Implement AI orchestration wedding demo

docs/api.md

@@ -0,0 +1,37 @@
+# API
+
+Primary orchestrator/demo endpoints:
+
+- `POST /tasks`
+- `PATCH /tasks/{task_id}`
+- `POST /notes`
+- `POST /reservations`
+- `PATCH /reservations/{reservation_id}`
+- `GET /dashboard`
+- `GET /tools`
+- `POST /transcriptions`
+- `GET /transcriptions/{job_id}`
+- `GET /pending-actions`
+- `PATCH /pending-actions/{action_id}`
+- `POST /pending-actions/{action_id}/approve` generates the API call; it does not send it
+- `POST /pending-actions/{action_id}/reject`
+- `WS /ws/transcript`
+
+Generated Wed backend tools:
+
+- `create_task` -> `POST /tasks/`
+- `update_task` -> `PATCH /tasks/{task_id}`
+- `create_note` -> `POST /notes/`
+- `update_note` -> `PATCH /notes/{note_id}`
+- `create_guest_reservation` -> `POST /reservations/guest`
+- `update_reservation` -> `PATCH /reservations/{reservation_id}`
+
+Transcript chunks:
+
+```json
+{
+  "speaker": "Bride",
+  "timestamp": "2026-06-26T12:00:00Z",
+  "text": "Create a task for Maria to confirm the photographer."
+}
+```