to your README or something.

ERROR: Could not build wheels for pyaudio, which is required to install pyproject.toml-based projects

of Google TTS will work better, it will reduce the number of services required.
See Google TTS Console here

If a user sends a message in his native language, allow to translate it to the learning language

When the tutor provides an assignment, for example:

Hier, nous (faire) ____________ une promenade dans le parc.

The underscores are read one by one ("underscore, underscore, underscore, ..."). Saying "underscore" or perhaps "mmm" should be the right thing.

Probably best handled once #2 is done

This is in order to allow running companion on a remote server and access it via mobile

Allow playing audio to be stopped or paused.

• Expand text-box when text no longer fits the box
• Allow Shift+Enter to make new line

It would feel more realistic if the tutor speaks first

Hi I tried the code with a different version of Python and still am having the same issue there is no response and I can't see or hear what the ai sending hope you will fix this issue or guide us to a solution for that
thank you

It'll be easier to read

When correcting or explaining, tutor might not ask further question, ending the session flow. for example:

Non, la forme interrogative correcte pour cette phrase serait "Aimons-nous le chocolat?" (avec le verbe "aimer" conjugué avant le sujet "nous").

Instead of creating config file, create a page which assists the user to configure the app, and create the config file behind the scene.

As can be seen from the following memory snippet:

  {
    "role": "assistant",
    "content": "Bonjour Shaked! Comment ça va? Pour notre leçon aujourd'hui, je vous propose trois sujets qui conviennent à votre niveau DELF A1.2: \n1. Les verbes réguliers au présent \n2. Les articles définis et indéfinis \n3. Les adjectifs possessifs. \nQu'est-ce que vous préférez?",
    "recording": [
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_0.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_1.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_2.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_3.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_4.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_5.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_6.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_7.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_8.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_9.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_10.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_11.mp3"
    ],
    "user_recording": null
  },
  {
    "role": "user",
    "content": "Et Je choisi les verbe réguliers.",
    "recording": [
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_0.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_1.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_2.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_3.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_4.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_5.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_6.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_7.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_8.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_9.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_10.mp3",
      "/Users/shakedzy/repos/companion/tmp/bot_speech_1_11.mp3"
    ],
    "user_recording": "/Users/shakedzy/repos/companion/tmp/user_recording_2.mp3"
  },

Current authentication to Google is done by installing the APK and authenticating through it. This will not be possible when running as an app on Streamlit

For #13 (select tutor from an init page, more tutor profile pictures are required. Probably 3 females and 3 males should be sufficient). Currently only a single female profile exist.
Should be created with DALL-E/Midjourney.

Understand and explain how to install prerequisites on Windows

There are no docstrings in the code