FD Portfolio Optimizer

Hybrid AI System — PSO Algorithm + Groq LLM + 2 Live Agents + FD vs Alternatives Comparator

---

Architecture

User Input (amount, risk, tenure, tax_slab)
           │
           ▼
┌──────────────────────────┐
│   Optimizer (Frontend)   │  ← Slider UI · risk profile · tenure · tax bracket
└────────────┬─────────────┘
             │ POST /optimize
             ▼
┌──────────────────────────┐
│  FastAPI  (api_new.py)   │  ← Runs on localhost:8002
└────────────┬─────────────┘
             │
    ┌────────┴────────────────────┐
    ▼                             ▼
┌──────────┐              ┌──────────────────┐
│  PSO     │              │  Groq LLM        │
│Algorithm │              │  llama-3.1-8b    │
│60 part.  │              │  Bank Selection  │
│200 iter. │              │  Rate Decision   │
└────┬─────┘              └──────┬───────────┘
     │                           │
     └──────────┬────────────────┘
                ▼
   ┌─────────────────────────┐
   │  FD vs Alternatives     │  ← Comparator Agent
   │  Comparator Agent       │    FD · Debt MF · SGB · RD · PPF
   │  (Fisher equation,      │    Pre-tax / Post-tax / Real return
   │   CPI inflation, tax)   │
   └────────────┬────────────┘
                ▼
  { allocation[], summary{}, comparison_rows[], bank_recommendation, rate_decision }
                │
                ▼ sessionStorage
     ┌─────────────────────┐
     │  Browser Dashboard  │
     │  (public/ folder)   │
     │                     │
     │  Portfolio.html     │  ← Allocation donut · FD ladder · Comparator table
     │  Agents.html        │  ← Bank Selection Agent · Rate Decision Agent
     │  Analytics.html     │  ← Bank table · maturity timeline
     └─────────────────────┘

---

Project Structure

AI/
├── api_new.py             ← Main FastAPI server (port 8002) — use this
├── api/
│   └── index.py           ← Legacy FastAPI (port 8001)
├── api_vercel.py          ← Vercel serverless version
├── vercel.json            ← Deployment config
│
├── fd_agents/
│   ├── api_new.py         ← FastAPI server (port 8002) ← START THIS
│   ├── simple_main.py     ← Direct pipeline (PSO + report, no CrewAI)
│   ├── main.py            ← CLI entry point
│   ├── tasks.py           ← Task definitions
│   ├── requirements.txt
│   ├── .env               ← GROQ_API_KEY goes here
│   └── agents/
│       ├── pso_optimizer.py     ← PSO algorithm (core engine)
│       ├── data_collector.py    ← Bank rate data + tools
│       ├── user_advisor.py      ← Report generation
│       ├── bank_selector_agent.py  ← Agent 1: Bank recommendation
│       ├── rate_decision_agent.py  ← Agent 2: Rate timing decision
│       └── comparator_agent.py     ← FD vs alternatives comparison
│
└── public/                ← Frontend (dark theme, Vercel deployed)
    ├── Login.html         ← Landing / sign-in
    ├── Overview.html      ← Portfolio summary dashboard
    ├── Optimizer.html     ← Input: amount · risk · tenure · tax bracket
    ├── Portfolio.html     ← Allocation · FD ladder · Comparator table
    ├── Analytics.html     ← Bank table · maturity timeline
    ├── Agents.html        ← 2 live AI agents · reasoning cards
    ├── Renewal.html       ← Renewal & tax (Form 15G)
    └── assets/
        ├── app.css        ← Dark theme styles
        └── shell.js       ← Nav · sidebar · routing

---

Quick Start

Step 1 — Get FREE Groq API Key

1. Go to https://console.groq.com
2. Sign up (free, no credit card)
3. API Keys → Create Key
4. Copy the key (starts with gsk_)

Step 2 — Setup

cd fd_agents
pip install -r requirements.txt

# Add your key to fd_agents/.env
echo "GROQ_API_KEY=gsk_your_key_here" > .env

Step 3 — Start API server

cd fd_agents
uvicorn api_new:app --host 0.0.0.0 --port 8002 --reload

Step 4 — Open frontend

# Serve public/ with any static server, e.g. Live Server in VS Code
# Open http://localhost:5500/public/Login.html

Step 5 — Run CLI (optional)

cd fd_agents
python simple_main.py

---

API Endpoints

# Health check
GET  http://localhost:8002/health

# Run optimization
POST http://localhost:8002/optimize
Content-Type: application/json

{
  "amount": 1000000,
  "risk_profile": "moderate",
  "tenure_months": 12,
  "tax_slab_pct": 30,
  "name": "Investor"
}

Response:

{
  "success": true,
  "allocation": [
    {
      "bank_name": "Bajaj Finance",
      "allocated_amount": 207755,
      "weight_percent": 20.78,
      "interest_rate": 8.35,
      "interest_earned": 17347,
      "maturity_amount": 225102,
      "dicgc_insured": true,
      "rating": "AAA"
    }
  ],
  "summary": {
    "total_investment": 1000000,
    "total_interest_earned": 82008,
    "total_maturity_amount": 1082008,
    "expected_annual_return_pct": 8.20,
    "banks_used": 8,
    "dicgc_fully_compliant": true
  },
  "comparison_rows": [
    { "instrument": "FD (Best)", "pre_tax_pct": 8.20, "post_tax_pct": 5.74, "real_return_pct": 0.23, "risk_score": 2, "liquidity_score": 7 },
    { "instrument": "SGB",       "pre_tax_pct": 10.55,"post_tax_pct": 9.80, "real_return_pct": 4.08, "risk_score": 3, "liquidity_score": 4 },
    { "instrument": "PPF",       "pre_tax_pct": 7.10, "post_tax_pct": 7.10, "real_return_pct": 1.52, "risk_score": 1, "liquidity_score": 3 }
  ],
  "comparison_text": "...",
  "inflation_used": 5.5,
  "bank_recommendation": "...",
  "rate_decision": "..."
}

---

FD vs Alternatives Comparator

Compares Fixed Deposits against 4 other instruments using tax-adjusted and inflation-adjusted real returns:

Instrument	Tax Treatment	Notes
FD (Best)	Taxed at income slab	DICGC insured up to ₹5L per bank
Debt Mutual Fund	Taxed at income slab (post Apr 2023)	No lock-in, high liquidity
SGB (Sovereign Gold Bond)	Gold appreciation tax-free at maturity; 2.5% coupon taxed at slab	8-year maturity
Recurring Deposit	Taxed at income slab	Monthly commitment
PPF	EEE — fully exempt	15-year lock-in

Real return formula: ((1 + nominal) / (1 + inflation)) - 1 (Fisher equation)

CPI inflation fetched live from data.gov.in; falls back to 5.5% if unavailable.

---

AI Agents

Agent	What it does
Bank Selection Agent	Recommends banks based on risk profile, tenure, ratings
Rate Decision Agent	Advises book-now vs wait based on RBI policy cycle
Comparator Agent	Calculates post-tax real returns across 5 instruments

---

8 Partner Banks

Bank	Type	Rating	12mo Rate
Bajaj Finance	NBFC	AAA	8.35%
Shriram Finance	NBFC	AA+	8.30%
Suryoday SFB	Small Finance	AA	8.25%
Mahindra Finance	NBFC	AA+	8.20%
Unity SFB	Small Finance	AA	8.15%
Utkarsh SFB	Small Finance	AA	8.10%
Shivalik SFB	Small Finance	A+	8.00%
Jana SFB	Small Finance	A+	7.90%

---

Cost Breakdown

Resource	Cost
Groq API (llama-3.1-8b-instant)	FREE
PSO Algorithm	FREE (pure Python)
FastAPI + Uvicorn	FREE
Vercel hosting	FREE tier
Total	Rs 0

---

Why Not Full CrewAI?

CrewAI builds ~29,000 token prompts per pipeline run. Groq free tier allows 6,000 tokens/minute — a single CrewAI run exceeds that limit 5×.

The hybrid approach uses ~500 tokens per run (98% reduction):

• PSO handles all math (no LLM needed)
• Comparator agent runs pure Python (no LLM)
• One direct Groq call for bank & rate advisory
• Result: same quality, always within free tier limits

---

Built with PSO Algorithm · Groq Llama 3.1 · FastAPI · Blostem Platform