c2141a127a
- Remove 'coverage' from worklist description across all docs - Add whitepaper v2, documentation gap whitepaper, gate demo, sample - Add TERAX.md, Claude Code settings, test-data generator - Add settings.local.json to .gitignore Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
127 lines
4.8 KiB
Markdown
127 lines
4.8 KiB
Markdown
# Signal — Terax AI Context
|
|
|
|
> Dev and terminal work only. Strategy, brand, and pilot decisions stay in Claude Code.
|
|
> Use @path references in the AI Composer to pull in specific files before asking questions.
|
|
|
|
---
|
|
|
|
## What Signal Does
|
|
|
|
Signal is a B2B CGM documentation readiness tracker for DMEPOS suppliers. A supplier exports their order data as a CSV (from Brightree or similar), Signal evaluates each patient against documentation requirements (6-month qualifying visit, PECOS enrollment, SWO, resupply eligibility), and produces a stoplight report so staff can prioritize outreach before claims are submitted.
|
|
|
|
Signal identifies gaps. It does not contact anyone.
|
|
|
|
**Core value:** Shifts supplier staff time from reactive (denial appeals after the fact) to proactive (preventing denial conditions before supplies ship).
|
|
|
|
---
|
|
|
|
## PHI Architecture — Non-Negotiable
|
|
|
|
- STTIL never stores patient names, SSNs, DOBs, or contact info
|
|
- Sole crosswalk key: `patient_id` (the DME's internal MRN or account number)
|
|
- AI/calculation layer sees: `patient_id`, `device_type`, `shipment_date`, `quantity`, `payer` — nothing else
|
|
- All logs hash `patient_id` before storing — never raw
|
|
- If a proposed change would require storing additional PHI fields, stop and flag it
|
|
|
|
---
|
|
|
|
## Level 1 Stack (Active)
|
|
|
|
| Layer | Tool |
|
|
|-------|------|
|
|
| Hosting | Hostinger VPS — data never leaves VPS |
|
|
| Language | Python (FastAPI) |
|
|
| Database | PostgreSQL (encrypted at rest, audit logs) |
|
|
| Orchestration | n8n self-hosted (24-hr batch trigger) |
|
|
| Notifications | Self-hosted Mailcow — encrypted email to DME staff only |
|
|
| Frontend | signal-ui/ |
|
|
|
|
---
|
|
|
|
## Active Files — Start Here
|
|
|
|
```
|
|
python-backend/core/coverage_calculator.py # main eligibility logic
|
|
python-backend/core/audit_logger.py # hashed logging
|
|
python-backend/config/payer_rules.json # wear days and payer rules — reference for all calculations
|
|
signal-ui/ # frontend
|
|
research/dmepos-research-v3.md # full market context + verified stat index (updated May 2026)
|
|
research/cgm-market-research-synthesis-v1.md # stat verification summary (May 2026)
|
|
docs/hipaa-deployment-analysis-v1.md # compliance reference
|
|
docs/sttil-brand-system-v1.md # brand (read-only reference)
|
|
pitch/ # pitch assets and whitepaper
|
|
```
|
|
|
|
---
|
|
|
|
## Market Context — Key Verified Stats (May 2026)
|
|
|
|
Use these. Do not invent or derive alternatives.
|
|
|
|
| Stat | Source | Use For |
|
|
|------|--------|---------|
|
|
| 30.86% pre-pay error rate | CGS MAC Jurisdiction B Q2 2024 | Denial prevention pitch — measures near-submission risk |
|
|
| 32.8% error rate / 68.6% from docs | CERT 2019 | LinkedIn and public copy — "nearly 1 in 3", "over two-thirds from docs" |
|
|
| 67.6% absent documentation | CMS 2024 MLN | Whitepaper / gate framing — post-payment audit |
|
|
| $1.9B DMEPOS improper payments | OIG FY2024 | Market context only |
|
|
|
|
**Do not use:** "94.2%" (derived, not citable), "63% write-off" (modeled, not citable).
|
|
|
|
Full verified stat table: `research/dmepos-research-v3.md` Section 14.
|
|
|
|
---
|
|
|
|
## Urgency Anchors (Active)
|
|
|
|
- Synapse Health mandatory for Medicare CGM: August 1, 2026
|
|
- PA exemption cycle (90% affirmation rate): June 1, 2026 first cycle
|
|
- 7 new HCPCS codes added to Required PA List: April 13, 2026
|
|
- DMEPOS enrollment moratorium in effect: February 27, 2026
|
|
- CGM competitive bidding: January 1, 2028
|
|
|
|
---
|
|
|
|
## What Is Tabled — Do Not Build
|
|
|
|
| Item | Why |
|
|
|------|-----|
|
|
| Dexcom OAuth API | Requires vendor agreement + PHI scope expansion |
|
|
| Prescriber fax automation | Phase 2 — Level 1 manual outreach sufficient |
|
|
| Patient-facing SMS | PHI to third party — needs BAA + consent layer |
|
|
| Consortium / Level 2-3 features | Needs 15+ paying Level 1 suppliers first |
|
|
| Convex | Not needed until real-time sync requirements emerge |
|
|
|
|
---
|
|
|
|
## VPS — Common Terax Commands
|
|
|
|
```bash
|
|
# Connect
|
|
ssh root@72.62.134.75
|
|
|
|
# Check running Signal services
|
|
ssh root@72.62.134.75 "docker ps --format 'table {{.Names}}\t{{.Status}}\t{{.Ports}}'"
|
|
|
|
# View Signal API logs (live)
|
|
ssh root@72.62.134.75 "docker logs signal-api --tail 50 -f"
|
|
|
|
# Restart Signal API
|
|
ssh root@72.62.134.75 "docker compose -f /opt/signal/docker-compose.yml restart signal-api"
|
|
|
|
# Check Caddy (reverse proxy)
|
|
ssh root@72.62.134.75 "docker logs caddy --tail 30"
|
|
|
|
# Run local tests before deploy
|
|
cd python-backend && pytest core/ -v
|
|
```
|
|
|
|
---
|
|
|
|
## Dev Rules
|
|
|
|
- Work Level 1 scope only
|
|
- Reference `payer_rules.json` for all wear-day calculations — do not hardcode
|
|
- Use "resupply" not "refill" (CGM supply allowances are exempt from Medicare refill rules)
|
|
- Never suggest adding PHI fields beyond `patient_id`
|
|
- Flag any vendor integration that would require a new BAA
|
|
- If asked about Phase 2/3 scope, acknowledge and defer — do not spec or build
|