Roadmap
This is a rough idea of the goals. I move basic on what people request on Discord.
I’m working on
Voice mode. TTS streaming while LLM is still working. Then going to work to conversation mode. Then Twilio support. This is a 3-week stretch of stuff in June I am working on.
Possible next up
- Voice mode
- ASR, no wakeword, full voice
- Phone calls
- Wakeword free voice conversation
- dynamic stop/resume of speech when user interrupts
- Virtual Employees
- Won’t give full details yet here- this is imminent
- Create goal
- Oriented as EASY interface for business admins
- Avatar v2
- This was a tech demo, want to expand
- Avatars walk around virtual rooms to areas that match the task they are on
- Animated stuff
- Lipsync – supported in our avatar on patreon
- Games – we have an easter egg to already walk around first person
- Story Mode v2
- I took a swing and I missed. Scrap this system.
- Zork-style still (room based).
- Less story, more rules for the AI, let the AI fill in details more
- Story scene is minimal, just list of items
- Story engine SEPARATE from the chat history, bah
- Privacy Mode v2
- Wasn’t perfect
- It should be a simple toggle
- Eyeball icon locks to local or something
- Not private-locked chats, private locked UI as a whole
- Still keep private only chats
- Emotion system
- cant talk details yet
- Guided Tour
- in-app tour of features
- In the form of achievements – 10 tool calls, etc
- Shows all core features
- Memory system
- Multiple memory systems are out there that can be integrated
- mempalace is one (thanks @Cyber from Discord)
- possibly add memory decay to memories not accessed
- sometimes important mems dont get read much and disappear— bad
- LAYERED — I think of the primal memories: T-rex roar, the smell of popcorn, THEN the plot of the movie.
- Obsidian seems helpful
- Translations
- A plugin that translates the app in real time
- No language files, just real-time UI translation
- Tool updates
- Chain tools together (output -> input)
- Create tool workflows editor so we call one tool and it chains
- Make tools dynamic (only show whats needed, have AI call “list tools” and enable subsets.)