bug fixes

relevance + diversity evals, batched scoring, math tooltips

Prompt imporved

sensing calibration + ink via ollama

updated prompt

air-writing

added affect and gesture fix

-

-

refactored and added synthetic data

Add voice + air-writing conflict resolution

Stop blocking /chat on evals; let the UI poll for them

Streaming candidate picker + side-index feedback loops

Session priors (P(bucket) + P(type)) + inline eval UI

Add faithfulness + authenticity evals; run scoring off the response path

Add MMR reranking with conversation-context query fusion

imultimodal sensing → real stylistic constraints

Add PRESENT_STATE intent + head-shake turnaround

Route sub-intents to their own pools, rip out the LLM intent router

added deeper memories and new users.

dropped more bloat

Remove library bloat

Added scaffolding for evals

Code refactor for extensibility