Deploy Log
DeepSeek Chatbot
An AI chatbot that can read an entire GitHub repository and answer questions about it with clickable, line-accurate citations — built on DeepSeek V4 Pro with a custom context-selection engine, not just a chat wrapper.
repo-aware chat — cited, file-linked answers
tracing logic across files, line by line
code explanations with inline references
Import any public GitHub repo by URL and chat with the codebase
Clickable citations open a code snippet with a "View on GitHub" link
Real-time token-by-token streaming, no full-response waiting
Keyword-scored context selection under a configurable token budget
Full chat history persisted locally — no backend database
Custom cyan/teal design system with light, dark, and system themes
Golden Dragon
A full-stack restaurant ordering platform — real accounts, real Stripe checkout, table reservations, and an admin dashboard for running the business. Built and deployed as a complete production system, not a demo.
customer-facing homepage
categorized menu with cart
admin dashboard — live business stats
Online ordering with Stripe checkout or pay-at-pickup
Table reservations by date, time, and party size
Customer accounts with order history and saved addresses
Admin dashboard: orders, menu, reservations, and messages
JWT auth with bcrypt hashing and role-based access
Environment-based CORS across separately deployed services
▸ postmortem — what shipping this actually involved
- Configured Prisma against Supabase's connection pooler (transaction vs. session pooling) and worked around IPv4/IPv6 constraints on free-tier hosting
- Debugged CORS across a frontend and backend deployed as two separate services
- Tracked down a production-only React crash caused by an unguarded array assumption
- Built the Stripe payment flow to verify charges against Stripe's event log rather than trusting the dashboard alone