101 lines
5.9 KiB
Markdown
101 lines
5.9 KiB
Markdown
# Execution support prototype
|
||
|
||
The shared sidebar opens `/support` (Messages) and `/journal` (Journal), alongside
|
||
Reader and Research. These are UI-only prototypes with fictional contacts
|
||
and in-memory state. Reloading resets the prototype state. Tasks, communication status, outcomes, and
|
||
journal entries are shared between routes for the current browser session.
|
||
No messages, calendar events, medical records, or Rensheng pages are written.
|
||
|
||
## Reference and scope
|
||
|
||
The functional reference is the user's life brief
|
||
`briefs/2026-09-25-executive-support-design.md`, together with the public
|
||
[Beeper Copilot concept](https://beeper-copilot-lp.vercel.app/).
|
||
The prototype adopts the execution flows, not private facts from the life repo
|
||
or the reference site's deployment and privacy claims.
|
||
|
||
- Communication: unified conversations, a maximum of three initial actions,
|
||
deferral, waiting, dismissal, and a full list; source messages, summaries,
|
||
editable reply styles, explicit review, and simulated sending.
|
||
- Scheduling: choosing a candidate and reviewing a simulated calendar entry.
|
||
- Context: person notes with links back to the sample conversation and a local
|
||
record of the simulated outcome. No permanent trait or goal is inferred.
|
||
- Journal: free notes, resumption points, user-entered medication and weight
|
||
records; occurrence and recording times, append-only corrections, and a
|
||
selectable summary preview for a consultation.
|
||
|
||
The intended memory boundary remains Rensheng: execution events can eventually
|
||
be compiled into durable context after source comparison and user correction.
|
||
This prototype neither implements that compiler nor creates a second personal
|
||
context database. Missing records remain unknown; there are no medication,
|
||
dose, weight-target, or treatment recommendations.
|
||
|
||
The UI uses existing shadcn Base UI components and neutral theme tokens.
|
||
Shared account management continues to use the existing connection system;
|
||
the new communication and journal prototypes do not consume those accounts.
|
||
|
||
## Daily home
|
||
|
||
`/` presents a greeting, a large input, and a Tiimo-inspired flexible day flow grouped as
|
||
Start here, Routines, and If there’s room. There are no fixed time slots.
|
||
Tasks can be reordered, completed, deferred, restored, or followed as routine
|
||
steps. Communications open their Messages page at
|
||
`/support?view=all&conversation=<id>`; browser Back returns to the home with
|
||
shared outcomes intact. Journal entries and corrections appear on the home, and
|
||
reader samples open for reading. Research remains a separate work surface.
|
||
|
||
Activity blocks expand in place into controls suited to their content: routine
|
||
steps, a task's next action and resumption note, or a weight input that records
|
||
an entry in Journal. Completed and deferred activities stay in their original positions with visible
|
||
status and undo controls. Measurements retain the recorded value in place, and
|
||
reply previews retain their outcome, so the user need not remember what was done. Expanded activity selection
|
||
uses the `activity` URL parameter. This remains a session-only mock; a reload
|
||
restores navigation, not recorded activity data.
|
||
|
||
Messages uses a Beeper-inspired person list and visible original conversation,
|
||
with reply drafting and review next to source-backed context. On narrow screens
|
||
the list and conversation become separate views. Inbox, set-aside, waiting and
|
||
archive filters describe communication state; a simulated reply does not prove
|
||
that a separate promise was fulfilled. References include the saved
|
||
[Beeper Copilot captures](references/beeper-copilot/README.md),
|
||
[Beeper Desktop](https://blog.beeper.com/2025/02/24/try-out-the-new-beeper-desktop-and-ios-beta/),
|
||
and [open-loop research](gbrain-open-loops-research.md).
|
||
|
||
The input currently uses a deterministic English/Japanese mock interpreter, not an LLM.
|
||
It proposes typed actions for task creation/status/reordering, routine steps,
|
||
contact reply/review or deferral, notes/weight/medication records, and navigation.
|
||
Confirming a proposal dispatches the corresponding operation. Unknown or
|
||
ambiguous input does not change state. Message sending still goes through the
|
||
existing simulated review UI; this prototype performs no external send.
|
||
|
||
Examples: `Add grocery shopping`, `Defer Morning routine`, `Reply to Aoki`,
|
||
`Note: Resume from page 2`, `Weight: 65.2 kg`, and `Open Reader`.
|
||
|
||
The interaction reference is [Tiimo visual planning](https://www.tiimoapp.com/product/visual-planning),
|
||
reviewed on 2026-09-28: combined routines and tasks, flexible reordering, small
|
||
steps, and natural-language planning. Unlike its timed scheduling options, this
|
||
prototype deliberately keeps the day free of precise times as requested.
|
||
|
||
The home greeting includes a short prose brief derived from current task and
|
||
communication state. Completed or deferred items leave the brief; the next
|
||
contact opens directly from the text. A current resumption note takes precedence
|
||
over the optional-task suggestion. This is local templated copy, not an LLM report.
|
||
|
||
## Jev reply review
|
||
|
||
The message thread keeps the reply composer at the bottom on wide screens;
|
||
summary and source context live in the side panel. Expand Reply suggestions to
|
||
compare candidate wording before editing.
|
||
|
||
Opening Review reply calls an authenticated server function using AI SDK
|
||
`experimental_evaluate` with `typesafe-ai/jev`. Set `AI_GATEWAY_API_KEY` on
|
||
the server. The draft, recipient role, and original sample messages are sent
|
||
through Vercel AI Gateway to TypeSafe AI. The model evaluates whether the reply
|
||
answers the request, stays grounded, and uses an appropriate tone. The UI shows
|
||
boolean probabilities, not a calibrated pass/fail verdict or generated advice.
|
||
Editing the draft invalidates its evaluation. Missing configuration and provider
|
||
failures show an unavailable state; no mock approval is substituted. Review
|
||
results do not trigger sending, which remains a simulation.
|
||
|
||
See [Vercel’s Jev integration](https://vercel.com/changelog/typesafe-ai-jev-now-available-on-ai-gateway).
|