RxSignal

FDA Regulatory Intelligence Platform — search, monitor, and analyze public FDA data in real time with AI-assisted pharmacovigilance.

---

What is RxSignal?

RxSignal is a full-stack regulatory intelligence platform built on top of the FDA's public API (openFDA). The goal is to turn raw FDA safety data — scattered across multiple endpoints and dense JSON formats — into interactive dashboards, search tools, and AI-assisted analyses that researchers, healthcare professionals, and pharmacovigilance teams can actually use without writing a single line of code.

The platform is built around three principles: real data (everything comes directly from the FDA, nothing is fabricated), privacy (the AI model runs locally via LM Studio, no data leaves your machine), and speed (a smart PostgreSQL cache protects API rate limits and keeps the interface fast).

---

Features

📊 Dashboard

Platform-wide overview with live KPIs: total adverse event volume, top reported drugs, recall breakdown by severity class, and medical device event categories — all in one view.

⚠️ Adverse Events Explorer

Search the FAERS database (FDA Adverse Event Reporting System) for any drug. Visualizes the most frequently reported reactions in a ranked list, report volume over time as a line chart, serious vs. non-serious breakdown, and exports results to CSV.

💊 Drug Intelligence

Look up any FDA-registered drug by brand name, generic name, or NDC code. Displays the full FDA label (SPL) in a structured accordion — indications, warnings, boxed warnings, contraindications, dosage — alongside its regulatory approval history from Drugs@FDA.

🚨 Recall Monitor

Unified enforcement feed across drugs, devices, and food simultaneously. Filterable by domain and severity class (Class I / II / III). Refreshed every hour via cache.

🔬 Device Safety

Queries the MAUDE database (Manufacturer and User Facility Device Experience) to surface medical device adverse event reports broken down by event type — malfunction, injury, death.

🔍 NDC Lookup

Resolves any drug by its 10 or 11-digit NDC code, or by name. Returns active ingredients, dosage form, route of administration, labeler, and packaging variants.

🤖 AI Assistant

A chat interface that accepts plain English questions — "what are the most serious side effects of Ozempic?" — translates them into precise openFDA queries using Mistral 7B running locally via LM Studio, executes the search, and returns a plain-language safety summary. All processing happens on your machine.

---

Tech Stack

Layer	Technology	Role
Frontend	Next.js 14 (App Router) + Tailwind CSS	UI, routing, SSR
Backend	Django 5 + Django REST Framework	REST API, business logic, openFDA proxy
Database	PostgreSQL 16	Persistence, FDA query cache
AI	Mistral 7B via LM Studio	NL→query translation, safety summarization
Data	openFDA API (api.fda.gov)	Source of truth — public FDA data
Auth	JWT via SimpleJWT	Stateless authentication
Infra	Docker Compose	Full local orchestration

Data Flow

Browser → Next.js → Django REST → PostgreSQL (cache check)
                                       ↓ cache miss
                                  openFDA API → cache write → response

                    ↓ AI path
              LM Studio (Mistral) → FDA query → results → summary

---

Getting Started

Prerequisites

• Docker Desktop — for the database
• Python 3.11+
• Node.js 20+
• LM Studio — optional, only needed for the AI Assistant

1. Clone the repository

git clone https://github.com/your-username/rxsignal.git
cd rxsignal

2. Start the database

docker compose up db -d

3. Django Backend

cd backend
python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env            # edit if needed
python manage.py migrate
python manage.py runserver
# → http://127.0.0.1:8000

Recommended: Register a free API key at open.fda.gov/apis/authentication and add it to .env as OPENFDA_API_KEY=your_key. This raises your rate limit from 1,000 to 120,000 requests/day.

4. Next.js Frontend

cd frontend
npm install
cp .env.local.example .env.local
npm run dev
# → http://localhost:3000

Windows users: Make sure NEXT_PUBLIC_API_URL=http://127.0.0.1:8000 in your .env.local — Windows resolves localhost to IPv6 (::1) which Django doesn't bind to by default.

5. AI Assistant (optional)

1. Open LM Studio
2. Download Mistral 7B Instruct v0.3 (Q4_K_M quantization recommended)
3. Go to Local Server tab → Start Server
4. The assistant connects automatically at http://localhost:1234/v1

---

API Reference

FDA Data

Method	Endpoint	Description
GET	/api/fda/adverse-events/	FAERS adverse event reports
GET	/api/fda/drug-labels/	FDA drug labels (SPL)
GET	/api/fda/ndc/	NDC directory lookup
GET	/api/fda/drugsfda/	Drugs@FDA approval history
GET	/api/fda/recalls/	Unified drug/device/food recalls
GET	/api/fda/device-events/	MAUDE device adverse events
GET	/api/fda/dashboard/	Aggregate KPI stats

AI

Method	Endpoint	Description
POST	/api/ai/query/	NL prompt → FDA query → summary
POST	/api/ai/summarize/	Summarize a set of FDA results
GET	/api/ai/history/	Recent AI query log

Auth

Method	Endpoint	Description
POST	/api/auth/register/	Create account
POST	/api/auth/login/	Login → JWT tokens
POST	/api/auth/token/refresh/	Refresh access token
GET	/api/auth/me/	Current user profile

---

Cache Strategy

All openFDA responses are cached in PostgreSQL to protect rate limits and keep the UI fast:

Endpoint	TTL
Adverse events	6 hours
Drug labels	24 hours
NDC directory	24 hours
Recalls / enforcement	1 hour
Count / aggregate queries	12 hours

---

Project Structure

rxsignal/
├── docker-compose.yml
├── backend/
│   ├── config/                  # Settings, URLs, WSGI
│   ├── apps/
│   │   ├── core/                # Auth, users, watchlists
│   │   ├── fda/                 # openFDA client, cache, endpoints
│   │   │   ├── client.py        ← HTTP client + cache-aside logic
│   │   │   ├── models.py        ← FDAQueryCache model
│   │   │   └── views.py         ← All FDA REST views
│   │   └── intelligence/        # Mistral / LM Studio
│   │       ├── client.py        ← NL→query + summarization
│   │       └── views.py         ← /api/ai/ endpoints
│   ├── requirements.txt
│   └── .env.example
└── frontend/
    ├── app/
    │   ├── page.tsx             # Dashboard
    │   ├── adverse-events/      # FAERS explorer
    │   ├── drugs/               # Drug search + detail page
    │   ├── recalls/             # Recall monitor
    │   ├── devices/             # Device safety
    │   ├── ndc/                 # NDC lookup
    │   └── assistant/           # AI chat
    ├── components/
    │   ├── layout/              # Sidebar, Topbar
    │   └── ui/                  # Shared components
    └── lib/
        └── api.ts               # HTTP client → Django

---

Roadmap — Phase 2

• Drug shortage monitor (/drug/shortages.json)
• Device 510(k) + PMA regulatory pathway explorer
• Food adverse events module (CAERS)
• Celery + Redis for scheduled signal detection jobs
• Email / webhook alerts for watched drugs
• Frontend authentication UI
• Saved search and watchlist persistence

---

Data Sources

All data is sourced from openFDA — a public initiative by the U.S. Food & Drug Administration. Data has not been validated for clinical or production use. RxSignal is a research and analytics tool, not a substitute for professional medical guidance.

---

License

MIT