Skip to content

Instantly share code, notes, and snippets.

@JarbasAl
Last active September 25, 2022 14:32
Show Gist options
  • Select an option

  • Save JarbasAl/c20a25f2de6e58aee9e5be94ef7fd267 to your computer and use it in GitHub Desktop.

Select an option

Save JarbasAl/c20a25f2de6e58aee9e5be94ef7fd267 to your computer and use it in GitHub Desktop.
TODOs, WIP, Ideas....

Random urls and utils

TODO

OCA

https://github.com/OpenVoiceOS/ovos-config-assistant

  • extend to also use ovos_config package
  • make a UI using PyWebIO similar to local backend UI
  • multiple-core support
    • installed in same device, ovos-core is a dependency and always available
    • manage multiple cores, native neon support (auto config button)
    • print and manage all ovos.conf + mycroft.conf aspects per core
    • allow mix and match, eg, neon-speech and all other services stock ovos (?)
  • OPM support, populate options based on installed plugins
    • dropdown to select saved configs
    • prepopulated from OPM interface per lang
    • allow use to save their own configs
    • form based on template (TODO, needs work OPM side)
  • rich intent plugin support
    • pre train models
    • benchmarks
    • match all installed skills and suggest best config
  • rich wake word model support
    • vosk plugin auto config with just a string
    • pocketsphinx auto config, threshold testing interface
    • import datasets
      • precise community
      • local backend
    • export datasets
      • download a dataset ready for precise training
    • auto generate snowboy ww models
      • configurable trainer url
      • try several combos of 3 samples, benchmark all models, pick best
      • allow resuming (remember previous used combinations and benchmarks)

New plugins

coref

intents

postag

  • stanza / natex-py - can be provided together with the intent plugin described above

Ideas

New ovos pipelines

OOVSkills

  • out-of-vocabulary intents skill class - OOVSkills
  • triggered before common_query framework
  • match an intent using something like https://github.com/GT4SD/zberta
    • this gives us an intent classification without a registred handler
    • "handle this question" becomes "handle this intent somehow"
    • OOVSkills can use this info for a richer common_query experience
  • send the intent_tag + utterance to OOVSkills

Tech demos

  • FM output
    • broadcast OVOS as a radio station
      • no speakers connected to pi
      • any old Radio for playback
        • integrate with old cars
      • broadcast to multiple devices
      • automated radio station fun, live interaction with AI?
    • https://github.com/F5OEO/rpitx
  • VHF output via composite
  • FPS game
  • RPG game
    • dedicated embbeded ovos-core as a lib
      • hivemind like integration for multiple users (in-game NPC)
      • shared_skills + skills per NPC, TTS per NPC
    • you are the player character / part of the game world
    • each NPC connected to a ovos-core/hivemind instance
  • Voice Frotz
  • Voice First Interactive Fiction
    • resume dev on https://github.com/JarbasAl/TNaLaGmes game engine
      • more inbuilt objects / intents
    • make a few prototypes
      • oregon trail port (WIP) <- simple game engine loop
      • interactive world with objects <- rooms / map / objects base interactive fiction engine
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment