Inou - Medical Imaging Second Opinion Platform

Vision

Patient-facing medical data platform with AI-powered second opinions. Upload your scans, connect to AI, get peace of mind.

Target user: Non-technical parent/spouse/caregiver who knows nothing about medical imaging but wants answers.

Brand: inou.com ("eye-now" - vision/seeing + immediate second opinion)

Scope: Human + veterinary imaging. Not just imaging - labs, vitals, supplements, genome, photos, observations.

---

Quick Reference

Service	URL
Portal	https://inou.com
Viewer	https://inou.com:8767
API	https://inou.com/api/*

Machine	IP	Role
Ubuntu Server	192.168.1.253	Source, Linux builds, Production
Mac	local	macOS bridge builds
Windows	192.168.1.9	Windows bridge builds

---

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         User                                     │
└───────────────────────────┬─────────────────────────────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        │                   │                   │
        ▼                   ▼                   ▼
┌───────────────┐   ┌───────────────┐   ┌───────────────┐
│ Portal (443)  │   │ Viewer (8767) │   │ Claude Desktop│
│ Web UI, Auth  │   │ DICOM viewer  │   │ + Bridge      │
└───────┬───────┘   └───────┬───────┘   └───────┬───────┘
        │                   │                   │
        │                   │                   │
        └───────────────────┼───────────────────┘
                            │
                    ┌───────▼───────┐
                    │  REST API     │
                    │  (viewer)     │
                    └───────┬───────┘
                            │
                    ┌───────▼───────┐
                    │   SQLite DB   │
                    │   + Objects   │
                    └───────────────┘

Services (2 total):

• Portal (443) - Web UI, authentication, proxies /api/* to viewer
• Viewer (8767) - DICOM viewer, REST API, image serving

Components

Component	Binary	Purpose
Portal	inou-portal	Web UI, auth, profile management, proxies API
Viewer	inou-viewer	DICOM viewer, REST API, image serving
Bridge	inou_bridge_*	Connects Claude Desktop to REST API (per-platform)
Lab Scrape	inou-lab-scrape	MyChart lab results scraper
Lab Import	inou-lab-import	Lab results importer
DICOM Import	import-dicom	DICOM to PNG importer

Bridge

The bridge connects Claude Desktop (stdio/MCP protocol) to the Inou REST API:

Claude Desktop → inou_bridge (stdio) → HTTPS → inou.com/api/*

Configuration (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "inou": {
      "command": "/Users/USERNAME/bin/inou_bridge_darwin_arm64",
      "args": ["--server=https://inou.com", "--account=ACCOUNT-GUID"]
    }
  }
}

The bridge translates MCP tool calls from Claude Desktop into REST API requests. There is no separate MCP server - the viewer's REST API serves all data.

---

API Tools

Tools available via the bridge (prefixed with inou:):

Tool	Description
list_profiles	Show patient profiles accessible to account
list_studies	Show all studies for selected profile
list_series	Show series in a study (filter by T1, T2, AX, etc.)
list_slices	Show slices with position info
fetch_image	Get image as base64 with metadata
list_lab_tests	List available lab test names
get_lab_results	Get lab results with date range/latest filters
get_version	Check bridge/server versions

Database Schema

-- Core tables
patients (id, guid, name, dob, sex, created_at)
studies (id, patient_id, study_date, study_desc, folder_path)
series (id, study_id, series_number, series_desc, modality, orientation_type)
slices (id, series_id, instance_number, slice_location, orientation_type, 
        tr, te, png_path, rows, cols, pixel_spacing_row, pixel_spacing_col,
        image_position_x/y/z, image_orientation)

-- Auth tables
accounts (id GUID, email, email_hash, name, last_login, created_at)
account_profiles (account_id, profile_id, role, email, created_at)
sessions (token, account_id, profile_id, created_at, expires_at)
auth_codes (email, code, created_at, expires_at)

-- Lab tables
lab_results (id, profile_id, test_name, value, units, ref_range_low, 
             ref_range_high, collected_at, source)

---

Build System

All binaries built with FIPS 140-3 cryptographic compliance (requires Go 1.24+).

Build Flow

Source (.253)  ──sync──►  Mac  ──build──►  darwin binaries  ──deploy──►  .253
     │                                                                      ▲
     └────────sync──────►  Win (.9)  ──build──►  win binaries  ──deploy────┘

Binaries

Binary	Platform	Build Location
inou-portal	Linux	.253
inou-viewer	Linux	.253
import-dicom	Linux	.253
inou-lab-scrape	Linux	.253
inou-lab-import	Linux	.253
inou_bridge_darwin_arm64	macOS ARM	Mac
inou_bridge_darwin_amd64	macOS Intel	Mac
inou_bridge_win_amd64.exe	Windows x64	.9
inou_bridge_win_arm64.exe	Windows ARM	.9
inou_bridge_win_386.exe	Windows x86	.9

Build Commands

Linux (on .253):

cd ~/dev/inou
make deploy

Mac bridges:

~/dev/inou/build-mac.sh

Windows bridges (from Mac):

sshpass -p 'Helder06' ssh johan@192.168.1.9 "C:\dev\inou\build.bat"
sshpass -p 'Helder06' scp johan@192.168.1.9:'C:/dev/inou/bin/inou_bridge_win_*.exe' /tmp/
scp /tmp/inou_bridge_win_*.exe johan@192.168.1.253:/tank/inou/bin/

Verify FIPS:

ssh johan@192.168.1.253 "/tank/inou/status.sh"

---

Service Management

ssh johan@192.168.1.253
/tank/inou/start.sh    # Start all services
/tank/inou/stop.sh     # Stop all services
/tank/inou/status.sh   # Check status and FIPS compliance

Logs

tail -f /tank/inou/logs/portal.log
tail -f /tank/inou/logs/viewer.log

---

Directory Structure

Source (.253)

~/dev/inou/
├── portal/           # Web portal + auth
├── viewer/           # DICOM viewer + REST API
├── import_dicom/     # DICOM importer
├── mcp-client/       # Bridge client source
├── scrape_mychart/   # Lab scraper + importer
├── lib/              # Shared crypto library
├── go.mod, go.sum
├── Makefile
├── README.md
├── journal.md
└── build-mac.sh

Production (.253)

/tank/inou/
├── bin/              # All compiled binaries
├── data/inou.db      # SQLite database
├── uploads/          # User uploads by GUID
├── objects/          # Converted images (GUID-based paths)
├── logs/             # Service logs
├── lang/             # Translation files
├── templates/        # Portal HTML templates
├── static/           # CSS, assets
├── static/robots.txt  # Search engine rules
├── master.key        # Encryption key
├── start.sh, stop.sh, status.sh

---

Security

Encryption

• FIPS 140-3 compliant (Go 1.24+ native module)
• AES-256-GCM encryption at rest
• TLS 1.3 for transport
• Deterministic encryption for searchable fields

Authentication

• Passwordless: email + one-time code (10 min expiry)
• Session cookies with configurable expiry
• Account GUID isolation (no cross-account access)

Data Isolation

• All queries filtered by profile_id derived from session
• SQL injection protection (parameterized queries)
• Access validation on all endpoints

---

Product Design

Terminology

Term	Meaning
Account	Human with email who logs in
Profile	A person whose data is stored (patient)
Exam/Study	One imaging session
Series	One sequence within an exam (T1, T2, FLAIR)

User Flow

1. Landing page → Create account (email + code)
2. Dashboard → Add profile (name, DOB, sex)
3. Profile → Upload exam (DICOM zip)
4. Connect Claude Desktop → "Analyze my inou"
5. AI reviews images, answers questions

Design

• Modern, warm, not clinical
• IBM Plex Sans font
• Desert color palette: warm beige (#F5EDE4), dark brown (#4A3728), terracotta (#C4704B)

Privacy (prominent)

• Images NEVER used to train AI
• Data NEVER shared
• Always encrypted
• You own your data - delete anytime

i18n

Languages: English, Русский, Nederlands, Español, Français, Deutsch

Detection priority: URL param → Cookie → Accept-Language → English

---

Data Types

Implemented

• Medical imaging (DICOM)
• Lab results

Planned

• Medical records / clinical notes (PDF)
• Medications / prescriptions
• Genetic data (VCF files)
• Vitals (BP, HR, SpO2, weight, glucose)
• Supplements
• Symptoms / pain journal

---

Roadmap

High Priority

• Position-based cross-series fetch
• find_slice_at_position tool
• Window/level controls (presets for T1, T2, FLAIR, DWI)
• Better error messages (SESSION_EXPIRED)

Medium Priority

• Batch montage fetch
• Measurement tools (distance, area)
• Series grouping/filtering
• Report system (save markdown, PDF export)

Lower Priority

• Expanded anatomical regions
• Report templates
• Persistent findings/annotations

Infrastructure

• Dedicated Inou server (separate from home lab)
• Encrypted ZFS pool for PHI
• Audit logging

---

Troubleshooting

FIPS Build Fails

• Ensure Go 1.24+ installed
• Use GOFIPS140=v1.0.0 environment variable
• FIPS requires native builds (no cross-compilation)

Bridge Connection Issues

• Check Claude Desktop config JSON syntax
• Verify account GUID is correct
• Test API: curl https://inou.com/api/version

Service Issues

# Check if running
pgrep -a inou

# Check logs
tail -100 /tank/inou/logs/portal.log

# Restart
/tank/inou/stop.sh && /tank/inou/start.sh

---

Session Log

See journal.md for detailed session history.

---

Future: Unified Data Model

See docs/DATA_MODEL_V2.md for planned schema migration to single observations table.