buildAiContext previously loaded table metadata sequentially in a
for-loop (3 API calls per table × up to 50 tables = 150 serial
round-trips). Now uses Promise.all to fetch all tables' columns,
indexes, and foreign keys concurrently, reducing total wait time
from O(n × latency) to O(latency).
Introduce src-web (axum HTTP server) exposing all 43 database operations
as REST APIs with SSE streaming for AI, transfers, and SQL file execution.
Frontend adapts automatically between Tauri and HTTP backends via api.ts.
Includes password authentication, Docker packaging, and login page.
- Chat-style conversation UI with message bubbles
- Multi-turn dialog with context history
- SQL code blocks with "Apply" and "Copy" buttons
- Toggle via Sparkles button in toolbar
- Settings dialog preserved
t8y2