washai.us

Search for restaurants by name, cuisine, or location.

Get EVERYTHING about a restaurant in one call: store info, full menu with modifiers and prices, hours, services, AI instructions, guardrails, availability.

Place an order at a restaurant.

Make a reservation at a restaurant.

Add a customer to the restaurant waitlist.

Get revenue summary for a date range: gross/net revenue, orders, avg ticket, breakdowns by order type.

Get top selling menu items by revenue or quantity for a date range.

Compare sales between two periods (today vs yesterday, this week vs last, this month vs last).

Get menu items with prices, descriptions, and stock status.

Get all menu categories with item counts.

Get all items currently 86'd (out of stock).

86 (mark out of stock) or un-86 (mark back in stock) a menu item.

Get recent orders.

Get all active (non-completed, non-cancelled) orders grouped by status.

Search customers by name, phone, or email.

Get detailed info about a customer: spending, orders, visits, segments, notes.

Get top customers by spending or order count.

Get all reservations for today, grouped by status.

Get upcoming confirmed/pending reservations (today and future).

Get reservation analytics (total, confirmed, completed, cancelled, no-show) for a date range.

Get current waitlist queue status: how many waiting, notified, checked in, seated, and estimated wait time.

Get the active waitlist queue with names, party sizes, and wait times.

Get waitlist analytics: total, seated, no-shows, conversion rate, avg wait times.

Get all active delivery orders, grouped by joint fulfillment state (Preparing, Awaiting dispatch, Dispatching, Courier confirmed, On the way, etc.).

Get today's delivery summary: total deliveries, completed, in progress, revenue from deliveries.

Get full details for a specific order, including items, modifiers, and available status transitions.

Get the total number of orders placed today.

List all customer segments (VIP, regulars, new, at-risk, etc.) with member counts.

Get a list of menu items that are missing images or descriptions.

Get recent task runs for a location.

Create a new task run record.

Update an existing task run record (e.g., mark running → completed or failed).

Get pending background tasks for a location (created by dashboard delegation).

Update the status of a background task (mark as in_progress, completed, or failed).

Audit which OpenPOS features are enabled vs available for a location.

Get current SEO state: store description, cuisine tags, highlights, visibility settings, website meta tags.

Update store's AI-discoverable metadata: description, cuisine tags, highlights, price range, store instructions.

Deep menu performance analysis from analytics_item_performance.

Generate the "Daily Brief" report for a location — yesterday's revenue, orders, AOV, top items, day-over-day comparison.

Generate the weekly review report — week-over-week revenue, top movers, anomalies, recommended focus areas.

Get the AI Manager service configuration for a location: enabled services, schedules, communication channels, quiet hours.

Update the AI Manager service configuration for a location.

Low-level send to the store owner via email, SMS, or push — raw delivery, NO dedupe/cooldown/quiet-hours.

Record an internal event in the activity audit log.

Recent ai_action_queue entries (proposed/applied/rejected/expired) for this location — what the AI has been doing.

Sense the store — your complete live picture, start here.

The authoritative list of every action you can propose, with each one's exact payload shape (field names, types, enum values, required vs optional) — generated live from the validation schemas, so it never drifts.

Understand what's still open for the owner — to inform your play, NOT to notify (the system reminds the owner about these on cadence).

RECONCILE PAST WORK — call FIRST each cycle, before proposing anything new.

The store's active/proposed growth goal + progress, pace, and the journey of rungs achieved.

Reconcile the store's goal STATE before you propose tactics — call this FIRST on a goal-driven tick.

What the AI has delivered for this store over the period — proven revenue, labor saved, volume handled (the value ledger).

Proactive growth plays that currently fit this store (win-backs, promos, SEO/GEO, menu, catering, etc.), pre-filtered by store state.

Customers who have not ordered within the lapse window — win-back candidates.

Owner-set boundaries the agent MUST operate within: marketing budget tier/cap, hard-nos (things never to do), optional direction.

Run any CALENDAR-due periodic work for this store: daily/weekly/monthly manager reports + email digests.

Detect LLM operator work due for this store that YOU (the engine) should run: { welcome: bool (store newly configured, never onboarded), resumes: [{questionId, question, answer, resumeWith}] (owner answered a question you were waiting on — continue the in-flight skill/work) }.

Mark the welcome package as completed for this store (idempotency marker so get_due_operator_work stops returning welcome=true).

Mark an answered owner question as resumed (so get_due_operator_work stops returning it).

Owner-supplied knowledge entries (story, brand voice, constraints, etc.).

Recall the curated facts (owner-taught + AI-learned) most RELEVANT to your current decision — pass a query, tags, or a topic.

Record a durable fact you just learned about this store (e.g.

Recall persisted external research for this store (competitor_pricing, local_events, review_sentiment, peer_benchmark, cuisine_trends, ...).

Persist a research finding you just gathered (via WebSearch/WebFetch) so it's saved for future cycles instead of re-crawled.

This store's own sales rhythm from order history (default 90 days), bucketed in the store's LOCAL timezone: busiest/slowest day-of-week, busiest hour, full day + hour distributions, and the last-7-days-vs-prior-7 trend.

Current loyalty program(s) for this location (any non-archived).

Current promotions for this location.

Catering menu items for this location.

Reach the store owner for a NOVEL, owner-worthy event (a fresh insight, an anomaly, a notable result you just produced).

Engine-only: atomically claim the next pending invocation from the queue.

Engine reports successful completion of a pending invocation.

Engine reports failure.

Hand work to the remote AI agent for background processing.

See what delegations are in flight (still working), completed but not yet shown to owner, or recently done.

Get the public no-auth approval URL for a pending action id.

Re-check whether an applied action actually took effect — reads the change back from canonical state.

The universal write path — your hands.

Owner-present chat: apply an action immediately (low/medium risk) or queue as pending_confirmation (high risk).

Pose a question to the owner asynchronously.