← Back to SumQ
SumQ

Smart Unified Model Queries

One question. Five AI models. One clear, grounded, trustworthy answer.

Every AI model has different strengths, blind spots, and knowledge cutoffs. Asking only one model means getting one perspective — often confidently wrong. SumQ queries all of them simultaneously, grounds every answer in live web data, learns from your history, and synthesises a consensus with a deterministic confidence score.

Five models, queried simultaneously

Claude

claude-sonnet-4-6

Writing · Synthesis · Reasoning

ChatGPT

gpt-4o

Analysis · Code · Structured output

Gemini

gemini-3.6-flash

Long context · Current events

Perplexity

sonar-pro

Built-in live search · Citations

Grok

grok-3

Real-time X/Twitter data

All models are optional — SumQ works with whichever keys you configure. Consensus improves with more active models.

How it works — 10 steps

01

Intent Classification

Every query is classified across 12 semantic categories — news, research, technical, strategic, financial, creative and more — then routed to the right search providers and models automatically.

02

Live Search Grounding

Before querying any AI, SumQ searches the live web using Tavily, Serper, Brave and more in parallel. Every answer is grounded in current information, not stale training data.

03

User Memory + RAG

SumQ learns from your query history — inferring your interests, expertise level, and preferred depth. Personal documents you upload are indexed and automatically injected as context when relevant.

04

Parallel LLM Queries

All active models receive the grounded prompt simultaneously. Each brings distinct strengths: Grok adds real-time X data, Perplexity adds its own search layer, Claude leads synthesis.

05

Trust Intelligence

Each model response is scored A–D before synthesis. Citations are verified against sources. Hallucination risk is flagged with specific claims. All deterministic — no AI guesswork.

06

Consensus + Confidence

A 6-signal formula computes confidence — semantic overlap (35%), source grounding (25%), query type (20%), consistency (10%), model count, and search mode. Claude synthesises into one authoritative cited answer.

07

SumQ News

Multiple sources, one story. SumQ clusters articles about the same event, removes duplicates, extracts facts, compares source framing, and produces one clear evidence-based summary. Evidence confidence rated HIGH / MODERATE / LIMITED.

08

Portfolio Intelligence

Upload your CAS statement from CAMS or KFintech to import all your mutual funds at once. SumQ fetches live NAVs from AMFI, calculates P&L, and runs AI portfolio analysis across 5 models simultaneously.

09

Trend Intelligence

Pick a niche and platform — SumQ generates a structured weekly trend report: rising/declining topics, hashtags, content opportunities, India-specific signals, and week-over-week diff.

10

WhatsApp Task Agent

Full SumQ pipeline accessible via WhatsApp. Set reminders in natural language, get answers on the go, type "portfolio" to get your P&L summary. Powered by Meta WhatsApp Cloud API.

Phase 2 · Trust Intelligence

Response Ranking

Every model scored A–D on 7 signals before synthesis — length, citations, structure, specificity, directness, readability, relevance. Best responses weighted more heavily.

Citation Verification

Cross-checks every [1][2] citation in the final answer against source content using term overlap analysis. Reports a verification rate.

Hallucination Detection

Flags specific statistics, absolute claims ("always", "never", "the only"), and date assertions that have no source grounding. Low/Medium/High risk.

Deterministic Confidence

6-signal formula replaces AI self-reported confidence. Reproducible — same inputs always produce the same score. Calibrated thresholds: 72+ High, 50–71 Medium, below 50 Low.

Phase 3 · Knowledge & Personalisation

User Memory

Automatically infers your interests, expertise level, and depth preference from query history. Silently enriches every prompt with your context — no setup needed.

RAG Document Search

Upload your documents and SumQ indexes them into searchable chunks. Every query automatically retrieves the most relevant passages and injects them alongside live web data.

SumQ News

Multi-source story clustering — facts, source comparison, evidence confidence

Portfolio Intelligence

CAS import, live NAV, P&L, AI portfolio analyst across 5 models

Trend Intelligence

Pick a niche and platform — SumQ generates a structured weekly trend report with rising/declining topics, hashtags, content opportunities, and India-specific signals. Weekly diffs show what changed.

Cross-device Sync

API keys and preferences sync across all devices via Google account. Open SumQ on mobile and your keys are ready — no re-entry needed.

Deterministic confidence scoring

Unlike other platforms that ask their model to self-report confidence, SumQ computes it from six measurable signals:

Model agreement35%

Vocabulary overlap between model responses

Source grounding25%

Number of live sources backing the answer

Query type20%

Factual vs opinion baseline

Response consistency10%

Similarity in response length and scope

Model countflat

+5 for 2 models, +10 for 3+

Search modeflat

News −10, deep research +5

72–99 → High50–71 → Medium0–49 → Low

Roadmap

V1 ✓ Live

Multi-LLM Consensus Search

5 models · 5 search providers · deterministic scoring · WhatsApp Task Agent · Daily Brief

V2 ✓ Live

Consensus Intelligence

Response ranking A–D · citation verification · hallucination detection · query planner · cost estimation

Pre-2b ⟳ Building

Meta Business Verification

Register as sole proprietor · Meta WhatsApp Business approval · dedicated phone number

V2b ⟳ Building

WhatsApp Task Agent MVP

Natural-language reminders · task management · daily news brief · conversation context · duplicate protection

V3 ⟳ Building

Knowledge & Personalisation

User memory · RAG document search · Trend Intelligence · cross-device sync · citation graph

V4 ⟳ Building

AI Orchestration Engine

Cost routing R(m,q) = Cost × Latency × HallucinationRisk / Quality · Streaming SSE · Smart model routing · Self-learning EMA registry · Patent: 15 claims, 5 inventions

V5 ◇ Future

Stock & Financial Intelligence

80+ Indian stocks · RSI/MACD/MA technicals · BULL/NEUTRAL/BEAR consensus · Reddit + Google Trends sentiment · NSE/BSE tickers

V5b ◇ Future

Precious Metals Intelligence

Gold & Silver · Technical + Macro + Prediction engines · TradeRiskScore · Entry/Stop/Target setup · Signal history

V6 ◇ Future

SumQ News

Story clustering · multi-source deduplication · FACT/CLAIM/OPINION classification · Perspective Check · 3-model analysis · Evidence confidence

V7 ◇ Future

Portfolio Intelligence

CAS import (CAMS/KFintech) · AMFI live NAV · P&L calculation · 5-model AI analyst · broker connection architecture · AddInvestments flow

V8 ⟳ Building

Broker Connections

OAuth2 broker integration · Zerodha (pending multi-user approval) · Upstox · Angel One · Account Aggregator (SEBI AA framework) · encrypted token storage

V9 ○ Planned

AI Action Platform

n8n workflow automation · MCP connectors — Calendar · Gmail · Notion · WhatsApp action triggers

V10 ◇ Future

Enterprise AI Operating System

Multi-tenant deployment · AI governance · SumQ Knowledge Network · White-label

Built by

S

Saket

Founder · Product Builder · AI/ML Enthusiast · Bengaluru, India

Product management leader turned full-stack AI product builder. Building SumQ from zero — product strategy, architecture, UI/UX, and engineering — to make AI answers trustworthy at scale.

The goal: not to pick the best AI model, but to make all models check each other. One SumQ brain. Multiple interfaces.

Get in touch

Open to partnerships, enterprise pilots, investor conversations, and feedback.

Try SumQ →

sumq.in · Built in India 🇮🇳

Home·Roadmap·Privacy·saket.s0308@gmail.com