Project Overview
FinSage is a conversational AI finance coach that unifies a person’s bank, credit, and investment data, runs goal-based simulations, and returns clear, actionable steps to reach milestones like buying a home or funding education. Think of a small matchstick that becomes a torch—FinSage turns scattered numbers into a bright, goal‑driven plan using live data via Fi MCP, an AI microservice, and an interactive React dashboard.

The Problem
Most finance apps report balances; they don’t tell whether a goal is on track or what to change today. Users juggle multiple accounts, SIPs, and loans without a confident, personalized plan, leading to guesswork and delayed decisions.

The Solution
FinSage connects financial accounts through Fi MCP, understands questions conversationally, simulates outcomes, and prescribes precise adjustments. Ask “Can I buy a ₹50L flat in 5 years?” and get “Increase your SIP by ₹6,000/month, keep an 11–12% return assumption, and you’ll close the shortfall by Month 53,” plus a chart and shortfall breakdown.

What We’re Building (Hackathon Scope)
A chat-first experience that answers “Will I make it?” and “What must I change today?”

A goal planner with SIP, timeline, and what‑if simulations (rate, tenure, contribution).

Fi MCP–backed data ingestion to ground advice in real balances, holdings, and flows.

A dashboard with progress, shortfall, and portfolio visuals; export to CSV/JSON/PDF.

Key Features
Conversational AI (LangChain + Gemini) for finance‑aware, stepwise guidance

SIP auto‑optimizer that proposes exact monthly deltas to hit a target on time

What‑if scenarios for contribution, tenure, return; median/P10/P90 paths

Portfolio‑aware insights using Fi MCP (bank, investments, cards, loans)

Smart alerts when a goal drifts off course, with transparent next steps

One‑click exports for reports and sharing

How It Works (Architecture)
Frontend (React + Tailwind + Plotly): Chat UI, goal flows, interactive charts

Backend (Node API Gateway): Auth, validation, routing, and orchestration

AI Service (FastAPI):

LangChain + Gemini for conversation and reasoning

Simulation engine (NumPy/Pandas) for SIP, timeline, projections

Export service for CSV/JSON/PDF artifacts

Fi MCP Integration: Secure connectors and transformers to normalize multi‑account data into a unified profile consumed by the AI service

**High‑Level Flow: **
User → React Frontend → Node API Gateway → AI Service (chat, simulate, export) → Fi MCP (live data) → Back to UI as insights, charts, and actions.

Why It’s Different
Goal‑first, action‑oriented: moves beyond static dashboards to “what to change today.”

Data‑grounded: advice is driven by Fi MCP’s consolidated, real‑time financial context.

Transparent math: clear formulas, assumptions, and scenario ranges (not black‑box).

Designed for momentum: alerts, nudges, and exports make action easy and repeatable.

Success Metrics
Time‑to‑insight: <3 seconds for chat + simulation response

Prescriptive coverage: >80% of common retail goals addressed (home, car, education)

Accuracy: ±1–2% vs benchmark calculators on SIP FV and timeline math

Engagement: % of users who act on at least one recommendation within a week

Risks & Mitigations
Data variability: normalize via transformers and robust schema validation

Model drift/ambiguity: strict prompt templates and guardrails; fallbacks to deterministic math

Performance: cache portfolio snapshots; precompute common scenarios; stream partial results

Compliance/security: scoped tokens, least‑privilege access, and secure secret management

Roadmap (Post‑Hackathon)
Multi‑goal optimization (cashflow‑aware) and debt payoff strategies

Deeper risk profiling and scenario stress tests

Email/WhatsApp nudges and calendarized savings tasks

Advisor‑friendly reports and shared planning links