786team · Hackathon

Customer-service AI with computer vision, built with Lovable for the Emerge Hackathon (April 2026).

Team

• Louis Rodriguez (@LouisRodriguez12101815)
• Lalla
• Corey Duke

Project (working concept)

An agentic customer-service assistant that understands photos, not just text. Customers upload images of damaged / wrong items; a VLM-powered agent assesses the issue, looks up the order, applies policy, and resolves the ticket (refund / replacement / escalate) in seconds. Lovable scaffolds the UI + Supabase backend; vision calls go to a hosted VLM (OpenAI / Anthropic / Google). Not final — scope may pivot during the build.

Links

• Lovable workspace: TBD (paste link once workspace is created)
• NotebookLM / shared docs: TBD
• Demo: TBD

Repo layout

web/              # Lovable-generated frontend (auto-synced, TBD)
server/           # Agent loop + API clients
  clients/        # one stub per external service (OpenAI, Gemini, Telegram, ...)
scripts/          # Dev utilities
  check_apis.py   # pings every configured API and prints status
docs/             # Architecture notes, APIs, team inputs
  apis.md         # Signup links + env vars for every service
  team-inputs/    # Raw team brainstorms (Lalla, Corey)

Quickstart — localhost demo in one command

git clone https://github.com/786team/hackathon.git
cd hackathon
./scripts/run.sh

Then open http://localhost:8000. You'll see:

• API connectivity status for all 12 services (red until you add keys)
• An Analyze pane that runs vision on any image URL
• A Market comps pane that fans out to eBay Sold + Google Shopping + Google Lens No keys required — the app falls back to realistic mocks automatically. Fill in .env and refresh to see real data.

API connectivity CLI

For a terminal-only check:

source .venv/bin/activate
python scripts/check_apis.py

See docs/apis.md for signup links.

Working agreements

• main is protected. All changes via PR.
• One approving review before merge (any teammate can approve).
• No force pushes, no direct commits to main.
• Never commit .env files or API keys. Use .env.example for docs.
• Use issue templates for bugs and feature ideas.
• Small, frequent commits > big end-of-day dumps.

License

MIT — see LICENSE.