Screenshot Scanner

One tap. Your screenshots, documents, folders, and repos become searchable memory for your AI agents.

You upload something — a screenshot, a .md file, a source folder, a GitHub repo URL. The app reads it with a vision model, decides what it is, and files it into your Agent Memory hub as a clean, searchable wiki page, skill, memory note, or code graph. No manual formatting. No API memorization. No junk drawer.

---

Try it now — 3 commands

git clone https://github.com/Mikey25050/screenshot-scanner.git
cd screenshot-scanner
./configure.sh    # answers a few questions, writes your config
docker compose up -d --build

Open http://localhost:8090 and upload a screenshot. Done.

New here? Read the full walkthrough: SETUP.md — what you need, both install options, how to verify, costs, and troubleshooting.

---

What you need (before you start)

You need	Why
Docker (or Python 3.10+)	to run the app
An Agent Memory hub	that's where everything gets filed (e.g. TencentDB Agent Memory)
Your own DeepSeek API key	the vision "brain" — your key, your bill, nothing charged through this project
3 IDs from your hub (team-…, usr-…, agt-…)	the app uses them to file content into your team

That's it. No database, no other services, no pip install — the server is pure Python standard library.

---

How it works

you (browser) ──upload──▶ scanner ──vision + classify──▶ your hub
   screenshot / doc / dir / repo      wiki · skill · memory · code-graph
                                    (searchable by all your agents)

• Screenshots are compressed in your browser before upload (10× smaller)
• Low-confidence items are parked for review — never silently filed
• Repos and folders become symbol-level code graphs your agents can query

---

What it does

• 📷 Screenshots — vision-classified and filed automatically
• 📄 Documents — .md / .txt / .json, no vision needed
• 🗂 Folders & repos — turned into searchable code graphs
• 🧠 Skills — upload a SKILL.md, filed as a skill
• 🗄 Pending inbox — review low-confidence items, file or discard
• 📊 Mission Control dashboard — live stats at /dashboard
• 📱 Phone-friendly — installable web app (PWA), works over VPN

---

Docs

Doc	What's in it
SETUP.md	full installation guide for your own server (Docker or bare metal)
ROADMAP.md	what's planned next (v6.1, v7.0)
SECURITY.md	security model + how to report a vulnerability

Configuration: copy config.example.json → config.json, or use SCANNER_* environment variables (all documented in SETUP.md).

---

API at a glance

Endpoint	What it does
POST /scan	upload a screenshot or document → classified and filed
POST /skill	upload a SKILL.md → filed as a skill
POST /graph	{"url": …} repo or {"path": …} folder → code graph
GET /graphs	list your team's code graphs
GET /pending · POST /review	review inbox (file / discard)
GET /dashboard · GET /stats	Mission Control + live stats
GET /ping	health check

---

Security

• Run it on a trusted network (LAN / VPN). Don't expose it to the public internet without protection.
• Optional token: set SCANNER_AUTH_TOKEN and every request needs it (except health checks).
• Credential files (.ssh, .aws, .env*, keys…) are never ingested — see SECURITY.md.
• Your keys live only in your own .env / config.json — both gitignored, never committed.

License

MIT — see LICENSE.