v0.7.0: Geo-tagging, place awareness, and departure reminders

Version 0.7.0 delivers a full geo-tagging feature that makes Chalie location-aware. Transcripts and episodes are now stamped with GPS coordinates and resolved place names, and a new subconscious step learns location-based behavioural patterns from the user’s history.

Every transcript write automatically picks up the device location from locale_service, and episodes aggregate the dominant place visited during their window. The data graph gains a KIND_PLACE entity backed by a new PlaceAbility, letting users save, list, and delete named places with cosine-supersede contradiction handling.

The scheduler extends its input schema with a destination_location field. When an event includes a place, Chalie estimates travel time using a layered speed model—historical median, heartbeat deltas, then a 30 km/h fallback—and prepends a departure advisory to the notification.

Under the hood, a GeoPatternProcessor runs as a newly added seventh step in the subconscious worker. It performs a single-pass LLM extraction of location-based behavioural patterns, driven by a dedicated cursor that requires at least 30 transcripts before firing.

On the UI side, three regressions from the brain redesign were patched: the providers-only lock now correctly restricts navigation until a provider is saved, model lists auto-fetch on modal open and platform tab switches, and the sidebar brand logo once again links to the chat home. Architecture docs were updated to reflect all new abilities, steps, and glossary terms.

• Transcripts auto-tagged with GPS coordinates and place names on every write

• New PlaceAbility supports save, list, get, and delete of named places

• Departure reminders with layered speed estimation (historical median → heartbeat deltas → 30 km/h default)

• LLM geo pattern learning added as subconscious worker step 7, with a 30-transcript minimum delta

• Providers-only lock restored with automatic unlock on first provider save