RiskShield — AI-Powered Risk Management System

Analyze SRS & BRD documents with 8 parallel AI agents, full visualization dashboard, and an interactive Copilot. 100% free. No backend. No data leaves your browser.

---

Tech Stack

Layer	Technology
UI	React 18 + Recharts
AI	Gemini 2.5 Flash (Google AI Studio — free)
RAG	In-memory TF-IDF (browser-native, zero cost)
Build	Vite 5
Hosting	Any static host (Vercel, Netlify, GitHub Pages)

---

Project Structure

riskshield/
├── .env                          # Default env vars (safe to commit)
├── .env.local                    # Your secrets — DO NOT COMMIT
├── .env.example                  # Template for new contributors
├── .gitignore
├── package.json
├── vite.config.js
├── index.html
│
└── src/
    ├── main.jsx                  # React entry point
    ├── App.jsx                   # Root component / screen router
    ├── config.js                 # Central config (reads .env)
    │
    ├── styles/
    │   ├── global.css            # Global resets + keyframes
    │   └── tokens.js             # Design tokens (colors, fonts)
    │
    ├── agents/
    │   ├── geminiClient.js       # Gemini API wrapper + retry logic
    │   ├── prompts.js            # All LLM system prompts
    │   └── pipeline.js           # Multi-agent orchestration
    │
    ├── utils/
    │   ├── rag.js                # TF-IDF RAG engine
    │   └── fileParser.js         # File → text extraction
    │
    ├── hooks/
    │   └── usePipeline.js        # React hook for pipeline state
    │
    ├── components/
    │   ├── ui/
    │   │   └── Atoms.jsx         # Shared UI primitives
    │   └── dashboard/
    │       ├── DashboardVisuals.jsx   # 9 charts + executive summary
    │       ├── RiskTable.jsx          # Filterable risk register
    │       └── Copilot.jsx            # Conversational AI assistant
    │
    └── views/
        ├── UploadView.jsx        # Landing + document upload
        ├── AnalysisView.jsx      # Live pipeline progress
        └── DashboardView.jsx     # Tabbed dashboard container

---

Quick Start

1. Get a Free Gemini API Key

Go to aistudio.google.com → Create API Key → Copy it.

Free tier: 10 RPM · 250 RPD · 1M TPM — more than enough.

2. Clone & Install

git clone https://github.com/your-org/riskshield.git
cd riskshield
npm install

3. Set Environment Variables

cp .env.example .env.local
# Edit .env.local and paste your Gemini API key

VITE_GEMINI_API_KEY=AIzaSy...your_key_here

4. Run

npm run dev
# Opens at http://localhost:3000

---

Agent Pipeline

Upload → DocParser → RAGIndexer → RiskAnalyst
                                       ↓
                           ┌─── PolicyAgent (parallel) ───┐
                           └─── MitigationAgent (parallel) ┘
                                       ↓
                               SummaryGenerator → Dashboard

Agent	LLM?	Runs	Purpose
DocParser	✗	Parallel	Extract text from SRS/BRD/Policy files
RAGIndexer	✗	Sequential	Build TF-IDF vector index
RiskAnalyst	✓	Sequential	Identify risks across 6 dimensions
PolicyAgent	✓	Parallel	Validate risks against company policy
MitigationAgent	✓	Parallel	Generate mitigation strategies
SummaryAgent	✓	Sequential	Generate executive summary

Peak concurrent LLM calls: 2 → well within 10 RPM free limit.

---

Environment Variables Reference

Variable	Default	Description
VITE_GEMINI_API_KEY	(required)	Your Google AI Studio API key
VITE_GEMINI_MODEL	gemini-2.5-flash	Gemini model to use
VITE_GEMINI_MAX_TOKENS	8192	Max tokens per response
VITE_GEMINI_TEMPERATURE	0.15	Lower = more deterministic
VITE_RAG_CHUNK_SIZE	500	Words per document chunk
VITE_RAG_CHUNK_OVERLAP	60	Overlap between chunks
VITE_RAG_TOP_K	7	Top-K chunks to retrieve
VITE_AGENT_STAGGER_POLICY_MS	1200	Delay before policy agent (ms)
VITE_AGENT_STAGGER_MITIGATION_MS	2400	Delay before mitigation agent (ms)
VITE_AGENT_SUMMARY_DELAY_MS	1000	Delay before summary agent (ms)

---

AWS Integration (Optional — Free Tier)

RiskShield has a single boolean toggle that switches between fully local and cloud-backed mode.

The toggle

In src/config.js (or via .env.local):

// Hard-code it directly:
export const USE_AWS = true;   // ← flip this

// Or use env var:
// VITE_USE_AWS=true  in .env.local

USE_AWS	Behaviour
false (default)	100% local, no AWS needed, works offline
true	Activates S3, DynamoDB, Lambda, CloudWatch

What each AWS service does

Service	Free tier	Role in RiskShield
Lambda	1M req/mo · 400K GB-sec	Handles all SDK calls — browser never holds credentials
S3	5 GB · 20K GET · 2K PUT/mo	Stores raw docs + risk report JSON with presigned download URL
DynamoDB	25 GB · 25 WCU/RCU/mo	Persists sessions (30-day TTL) + full Copilot chat history
CloudWatch	5 GB ingestion/mo	Records all 6 agent log events per session

Setup (3 steps)

Step 1 — Deploy AWS resources (one command)

aws cloudformation deploy \
  --template-file aws/cloudformation.yaml \
  --stack-name riskshield \
  --capabilities CAPABILITY_IAM \
  --parameter-overrides AllowedOrigin=https://your-domain.com

Step 2 — Upload the Lambda code

cd lambda
npm install
zip -r ../lambda.zip .
aws lambda update-function-code \
  --function-name riskshield-pipeline \
  --zip-file fileb://../lambda.zip

Step 3 — Copy CloudFormation outputs to .env.local

# Get outputs
aws cloudformation describe-stacks --stack-name riskshield \
  --query "Stacks[0].Outputs"

# Then in .env.local:
VITE_USE_AWS=true
VITE_AWS_LAMBDA_URL=https://abc123.lambda-url.us-east-1.on.aws
VITE_AWS_S3_BUCKET=riskshield-reports
VITE_AWS_DYNAMO_TABLE=riskshield-sessions
VITE_AWS_REGION=us-east-1

Done. Restart npm run dev — the dashboard will show a ☁ RS-XXXX session badge and a ↓ S3 Report download button when AWS is active.

Architecture

Browser
  │
  ├─ Gemini API  (direct, always — LLM calls)
  │
  └─ Lambda Function URL  (only when USE_AWS=true)
       │
       ├─ S3          raw-docs/{sessionId}/*.txt
       │              reports/{sessionId}/report.json  → presigned URL
       │
       ├─ DynamoDB    PK=sessionId  SK=META | CHAT#<ts>
       │              TTL = 30 days
       │
       └─ CloudWatch  /riskshield/pipeline/{sessionId}

Credentials never touch the browser. The Lambda function URL is the only public endpoint; it holds an IAM role scoped to just our S3 bucket, DynamoDB table, and CloudWatch log group.

---

npm run build
# Output in /dist — deploy to any static host

---

License

MIT